1. Neuroscience
Download icon

An excitatory lateral hypothalamic circuit orchestrating pain behaviors in mice

  1. Justin N Siemian
  2. Miguel A Arenivar
  3. Sarah Sarsfield
  4. Cara B Borja
  5. Lydia J Erbaugh
  6. Andrew L Eagle
  7. Alfred J Robison
  8. Gina Leinninger
  9. Yeka Aponte  Is a corresponding author
  1. Neuronal Circuits and Behavior Unit, National Institute on Drug Abuse Intramural Research Program, National Institutes of Health, United States
  2. Department of Physiology, Michigan State University, United States
  3. Institute for Integrative Toxicology at Michigan State University, United States
  4. The Solomon H. Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, United States
Research Article
  • Cited 0
  • Views 864
  • Annotations
Cite this article as: eLife 2021;10:e66446 doi: 10.7554/eLife.66446

Abstract

Understanding how neuronal circuits control nociceptive processing will advance the search for novel analgesics. We use functional imaging to demonstrate that lateral hypothalamic parvalbumin-positive (LHPV) glutamatergic neurons respond to acute thermal stimuli and a persistent inflammatory irritant. Moreover, their chemogenetic modulation alters both pain-related behavioral adaptations and the unpleasantness of a noxious stimulus. In two models of persistent pain, optogenetic activation of LHPV neurons or their ventrolateral periaqueductal gray area (vlPAG) axonal projections attenuates nociception, and neuroanatomical tracing reveals that LHPV neurons preferentially target glutamatergic over GABAergic neurons in the vlPAG. By contrast, LHPV projections to the lateral habenula regulate aversion but not nociception. Finally, we find that LHPV activation evokes additive to synergistic antinociceptive interactions with morphine and restores morphine antinociception following the development of morphine tolerance. Our findings identify LHPV neurons as a lateral hypothalamic cell type involved in nociception and demonstrate their potential as a target for analgesia.

Introduction

Responding appropriately to environmental stimuli is vital to an organism’s survival. Nociception facilitates survival via the detection of dangerous environmental stimuli, which organisms use to escape and avoid these threats (Bolles and Fanselow, 1980; Tovote et al., 2015). However, maladaptive processes following injury or infection can cause the transition to chronic pain, a clinical condition with great economic burden that is not well-addressed by current therapeutics (Price et al., 2018; Grace et al., 2014). The widespread failure of preclinical pain therapies to translate to the clinic may be due to the historical focus on studying acute, pain-stimulated nocifensive behaviors in naive animals such as paw withdrawal to heat, which are not maladaptive and necessitate the examination of off-target effects like sedation in separate assays (Negus et al., 2015). Rather than physical sensitization to painful stimuli, the more problematic components of chronic pain in humans are likely the loss of ability to perform standard daily life activities and development of comorbid depression (Asmundson and Katz, 2009; Cleeland and Ryan, 1994; Dworkin et al., 2005; Elman et al., 2013; Negus et al., 2006). As such, rodent studies searching for new analgesics have begun to investigate ethological behaviors like nesting that are suppressed by noxious stimulation (e.g., forgoing standard life activities) as well as the affective/emotional component of nociception with assays of noxious stimulus-induced aversion (e.g., comorbid depression) (Negus et al., 2015; Johansen et al., 2001; Corder et al., 2019). Identifying specific brain pathways capable of managing these multiple components of chronic pain behavior and developing strategies for targeting them for translational use will advance the search for novel analgesics.

Decades ago, the lateral hypothalamus (LH) was identified as a brain region responsive to noxious stimuli that is capable of controlling pain-related behavioral responses and modulating neuronal activity in the periaqueductal gray area (PAG) (Cox and Valenstein, 1965; Lopez et al., 1991; Dafny et al., 1996; Fuchs and Melzack, 1995; Behbehani et al., 1988). Pharmacological experiments have implicated various neurotransmitters and receptors in the regulation of nociception by the LH-PAG pathway, including α1- and α2-adrenoceptors, cannabinoid 1 receptors (CNR1), hypocretin 2 receptors (HCRT2), tachykinin 1 receptors (TACR1; neurokinin 1 [NK1] receptor), and substance P (Esmaeili et al., 2017; Holden et al., 2009; Holden et al., 2002; Holden and Naleway, 2001). However, characterizing the specific LH cell types associated with nociception or other behavioral processes has only recently been enabled by modern neurobiological approaches.

While the LH circuits controlling food intake and reward have received intense focus over the past several years (Jennings et al., 2015; Jennings et al., 2013; Qualls-Creekmore et al., 2017; Navarro et al., 2016; Barbano et al., 2016; Nieh et al., 2016), those governing nociception have been understudied by comparison. Thus, with its diverse array of neuronal populations (Mickelsen et al., 2019), uncovering genetically defined LH circuits that regulate pain behavior may bring forth novel therapeutic targets. We previously described a small population of fast-spiking glutamatergic LH neurons expressing parvalbumin (LHPV neurons) that forms functional excitatory synapses in the ventrolateral periaqueductal gray area (vlPAG) and regulates acute thermal and chemical nociception (Siemian et al., 2019; Kisner et al., 2018). However, the broader therapeutic potential of LHPV neurons and their specific targets within the vlPAG have not yet been fully assessed.

Using in vivo calcium imaging, we demonstrate that LHPV neurons exhibit time-locked responses to acute hot or cold stimuli as well as increased activity following the administration of a persistent inflammatory irritant. Additionally, we show that chemogenetic modulation of LHPV neurons alters not only reflexive nociceptive behaviors over a timescale of hours but also restores noxious stimulus-suppressed behavior and ameliorates noxious stimulus-associated negative affect. In models of persistent inflammatory or neuropathic pain, optogenetic activation of LHPV neurons or their axonal projections in the vlPAG attenuates nociception. Furthermore, neuroanatomical tracing using modified rabies virus revealed that LHPV neurons preferentially target nociception-suppressing glutamatergic neurons over nociception-facilitating GABAergic neurons in the vlPAG. Interestingly, we observed that activation of an LHPV neuron pathway to the lateral habenula (LHb) can mediate aversion-like behavior but not nociception, suggesting pathway-specific behavioral effects of these neurons. Finally, we report that LHPV neuronal activation evokes additive to synergistic antinociceptive interactions with morphine and restores morphine antinociception following the development of morphine tolerance. Our findings identify LHPV neurons as a lateral hypothalamic cell type intricately involved in nociception and demonstrate their potential as a novel target for analgesic treatment or for use in combination therapies with current analgesics.

Results

In vivo functional imaging of LHPV neurons

LHPV neurons bidirectionally modulate responses to acute noxious stimuli (Siemian et al., 2019), but their activity in response to noxious stimuli in vivo has not yet been studied. To investigate this, we used the combination of in vivo endomicroscopy with a genetically encoded calcium indicator (GCaMP) to measure intensity fluctuations of calcium-sensitive fluorophores as an indicator of neuronal activity in LHPV cells during behavior. First, we expressed a green fluorescent calcium indicator in LHPV neurons by injecting a Cre recombinase-dependent viral vector driving the expression of GCaMP6s (Chen et al., 2013) in the LH of PvalbCre transgenic mice (Hippenmeyer et al., 2005). For detection of GCaMP6s fluorescence, we implanted a GRIN lens above the LHPV nucleus and interfaced the lens with a detachable miniscope (Figure 1a, b). In conjunction with established and open-source computational algorithms for data processing (Friedrich et al., 2017; Zhou et al., 2018; Pnevmatikakis and Giovannucci, 2017), we were able to visualize (Figure 1c) and extrapolate calcium (Ca2+) traces from individual LHPV neurons over periods of behavioral testing (Figure 1d).

Figure 1 with 2 supplements see all
In vivo functional imaging of LHPV neurons.

(a) Schematic configuration for deep-brain functional imaging from LHPV neurons in freely moving mice. Permission to publish miniscope drawing granted by Doric Lenses Inc. (b) Top: representative GRIN lens placement for functional imaging of LHPV neurons. Scale bar: 500 µm. Bottom: depiction of GRIN lens above GCaMP6s-expressing LHPV neurons. Scale bar: 200 µm. (c) Top: sample background-subtracted frame from a recording session. Bottom: spatial footprints of extracted neural segments. Scale bar: 100 μm. (d) Representative filtered traces from individual LHPV neurons. Dotted lines represent contacts with hot plate. (e) Z-scored Ca2+ traces of LHPV neurons (87 neurons, three mice) averaged across exposures to a 51°C hot plate or a room temperature control surface. Dotted line represents contact with plate or control surface. (f) Clustering of 87 units by mean max amplitude and mean area under the curve (AUC) following hot plate surface contact. Dotted lines indicate the thresholds for inclusion into cluster 1 (mean max amplitude ≥ 1 and mean AUC ≥ 0) or cluster 2 (mean max amplitude ≤ 1 and mean AUC ≤ 0). (g) Neurons in cluster 1 (n = 35/87) displayed time-locked increases in activity in response to the hot plate as compared to the control room temperature surface. Two-way repeated-measures ANOVA on average Z-score per second revealed a significant time × stimulus interaction (F(29, 986) = 10.47, p<0.0001). Bonferroni multiple comparisons post-test significant between-stimulus differences are represented by the bolded red line. Red and gray shaded areas represent s.e.m. Tan shaded region represents average contact time with hot plate stimulus. (h) Neurons in cluster 2 (n = 16/87) displayed average decreases in activity in response to the hot plate as compared to the control room temperature surface. Two-way repeated-measures ANOVA on average Z-score per second revealed a significant time × stimulus interaction (F(29, 435) = 7.61, p<0.0001). Bonferroni multiple comparisons post-test significant between-stimulus differences are represented by the bolded red line. Red and gray shaded areas represent s.e.m. Tan shaded region represents average contact time with hot plate stimulus. (i) Z-scored Ca2+ traces of LHPV neurons (53 neurons, three mice) averaged across exposures to a 4°C cold plate or a room temperature control surface. Dotted line represents contact with plate or control surface. (j) Clustering of 53 units by mean max amplitude and mean AUC following cold plate surface contact. Dotted lines indicate the thresholds for inclusion into cluster 1 (mean max amplitude ≥ 1 and mean AUC ≥ 0) or cluster 2 (mean max amplitude ≤ 1 and mean AUC ≤ 0). (k) Neurons in cluster 1 (n = 15/53) displayed time-locked increases in activity in response to the cold plate as compared to the control room temperature surface. Two-way repeated-measures ANOVA on average Z-score per second revealed a significant time × stimulus interaction (F(29, 406) = 5.94, p<0.0001). Bonferroni multiple comparisons post-test significant between-stimulus differences are represented by the bolded blue line. Blue and gray shaded areas represent s.e.m. Tan shaded region represents average contact time with cold plate stimulus. (l) Neurons in cluster 2 (n = 11/53, top) displayed average decreases in activity in response to the hot plate as compared to the control room temperature surface. Two-way repeated-measures ANOVA on average Z-score per second revealed a significant time × stimulus interaction (F(29, 290) = 2.05, p=0.0016). Bonferroni multiple comparisons post-test significant between-stimulus differences are represented by the bolded blue line. Blue and gray shaded areas represent s.e.m. Tan shaded region represents average contact time with cold plate stimulus. (m) Illustration of fluorescent trace deconvolution to estimated periods of neuronal firing. (n) Average deconvolved events per 5 min period following no injection (n = 46 neurons) or formalin injection in the hindpaw ipsilateral (n = 67 neurons) or contralateral (n = 51 neurons) to the brain hemisphere implanted with a GRIN lens. (o–q) Formalin induced fluctuations in LHPV neuronal activity in each phase of the formalin test. Mann–Whitney U-tests with Holm–Sidak correction for multiple comparisons revealed significantly higher Ca2+ event frequency following contralateral formalin injection in the (o) acute (p=0.048), (p) interphase (p=0.0078), and (q) inflammatory phases (p=0.048) relative to no injection, whereas no significant differences were found between ipsilateral formalin and no injection (acute p=0.80, interphase p=0.18, inflammatory p=0.86). Lines and error bars indicate mean ±95% CI. See also Figure 1—figure supplements 1 and 2

Figure 1—source data 1

LHPV neuronal responses to acute thermal stimuli.

https://cdn.elifesciences.org/articles/66446/elife-66446-fig1-data1-v1.xlsx
Figure 1—source data 2

LHPV calcium transient frequency during formalin tests.

https://cdn.elifesciences.org/articles/66446/elife-66446-fig1-data2-v1.xlsx

We first monitored Ca2+ dynamics in LHPV neurons (n = 87 neurons, three mice) in response to an acute thermal hot plate stimulus and clustered the neurons according to their response properties (Figure 1d−f). In a subset of the recorded LHPV neurons (‘Cluster 1,’ n = 35/87 neurons), we observed time-locked increases in fluorescence in response to the 51°C hot plate relative to a room-temperature innocuous stimulus of similar visual and tactile properties, suggesting that this subpopulation of LHPV neurons becomes active in response to a thermal stimulus (Figure 1g). Another subset of neurons (‘Cluster 2,’ n = 16/87 neurons) exhibited an average decrease in activity in response to the hot plate relative to control stimulus (Figure 1h). We observed a similar profile of time-locked responses to a 4°C cold stimulus relative to a control innocuous stimulus (Figure 1i, j). One subset of the recorded neurons (‘Cluster 1,’ n = 15/53 neurons) was significantly activated in response to the cold plate relative to control stimulus (Figure 1k), while another subset (‘Cluster 2,’ n = 11/53 neurons) displayed significantly lower activity following the cold plate stimulus as compared to the control stimulus (Figure 1l). Within each cluster, we trained a support vector machine (SVM) classifier using averaged 10 s traces following contact with the noxious (hot/cold) or neutral surfaces and tested whether it could predict the stimulus type when given unlabeled traces. Remarkably, neuronal activity from each cluster except cluster 2 from the cold plate test could decode the correct stimulus type above chance levels (Figure 1—figure supplement 1). Cluster 1 and 2 neurons were observed in each of the mice tested. Thus, we registered cells across the hot plate and cold plate sessions to examine whether LHPV neurons exhibited consistent response profiles across tests (Sheintuch et al., 2017). Of the 33 total neurons that were detected in both sessions, only 6 remained in the same cluster, and the area under the curves of the fluorescent traces of all 33 neurons did not significantly correlate between sessions, suggesting that the responses of LHPV neurons were generally variable across testing (Figure 1—figure supplement 2). Together, these results demonstrate that LHPV neuronal activity is modulated in response to acute thermal stimuli.

We next tracked LHPV neuronal activity over a longer timescale in response to a hindpaw injection of the chemical irritant formalin. Formalin induces discrete acute (0–5 min) and inflammatory (15–45 min) phases of pain behavior, separated by a brief interphase period (Alhadeff et al., 2018; Dubuisson and Dennis, 1977), allowing us to monitor changes in neuronal activity during each phase. Relative to recording sessions without formalin injection, we observed that the frequency of deconvolved Ca2+ transients (Figure 1m) appeared to be generally higher following formalin injections in the hindpaw contralateral to the brain hemisphere in which the GRIN lens was implanted (Figure 1n), and statistical analyses of Ca2+ event frequency within each period supported this observation (Figure 1o−q). Cell registration revealed that only three neurons were detected in more than three of these imaging sessions, thus we could not examine whether a neuron being in cluster 1 or 2 in the hot plate and cold plate tests impacted its response properties in the formalin tests (Figure 1—figure supplement 2). Together, these findings indicate that LHPV neurons display changes in spontaneous activity in response to several stimulus modalities, including both acute thermal stimuli and ongoing chemical inflammation.

LHPV neurons regulate sensory and affective aspects of pain over long timescales

We next examined whether manipulating LHPV neuronal activity can alter noxious stimulation-suppressed behavior and negative affect, which may be better indicators of clinical utility than stimulus-evoked behaviors (e.g., reflexive withdrawal to acute thermal stimuli). To investigate this, we targeted these neurons for chemogenetic manipulations by bilaterally injecting Cre recombinase-dependent viral vectors driving the expression of either the excitatory designer receptor hM3D, the inhibitory designer receptor hM4D, or the fluorophore mCherry as control into the LH of PvalbCre transgenic mice (Figure 2a). Activation of the designer receptors via administration of the ligand clozapine-N-oxide (CNO, 1 mg/kg, i.p.) evoked significant increases and decreases in PWLHP in LHPV:hM3D and LHPV:hM4D mice, respectively, as compared to mCherry controls, with effects detectable between 1 and 18 hr post-injection (Figure 2b). Thus, chemogenetic manipulations of these neurons alter nociception over a long timescale.

Chemogenetic modulation of LHPV neurons regulates pain-suppressed behavior and alters pain-associated negative affect.

(a) Representative images of hM3D, hM4D, or mCherry expression in LHPV neurons. Scale bars: 500 μm, widefield; 50 μm, zoom. (b) Chemogenetic activation and inhibition of LHPV neurons evoked long-lasting significant increases and decreases in thermal pain thresholds, respectively (n = 11 mice per group; two-way mixed-model ANOVA group × time interaction, F(16, 240)=14.15, p<0.0001). Significant differences from LHPV:mCherry mice were determined by Bonferroni multiple comparisons tests and are represented graphically, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. (c) Schematic for pain-suppressed nesting assay. (d) Chemogenetic activation of LHPV neurons prevented the reductions in nesting behavior induced by i.p. injection of 0.6% acetic acid (10 ml/kg). Two-way mixed-model ANOVA revealed a significant group × test interaction (n = 11 mice per group; F(4, 60) = 4.17, p=0.0048). Bonferroni multiple comparisons post-tests revealed no differences in normal nesting behavior from clozapine-N-oxide (CNO) injections (p=0.92), and that acetic acid injection decreased nesting behavior across groups when administered without CNO (p<0.0001). Administration of CNO before acetic acid increased nesting behavior in LHPV:hM3D mice relative to LHPV:mCherry control mice (p=0.008; Cohen’s d = 1.09) and tests without CNO (p=0.0002). (e) Schematic of the formalin place conditioning experiment. (f) Chemogenetic modulation of LHPV neurons altered the effects of formalin on place conditioning. Two-way mixed-model ANOVA revealed a significant group × test interaction (n = 11 mice per group; F(2,28) = 3.89, p=0.032). Bonferroni multiple comparisons post-tests showed significant shift in chamber preference in LHPV:mCherry (p=0.0001) and LHPV:hM4D mice (p<0.0001) but not LHPV:hM3D mice (p=0.094). (g) Time spent paw licking was altered by LHPV neuronal modulation (n = 10 LHPV:mCherry, 7 LHPV:hM3D, and 10 LHPV:hM4D mice; two-way mixed-model ANOVA group × time interaction, F(22, 264) = 1.99, p=0.0064). (h) Acute and inflammatory phase paw licking were differentially altered by LHPV neuronal activation and inhibition (n listed above; two-way mixed-model ANOVA group × phase interaction, F(2, 24) = 4.33, p=0.025). Activation of LHPV neurons in LHPV:hM3D mice decreased acute (p=0.0004, Cohen’s d = 2.07) but not inflammatory phase paw licking (p=0.50), whereas LHPV neuronal inhibition in LHPV:hM4D mice increased inflammatory (p=0.0049, Cohen’s d = 1.29) but not acute phase paw licking (p>0.99).

Pain in basic research is traditionally assessed by measuring ‘pain-stimulated behavior’ or the elicited reactions to noxious stimuli (e.g., paw withdrawal). However, clinical pain disorders often impact quality of life more profoundly by deterring actions normally performed when healthy (Cleeland and Ryan, 1994; Dworkin et al., 2005). Therefore, we next examined the effects of LHPV neuronal activity in a model of ‘pain-suppressed behavior,’ which measures a decrease in behavioral output following a noxious stimulus (Negus et al., 2006). Healthy mice normally collect nestlet pieces distributed throughout the home cage within 30 min and begin nest building, a natural behavior (Negus et al., 2015; Diester et al., 2021a; Diester et al., 2021b; Figure 2c); this was not affected by CNO administration across groups (Figure 2d, ‘CNO + saline’). However, administration of acetic acid (0.6%; i.p.) significantly decreases nesting behavior; this was apparent in all three groups without LHPV manipulations (Figure 2d, ‘Vehicle + acid’). Interestingly, activation of LHPV neurons in LHPV:hM3D mice prior to acetic acid injection prevented the reductions in nesting behavior (Figure 2d, ‘CNO + acid’) as compared to tests without CNO and to LHPV:Ctrl mice. In contrast, no changes were observed in LHPV:hM4D mice. Thus, LHPV activation not only decreases noxious stimulus-evoked behavior but also restores behaviors normally suppressed by noxious stimulation.

Pain results not only in overt behavioral changes but also negative affect, as made evident by the high comorbidity between pain and mood disorders (Asmundson and Katz, 2009; Elman et al., 2013). We sought to determine the role of LHPV neuronal activity on the affective, or emotional, component of a painful experience. For this, we used a place conditioning paradigm in which mice avoid a context paired with an aversive event (Johansen et al., 2001; Alhadeff et al., 2018; Figure 2e). After assessment of initial side preference of a two-chamber apparatus, we passively conditioned the mice by administering CNO (i.p.) with intra-plantar formalin to induce inflammation in the initially preferred side and CNO with intra-plantar saline in the initially less-preferred side. Mice were conditioned twice in each context on alternating days and then were given free access to both chambers during a post-test to assess changes in place preference. As expected, LHPV:Ctrl mice lost preference to the formalin-paired context as compared to pre-formalin preference levels (Figure 2f). However, activation of LHPV neurons during conditioning attenuated this loss of place preference, whereas inhibition of LHPV neurons during conditioning permitted the loss of place preference (Figure 2f). Furthermore, the time spent paw licking in these sessions was bidirectionally affected by chemogenetic LHPV neuronal activation or inhibition (Figure 2g, h). LHPV neuronal activation decreased paw licking during the acute but not inflammatory phase, whereas inhibition increased paw licking in the inflammatory but not acute phase. Together, these results support a role for LHPV neurons both in pain behaviors and associated negative affect.

Optogenetic activation of LHPV neurons attenuates persistent inflammatory pain-associated behaviors

Since LHPV neurons ameliorated moderately long-lasting behavioral effects of pain, we next sought to determine whether they could also alter nociceptive thresholds in traditional models of persistent pain behavior. We targeted LHPV neurons for optogenetic manipulations with bilateral injections of a Cre recombinase-dependent viral vector driving the expression of either channelrhodopsin (ChR2:tdTomato; light-sensitive neuronal activator) or GFP (control fluorophore) in the LH of PvalbCre transgenic mice and implanted optical fibers bilaterally above these neurons (Figure 3a). Activation of LHPV neurons in naive mice significantly increased paw withdrawal latency in response to a 51°C hot plate (PWLHP, Figure 3b). However, activating these neurons did not change paw withdrawal threshold in the von Frey filament test (PWTVF, Figure 3c), suggesting that these neurons regulate acute thermal but not mechanical nociception in healthy mice. Since LHPV neurons are glutamatergic (Siemian et al., 2019; Kisner et al., 2018) and activation of LH neurons expressing the vesicular glutamate transporter 2 (SLC17A6; LHVGLUT2) is aversive (Jennings et al., 2013), we also assessed the effects of LHPV neuronal activation in a real-time place preference (RTPP) assay in which photostimulation was paired with one-half of the behavioral arena. Activation of LHPV neurons was mildly aversive as mice spent significantly less time on the photostimulation-paired side (Figure 3d), suggesting that these neurons may play a role in reward- and aversion-like behaviors. Next, we injected complete Freund’s adjuvant (CFA), a well-known inflammatory reagent (Alhadeff et al., 2018; Fehrenbacher et al., 2012; Nagakura et al., 2003), into the right hindpaw to cause inflammation and induce persistent hypersensitivity. We observed a significant decrease in nociceptive thresholds for both thermal and mechanical stimuli following these CFA injections (Figure 3—figure supplement 1a, b). Interestingly, activation of LHPV neurons after CFA evoked significant increases in both PWLHP and PWTVF (Figure 3e, f). Additionally, activation of LHPV neurons no longer triggered place avoidance (Figure 3g). Furthermore, we observed that the magnitude of the PWLHP response depends on the photostimulus frequency (Figure 3—figure supplement 1c) and that LHPV neuron-mediated antinociception was not strictly photostimulus-bound (Figure 3—figure supplement 1d) as the antinociceptive effects persisted for several minutes after photostimulation ceased. Together, these results indicate that LHPV neuronal activation attenuates hypersensitivity to both thermal and mechanical stimuli following the onset of inflammation.

Figure 3 with 1 supplement see all
Optogenetic activation of LHPV neurons attenuates thermal and mechanical nociception following the induction of inflammatory pain.

(a) Representative images of ChR2 or GFP expression in LHPV neurons and optical fiber implants above the lateral hypothalamus (LH). Scale bars: 500 μm, widefield; 50 μm, zoom. (b) Optogenetic activation of LHPV neurons in naive mice triggers thermal antinociception (n = 9 ChR2 mice and 10 Ctrl mice). Two-way mixed-model ANOVA revealed a significant group × epoch interaction (F(2, 34) = 14.01, p<0.0001), and Bonferroni multiple comparisons post-test showed that LHPV:ChR2 mice had significantly higher PWLHP during the photostimulation epoch than LHPV:Ctrl mice, p<0.0001; Cohen’s d = 2.22. (c) Optogenetic activation of LHPV neurons in naive mice does not affect mechanical nociception (n = 9 ChR2 mice and 10 Ctrl mice). Two-way mixed-model ANOVA interaction, p=0.87. (d) Naive LHPV:ChR2 mice displayed significant real-time place avoidance to photostimulation relative to controls (n = 9 ChR2 mice and 10 Ctrl mice, t(17) = 3.15, p=0.0058, Cohen’s d = 1.43). (e) Optogenetic activation of LHPV neurons in mice 5 days following complete Freund’s adjuvant (CFA) injection triggers increases in PWLHP (n = 9 ChR2 mice and 10 Ctrl mice). Two-way mixed-model ANOVA revealed a significant group × epoch interaction (F(2, 34) = 15.05, p<0.0001), and Bonferroni multiple comparisons post-test showed that LHPV:ChR2 mice had significantly higher PWLHP during the photostimulation epoch than LHPV:Ctrl mice (p=0.0001; Cohen’s d = 2.08). (f) Optogenetic activation of LHPV neurons in mice 6 days following CFA injection triggers increases in PWTVF (n = 9 ChR2 mice and 10 Ctrl mice). Two-way mixed-model ANOVA revealed a significant group × epoch interaction (F(2, 34) = 11.28, p=0.0002), and Bonferroni multiple comparisons post-test showed that LHPV:ChR2 mice had significantly higher PWLHP during the photostimulation epoch than LHPV:Ctrl mice (p=0.003; Cohen’s d = 1.11). (g) LHPV:ChR2 mice did not display significant real-time place avoidance to photostimulation relative to controls 7 days post-CFA, p=0.75 (n = 9 ChR2 mice and 10 Ctrl mice). See also Figure 3—figure supplement 1.

LHPV neurons target excitatory circuits within the vlPAG to regulate pain behaviors

LHPV neurons send dense projections to the vlPAG (Kisner et al., 2018; Celio et al., 2013), where they form functional excitatory synapses. We next examined whether this LHPV→vlPAG pathway also regulates nociception in models of persistent pain behavior. To specifically target and manipulate the LHPV→vlPAG pathway, we bilaterally injected a Cre recombinase-dependent viral vector driving the expression of channelrhodopsin (ChR2:tdTomato), the light-sensitive neuronal silencer archaerhodopsin (ArchT:GFP), or the fluorophore GFP (control) into the LH of PvalbCre mice and implanted optical fibers bilaterally above the vlPAG to specifically manipulate the axonal projections of LHPV neurons (Figure 4a). In naive mice, activation of the LHPV→vlPAG pathway evoked increases in PWLHP but not PWTVF, similar to somatic manipulations (Figure 4b, c), whereas inhibition of the pathway decreased both nociceptive thresholds (Figure 4d, e). However, in contrast to somatic manipulations, no effects were observed for either LHPV→vlPAG activation or inhibition in the RTPP test (Figure 4f), suggesting that there were no changes in the overall affective state of the mice that may have contributed to these bidirectional effects on nociception. In healthy mice, we also observed that the magnitude of the PWLHP response during activation of the LHPV→vlPAG pathway depends on photostimulus frequency. Moreover, these responses were not affected by systemic administration of the cannabinoid receptor 1 (CNR1 or CB1) antagonist/inverse agonist rimonabant (3 mg/kg, i.p.; Figure 4g) despite the PAG being an important site for cannabinoid-mediated antinociception (Esmaeili et al., 2017; Finn et al., 2003; Maione et al., 2006). These results suggest that blocking CB1 receptors does not affect antinociception driven by LHPV→vlPAG circuitry.

Figure 4 with 3 supplements see all
LHPV→vlPAG pathway mediates nociception in models of chronic neuropathic and inflammatory pain.

(a) Representative images of ChR2, ArchT, or GFP expression in LHPV neurons and optical fiber implants above the ventrolateral periaqueductal gray area (vlPAG). Inset shows axons from LHPV neurons under the optical fiber. Scale bars: 500 μm, widefield; 50 μm, zoom; 100 μm, inset. (b) In naive mice, optogenetic activation of LHPV axonal projections in the vlPAG evokes thermal antinociception (n = 7 mice per group; two-way mixed-model ANOVA group × epoch interaction, F(2, 24) = 25.19, p<0.0001, Bonferroni multiple comparisons post-test during photostimulation epoch, p<0.0001; Cohen’s d = 2.85) but not (c) mechanical antinociception (p=0.38). (d) In naive mice, optogenetic inhibition of LHPV axonal projections in the vlPAG decreases both thermal (n = 7 mice per group; two-way mixed-model ANOVA group × epoch interaction, F(2, 24) = 11.96, p=0.0002, Bonferroni multiple comparisons post-test during photostimulation epoch, p<0.0001; Cohen’s d = 2.13) and (e) mechanical thresholds (two-way mixed-model ANOVA group × epoch interaction, F(2, 24) = 12.10, p<0.0002, Bonferroni multiple comparisons post-test during photostimulation epoch, p=0.038; Cohen’s d = 1.36). (f) Optogenetic activation or inhibition of the LHPV→vlPAG pathway did not affect real-time place preference behavior in naive mice (n = 7 mice per group; one-way ANOVA, F(2, 18) = 0.28, p=0.76). (g) LHPV→vlPAG activation-induced antinociception is dependent on photostimulus frequency but is not attenuated by the CB1 receptor antagonist rimonabant (3 mg/kg, i.p.; ‘3 RIM’). Two-way mixed-model ANOVA revealed a significant group × epoch interaction (n = 6 ChR2 mice and 7 Ctrl mice; F(7, 77) = 14.27, p<0.0001). Bonferroni multiple comparisons post-tests revealed between-group differences during the ‘50 Hz’ and ‘50 Hz + 3 RIM’ epochs (p<0.0001), but no within-group differences between these epochs (p>0.99). (h) Optogenetic activation of the LHPV→vlPAG pathway evokes increases in PWLHP on day 5 post-spared nerve injury (SNI) (n = 7 mice per group; two-way mixed-model ANOVA group × epoch interaction, F(2, 24) = 12.86, p<0.0001, Bonferroni multiple comparisons post-test during photostimulation epoch, p=0.0002; Cohen’s d = 2.04) and (i) PWTVF on day 6 post-SNI (two-way mixed-model ANOVA group × epoch interaction, F(2, 24) = 5.24, p<0.013, Bonferroni multiple comparisons post-test during photostimulation epoch, p=0.019; Cohen’s d = 1.03). (j) On day 7 post-SNI, optogenetic activation of the LHPV→vlPAG pathway did not affect real-time place preference behavior (n = 7 mice per group; p=0.39). (k) In a new cohort of naive mice, optogenetic activation of the LHPV→vlPAG pathway evoked thermal (n = 10 mice per group; two-way mixed-model ANOVA group × epoch interaction, F(2, 36) = 23.64, p<0.0001, Bonferroni multiple comparisons post-test during photostimulation epoch, p=0.0009) but not (l) mechanical antinociception (p=0.31). (m) Optogenetic activation of the LHPV→vlPAG pathway evokes increases in PWLHP on day 5 post-complete Freund’s adjuvant (CFA) injection (n = 10 mice per group; two-way mixed-model ANOVA group × epoch interaction, F(2, 36) = 19.65, p<0.0001, Bonferroni multiple comparisons post-test during photostimulation epoch, p<0.0001; Cohen’s d = 3.66) and (n) PWTVF on day 6 post-SNI (two-way mixed-model ANOVA group × epoch interaction, F(2, 36) = 24.63, p<0.0001, Bonferroni multiple comparisons post-test during photostimulation epoch, p<0.0001; Cohen’s d = 1.88). (o) On day 7 post-CFA injection, optogenetic activation of the LHPV→vlPAG pathway did not affect real-time place preference behavior (n = 9 mice per group; p=0.59). See also Figure 4—figure supplements 13.

We next investigated the effects of activating the LHPV→vlPAG pathway in the spared nerve injury (SNI) model of neuropathy. Five days post-SNI, we observed significant decreases in thermal and mechanical thresholds (Figure 4—figure supplement 1a, b). Photostimulation of the LHPV→vlPAG pathway evoked increases in both PWLHP and PWTVF (Figure 4h, i), and this remained during testing 25 days post-SNI (Figure 4—figure supplement 1c, d). Furthermore, no effects of LHPV→vlPAG pathway activation were observed in the RTPP test post-SNI (Figure 4j). Due to the modest effects of LHPV→vlPAG activation on mechanical thresholds, we predicted that this pathway may be more effective during inflammatory than neuropathic conditions. Therefore, in a new cohort of mice, we activated the LHPV→vlPAG pathway before (Figure 4k, l) and after (Figure 4m, n) the induction of inflammation by CFA (Figure 4—figure supplement 2a, b) and observed robust effects on both thermal and mechanical nociceptive thresholds. Similar to the SNI cohort, LHPV→vlPAG activation post-CFA did not affect RTPP, suggesting that there were no effects on reward- or aversion-related behaviors (Figure 4o). Furthermore, LHPV→vlPAG pathway-mediated antinociception post-CFA was dependent on photostimulus frequency but not strictly photostimulus-bound (Figure 4—figure supplement 2c, d). Together, these results show that the LHPV→vlPAG pathway regulates nociception in at least two models of persistent pain behavior and that its activation is more effective in attenuating inflammatory than neuropathic hypersensitivity.

Within the vlPAG, GABAergic and glutamatergic neurons play opposing roles in regulating nociception and defensive behavior (Samineni et al., 2017; Tovote et al., 2016). Although we previously showed that LHPV neurons form functional excitatory synapses with vlPAG neurons (Siemian et al., 2019), the identity of these post-synaptic targets remains unknown. Thus, we used a monosynaptic retrograde viral tracing strategy with a modified rabies virus (Wickersham et al., 2007a; Wickersham et al., 2007b) to identify the targets of LHPV neurons in the vlPAG. In Slc17a6Cre and Slc32a1Cre mice (Vong et al., 2011), we injected starter cells in the vlPAG with Cre recombinase-dependent helper virus containing rabies glycoprotein G and the EnvA receptor for avian sarcoma leukosis virus (TVA) to express the proteins required for uptake and monosynaptic propagation of modified rabies virus (Figure 5a). Three weeks later, we injected the EnvA-pseudotyped G-deleted rabies virus RVdG-mCherry(EnvA) into the vlPAG. After an additional 3 weeks, mice were perfused, and brains were processed for histological assessment. LH-containing sections were immunostained with an anti-parvalbumin antibody and imaged using confocal microscopy (Figure 5b). Quantitative analyses revealed that more vlPAGVGLUT2 neurons (14.89%; n = 63 of 423 neurons, three mice) than vlPAGVGAT neurons (6.96%; n = 22 of 316 neurons, six mice) are synaptically targeted by LHPV neurons (Chi-square = 11.18, ***p=0.0008, Figure 5c). Since activation of glutamatergic neurons in the PAG was shown previously to decrease pain (Samineni et al., 2017), our findings suggest a potential role for LHPV neurons as an excitatory input to glutamatergic vlPAG circuitry.

LHPV neurons preferentially target glutamatergic neurons in the ventrolateral periaqueductal gray area (vlPAG).

(a) Schematic for modified rabies viral tracing strategy. (b) Images from Slc17a6Cre (top row) and Slc32a1Cre (bottom row) brain slices showing the overlap of RVdG-mCherry(EnvA) with LHPV neurons. Scale bars: 20 μm. (c) Proportion of LHPV neurons that express or do not express RVdG-mCherry(EnvA) in Slc17a6Cre or Slc32a1Cre mice. LHPV neurons were connected to a greater proportion of vlPAGVGLUT2 neurons than vlPAGVGAT neurons (chi-square = 11.18, p=0.0008).

To determine whether other lateral hypothalamic circuits encode for nociception, we examined the effects of manipulating LH leptin receptor expressing (LHLEPR) neurons, which also project to the vlPAG, albeit a slightly more posterior region (Schiffino et al., 2019; Leinninger et al., 2009). In contrast to LHPV neurons, LHLEPR neurons are predominantly GABAergic and their vlPAG axonal projections are more broadly distributed than those of LHPV neurons. Cre-dependent viruses driving ChR2 or GFP expression were injected into the LH of LeprCre mice (Leshan et al., 2006; Figure 4—figure supplement 3a), and we found that activation of the LHLEPR→vlPAG pathway potentiated both thermal and mechanical nociception in healthy mice (Figure 4—figure supplement 3b−e). Moreover, activation of the LHLEPR→vlPAG pathway was rewarding as LHLEPR:ChR2→vlPAG mice spent more time on the photostimulation-paired side of the chamber than LHLEPR:Ctrl→vlPAG control mice (Figure 4—figure supplement 3f). These results demonstrate that activation of lateral hypothalamic glutamatergic (LHPV) and GABAergic (LHLEPR) populations that project to the vlPAG attenuates and potentiates nociception, respectively.

Activation of LHPV axonal projections to the LHb triggers aversion

LHPV neurons also target other brain regions including the LHb (Kisner et al., 2018; Celio et al., 2013). Therefore, we examined whether LHPV neurons also modulate nociceptive processing via projections to the LHb. For this, we bilaterally injected a Cre recombinase-dependent viral vector driving the expression of either channelrhodopsin (ChR2:tdTomato) or the fluorophores GFP or tdTomato (control) into the LH of PvalbCre mice and implanted optical fibers bilaterally above the LHb to specifically activate the LHPV→LHb pathway (Figure 6a). Interestingly, activation of this LHPV→LHb circuitry did not evoke changes in nociceptive responses to an acute noxious thermal or mechanical stimulus in healthy mice (Figure 6b−d) or in mice with SNI-induced neuropathy when tested at 5 and 25 days post-surgery (Figure 6—figure supplement 1). Because the LHb is a brain region associated with reward- and aversion-related behaviors (Stamatakis et al., 2016; Faget et al., 2018), we also sought to determine whether activation of the LHPV→LHb pathway triggers such behaviors. We found that activation of this pathway in healthy mice was aversive as mice spent significantly less time on the photostimulation-paired side in the RTPP assays (Figure 6e−g). These results are consistent with previous findings demonstrating that broad activation of lateral hypothalamic glutamatergic axonal projections in the LHb is aversive (Stamatakis et al., 2016). Together, these findings demonstrate that LHPV neurons encode for distinct behavioral outputs depending on their targeted downstream regions: nociceptive processing via projections to the vlPAG and aversion-related behaviors through connections to the LHb.

Figure 6 with 1 supplement see all
Activation of the LHPV→LHb pathway triggers aversion but not antinociception.

(a) Representative images of ChR2 and tdTomato expression in LHPV neurons and optical fiber implants above the lateral habenula (LHb). Scale bars: 500 μm, widefield; 50 μm, zoom. (b) Optogenetic activation of LHPV axonal projections in the LHb does not alter thermal nociception at 50 Hz photostimulus frequency (p=0.16) or (c) 20 Hz photostimulus frequency (p=0.23) in healthy mice (n = 13 ChR2 mice and 14 Ctrl [GFP/tdTomato] mice). (d) Optogenetic activation of LHPV axonal projections in the LHb does not alter mechanical nociception at 50 Hz (p=0.15) in healthy mice (n = 13 ChR2 mice and 14 Ctrl mice). (e) Optogenetic activation of the LHPV→LHb pathway evokes significant real-time place aversion (p=0.0041; Cohen’s d = 1.38) in a standard rectangular one-chamber testing apparatus (n = 10 ChR2 mice and 12 Ctrl mice). (f) Optogenetic activation of the LHPV→LHb pathway also evokes real-time place aversion in a three-chamber testing apparatus (n = 8 mice per group); two-way mixed-model ANOVA group × chamber interaction, F(2, 28) = 6.22, p=0.0058, Bonferroni’s multiple comparisons post-test revealed the LHPV:ChR2→LHb group spent less time in the photostimulation chamber (p=0.0089; Cohen’s d = 1.28) and more time in the no photostimulation chamber (p=0.016) than LHPV:Ctrl→LHb control mice, but no differences were observed in hall zone occupancy (p>0.99). (g) Representative heatmaps of LHPV:ChR2→LHb and LHPV:Ctrl→LHb mice in a three-chamber real-time place preference session. See also Figure 6—figure supplement 1.

Antinociceptive interactions between LHPV neuronal activation and morphine

Since activation of LHPV neurons appears to reduce nociception as monotherapy, we last sought to examine the interaction between the antinociception induced by these neurons and the μ-opioid pain reliever morphine. For this, we performed a dose-addition analysis of CNO and morphine in LHPV:hM3D and LHPV:Ctrl mice. First, we determined the individual dose-response curves of CNO and morphine using a cumulative dosing procedure. As expected, CNO evoked dose-dependent PWLHP increases in LHPV:hM3D (Figure 7a) but not LHPV:Ctrl mice (Figure 7b), whereas morphine produced dose-dependent increases in both groups (Figure 7a, b). Next, the two drugs were combined in fixed proportions (1:1, 1:3, and 3:1) according to their relative potencies (ED50) in the LHPV:hM3D group. For example, the 1:1 ratio consisted of one unit of the morphine ED50 (10.31 mg/kg) for every one unit of the CNO ED50 (0.78 mg/kg). Fractions of these mixtures (e.g., the combined 0.125 ×, 0.25 ×, 0.5 ×, and 1 × ED50 values of morphine and CNO) were administered consecutively by a cumulative dosing procedure to complete one dose-response curve test (Figure 7a, b). The shared dose-response curves were used to calculate the ED50 of each drug within each mixture; these equi-effective points were plotted on an isobologram to visualize the nature of each interaction (Figure 7c, d). For LHPV:hM3D mice, 1:3 and 1:1 morphine:CNO combinations fell within the range of additivity. Remarkably, the 3:1 morphine:CNO combination fell below the range of additivity, suggesting synergistic interactions between morphine and LHPV neuronal activation, indicating that activation of LHPV neurons enhanced the antinociceptive potency of morphine. Formal statistical comparison of expected and experimental ED50 values confirmed this observation (Student’s paired t-test: t(7) = 2.92, p=0.022). For LHPV:Ctrl mice, no combinations significantly differed from the range of additivity, suggesting that CNO did not affect the antinociceptive potency of morphine in control subjects.

Antinociceptive interactions between LHPV neuronal activation and morphine.

(a) Dose-response curves of clozapine-N-oxide (CNO) and morphine alone or in combinations of different fixed proportions in LHPV:hM3D and (b) LHPV:Ctrl mice in the hot plate test (n = 8 mice per group). (c, d) Isobolograms constructed from the data shown in panels (a) and (b). Each point represents the ED50 ± 95% CI of each drug alone or in a mixture; ordinates represent the ED50 value of morphine and abscissae represent the ED50 value of CNO. In LHPV:hM3D mice, the 3:1 morphine:CNO mixture was significantly more potent than predicted by the hypothesis of additivity (paired Student’s t-test, t(7) = 2.92, p=0.022). (e) Both groups of mice developed significant antinociceptive tolerance to 32 mg/kg morphine when administered twice per day for 3 days. Three-way mixed-model ANOVA revealed a significant morphine × test interaction (n = 8 mice per group; F(1, 14) = 134.7, p<0.0001), and Bonferroni multiple comparisons post-tests showed the antinociceptive effects of 32 mg/kg morphine were significantly lower on day 4 than day 1 (both p<0.0001). (f) Activation of LHPV neurons restored morphine potency, and further tolerance did not develop to combination treatment. Three-way mixed-model ANOVA revealed a significant treatment × group interaction (n = 8 mice per group; F(2, 28) = 42.10, p<0.0001). Bonferroni multiple comparisons post-tests revealed that there were between-group differences in PWLHP evoked on day 5 by CNO (p=0.0006) and morphine (p<0.0001) and on day 9 by CNO (p=0.016) and morphine (p<0.0001). However, no within-group differences were observed between day 5 and 9 in LHPV:hM3D mice during CNO (p>0.99) or morphine treatment (p>0.99).

Finally, using the same mice, we investigated the effects of LHPV neuronal activation following the development of morphine tolerance. We administered morphine (32 mg/kg, i.p.) twice per day for 3 days, which caused a significant decrease in morphine-induced antinociception (Figure 7e). On day 5, CNO (1 mg/kg) evoked a significant increase in PWLHP and restored morphine-induced antinociception as compared to control mice (Figure 7f). We then treated these mice once per day over the following three days with a combination of 1 mg/kg CNO and 32 mg/kg morphine to assess the potential development of tolerance to this combination. However, no differences were observed on the day 9 test in LHPV:hM3D mice as compared to day 5 (Figure 7f). Thus, activating LHPV neurons not only increases morphine potency acutely but also rescues morphine tolerance and may prevent subsequent tolerance development.

Discussion

The LH is an important site for numerous survival-critical processes such as sleep, feeding, and reward (Carter et al., 2009; Bonnavion et al., 2016; Stuber and Wise, 2016). New technologies have enabled the identification of specific lateral hypothalamic populations associated with certain behaviors and the understanding of how the activity of such neurons drives behavior and relates to external factors. However, the cell types mediating many other LH-associated behaviors have received less attention. Nociception has historically been a less LH-prototypical process than one such as feeding, but LH circuits were nevertheless previously shown to respond to noxious stimuli, to control nociception, and to affect downstream circuits in the PAG, a critical brain region for pain regulation (Cox and Valenstein, 1965; Lopez et al., 1991; Dafny et al., 1996; Fuchs and Melzack, 1995; Behbehani et al., 1988). Cell-type-specific optogenetic manipulations showed that a small cluster of fast-spiking glutamatergic LHPV neurons projects to the vlPAG and modulates acute nociception in a μ-opioid-independent manner (Siemian et al., 2019). However, much remained to be learned as to how LHPV neurons respond to noxious events and whether they could be targeted for therapies in scenarios outside of acute sensory stimulation.

One of the great challenges in understanding how dynamics in neuronal circuits control behavioral output is to determine when specific cell types are active, as well as the nature of the relationship between this activity and behavior. Although direct manipulations of neuronal activity followed by behavioral examination are important for understanding this relationship, measuring changes in the activity patterns of neurons in awake behaving mice provides information as to how this circuit functions in the absence of experimenter-driven input. Using functional imaging to measure calcium dynamics, we gained insight as to how the activity of LHPV neurons correlates with nociception and show for the first time that LHPV neurons exhibit an array of time-locked responses to acute noxious thermal events. The involvement of LHPV neurons in holding information related to noxious events is supported by the finding that the neuronal activity could be used to decode noxious from innocuous stimuli. LHPV neuronal activity was also altered during formalin-induced inflammation. Formalin injection, which causes discrete phases of acute and inflammatory forms of pain behavior, into the hindpaw contralateral to the imaged LH hemisphere evoked increases in calcium transient frequencies that were more pronounced than those observed during injection of the paw ipsilateral to the imaged LH hemisphere, likely reflecting decussation of the nociceptive signal at the spinal level (Dafny et al., 1996; Yoshida et al., 2019; Yamada et al., 2012). It is worthwhile to note the advantages of using a single-photon miniscope, which enables single-cell resolution of neuronal activity. Other methods such as fiber photometry would likely not have revealed the changes we observed in LHPV neuronal activity to acute thermal stimuli, for which there were heterogeneous responses across neurons, as well as to formalin, which was reflected as an elevated rate of calcium transients that were asynchronous across neurons. Together, our functional imaging data suggest that LHPV neurons may become active during noxious events to signal or suppress nociception in mice.

In chemogenetic experiments designed to assess the broader therapeutic potential of LHPV neuron manipulation in pain disorders, we found that these neurons can bidirectionally modulate thermal nociception over long timescales, and thus may represent potential targets for extended duration analgesia. Moreover, activation of LHPV neurons significantly attenuated acetic acid-reduced nesting behavior, demonstrating that this manipulation not only decreases sensory pain but also permits the resumption of species-specific natural behaviors that are suppressed by noxious events. This model may be analogous to clinical interventions that allow patients undergoing chronic pain to resume daily activities such as exercising or performing occupational duties as opposed to removing pain at the expense of a reduced motivational capacity. We also observed that inhibition of LHPV neuronal activity in hM4D-expressing mice did not decrease nesting behavior in control tests, suggesting that inhibition of these neurons does not cause pain directly, but likely rather enhances sensitivity to noxious stimuli. In support of this, LHPV neuronal activation reduced formalin-associated negative affective pain, whereas this was nearly enhanced by inhibiting LHPV neurons. Behavioral scoring showed that sensory pain behavior was also bidirectionally modulated in this experiment, suggesting that LHPV neurons modulate both sensory and affective experiences.

Optogenetic activation of LHPV neurons decreased both thermal and mechanical nociception following the induction of a commonly used inflammatory pain model. These effects were likely attributable to LHPV projections to the vlPAG as LHPV→vlPAG activation decreased thermal and mechanical thresholds in neuropathic and inflammatory models. In contrast, projections of LHPV neurons to the LHb regulated aversion as previously shown for the broader LH glutamatergic population (Stamatakis et al., 2016), but not nociception, suggesting that LHPV neurons regulate different behavioral outputs via different downstream projection areas. While systemic CB1 or µ-opioid antagonism does not affect LHPV→vlPAG activation-induced antinociception (Siemian et al., 2019), the finding that sustained antinociception following extended LHPV somatic or LHPV→vlPAG activation suggests that LHPV neurons may co-release neuropeptides that interact with downstream receptors to attenuate nociception or that other efferent circuits for antinociception are recruited during such activation that may function to decrease nociception. Importantly, in conjunction with the functional imaging data, the increased sensitivity observed upon inhibition of LHPV somas or the LHPV→vlPAG pathway suggests that LHPV neurons may become active in response to a noxious stimulus to decrease its severity. It is important to briefly note that while a study found that sustained activation of archaerhodopsin evokes spontaneous synaptic release in ex vivo preparations (Mahn et al., 2016), this phenomenon has not been observed during in vivo electrophysiological recordings or precluded the observation of behavioral effects in the direction associated with the loss of presynaptic input when using photoinhibition times equal to or longer than the ones we employed here (Jennings et al., 2013; Rozeske et al., 2018).

Although LHPV neurons are functionally connected to neuronal circuits within the vlPAG, the heterogeneous behavioral effects driven by the intermingled vlPAG neuronal populations made it challenging to draw a clear circuit map from LHPV neurons to behavior through the vlPAG pathway. For instance, activation of vlPAG glutamatergic and GABAergic neurons decreased and increased nociception, respectively (Samineni et al., 2017). Therefore, we used a retrograde monosynaptic rabies tracing strategy to identify the preferred post-synaptic vlPAG targets of LHPV neurons. We found a higher proportion of LHPV neurons labeled following uptake of RVdG-mCherry(EnvA) in vlPAGVGLUT2 compared to vlPAGVGAT neurons, suggesting that LHPV neurons may preferentially, yet not exclusively, target glutamatergic vlPAG neurons. In the context of previous work, excitatory input from LHPV neurons to vlPAGVGLUT2 neurons would thus form a discrete antinociceptive pathway. However, experiments using techniques such as ChR2-assisted circuit mapping (CRACM) from LHPV:ChR2+ axonal projections onto postsynaptic vlPAG neurons followed by single-cell RT-qPCR analysis will be needed to elucidate how LHPV neurons regulate vlPAG microcircuitry and how activation of this LHPV→vlPAG pathway modulates nociceptive responses to noxious stimuli. The opposing behavioral outcomes during photostimulation of a GABAergic LH population, LHLEPR neurons, in the vlPAG further demonstrate the complex, heterogeneous nature of LH→PAG pathways and highlight the need for future circuit characterizations. Moreover, it is still unknown whether LHPV axonal projections to their target regions follow a one-to-one or one-to-many architecture. This is certainly an important question that has yet to be determined. However, our behavioral data suggest that these might be independent LHPV populations since we did not observe aversive-like effects during LHPV→vlPAG stimulation or antinociception during LHPV→LHb stimulation. Furthermore, the LHPV→LHb data also demonstrate that the antinociception evoked by activating the LHPV→vlPAG pathway was not due to antidromic stimulation effects. Future experiments will be needed to determine that these are indeed independent populations of LHPV neurons.

In a final series of experiments, we investigated the antinociceptive interactions between LHPV neuronal activation and the μ-opioid receptor agonist morphine. For a novel analgesic therapy to be useful, it must meet one of these three criteria: (1) possess analgesic properties alone, (2) facilitate analgesic action of existing treatments, or (3) decrease unwanted effects of existing treatments to make them more suitable for extended use (Li and Zhang, 2011). Our observation that LHPV neuronal activation attenuates nociception suggests that this manipulation meets the first criterion. Therefore, our last experiments were designed to assess the remaining criteria. To address the second criterion, we performed a dose-addition analysis between morphine and LHPV DREADD receptor activation by CNO as a standard pharmacological agent. We found that, depending on the proportion of drugs in the mixture, LHPV neuronal activation and morphine produced additive to synergistic interactions on thermal antinociception. The combination exhibiting the highest level of synergism required only a small stimulation of LHPV neuronal activity to greatly enhance morphine’s potency. Importantly, we included a group of control mice without hM3D receptors, in which CNO did not alter morphine’s potency. To address the third criterion above, we last investigated the effects of activating LHPV neurons following the development of tolerance to morphine-induced antinociception effect in the hot plate test. Here, chemogenetic activation of LHPV neurons evoked significant antinociception in morphine-tolerant mice, and more importantly, significantly restored morphine-induced antinociception. Remarkably, similar antinociceptive effects were maintained through another period of concurrent LHPV neuronal activation and morphine administration. Together, these findings show that LHPV neuronal activation can synergistically enhance acute morphine antinociception and restore its antinociceptive effects following the development of tolerance. Thus, activation of these LHPV neurons could be used to reduce the effective antinociceptive dose of morphine, helping to attenuate unwanted side effects such as respiratory depression and slow the rate of morphine tolerance.

An important point warranting further discussion is the contrast between our observations of divergent clusters of response patterns in LHPV neuronal activity during the hot and cold plate tests and the dominant behavioral phenotype of pain suppression when bulk activating these neurons. First, there is precedent to this contrast between diverse spontaneous activity patterns and more uniform behaviors driven by causal manipulations. For instance, LHGABA neurons responded to food locations, appetitive, or consummatory behaviors in a heterogeneous manner by either increasing or decreasing in activity at each event, yet when these neurons were bulk activated, mice ate voraciously, and when these neurons were ablated, mice ate less (Jennings et al., 2015). In the current study, we observed that most neurons identified as cluster 1 or 2 in the hot plate assay did not maintain the same designation in the cold plate assay, and vice versa. Therefore, it seems likely that, in general, the responses of LHPV neurons are either not consistent over time or dependent on the type of stimulus applied. For instance, some LHPV neurons may specifically suppress heat pain, others may suppress cold pain, others may suppress chemical pain, and so on. As such, we think that the bulk activation of LHPV neurons with optogenetics or chemogenetics during stimulation with one specific noxious stimulus (e.g., heat) likely activates a stimulus-specific cluster of LHPV neurons (heat) as well as the clusters specific for other stimuli (cold, chemical, mechanical, etc.) to evoke antinociception, as opposed to having populations of LHPV neurons that are exclusively pronociceptive or antinociceptive, the effects of which could be potentially diluted during bulk activation of these neurons. However, further work will be needed to elucidate how noxious stimuli are responded to and encoded by LHPV neuronal activity.

Here, we provide a detailed characterization of LHPV neurons, clearly demonstrating that these neurons modulate nociception through a distinct downstream circuit. Moreover, we measured and correlated LHPV neuronal activity patterns during noxious events. Finally, we found that chemogenetic modulation of these neurons could potentially be used as a standalone analgesic therapy or in combination with current analgesics such as morphine. These results support the continued investigation of LHPV neurons as a target for novel analgesics and warrant new efforts to identify neuronal populations in humans for targeting in clinical settings.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Genetic reagent
(Mus musculus; male/female)
PvalbCreThe Jackson LaboratoryRRID:IMSR_JAX:008069C57BL/6J background
Genetic reagent
(M. musculus; male/female)
Slc32a1CreThe Jackson LaboratoryRRID:IMSR_JAX:028862C57BL/6J background
Genetic reagent
(M. musculus; male/female)
Slc17a6CreThe Jackson LaboratoryRRID:IMSR_JAX:028863C57BL/6J background
Genetic reagent
(M. musculus; male/female)
LeprCreM.G. Myers Jr., University of Michigan Medical SchoolRRID:IMSR_JAX:032457C57BL/6J background
AntibodyAnti-DsRed, rabbit polyclonalTakara Bio, IncCat # 632496
RRID:AB_10013483
(1:1000)
AntibodyAnti-parvalbumin (PVALB), guinea pig polyclonalSwantCat # GP72; RRID:AB_2665495(1:300)
AntibodyAnti-rabbit Alexa Fluor 488, goat polyclonalThermo Fisher ScientificCat # A11034
RRID:AB_2576217
(1:500)
AntibodyAnti-guinea pig Alexa Fluor 488, donkey polyclonalJackson ImmunoResearch LaboratoriesCat # 706-545-148; RRID:AB_2340472(1:500)
AntibodyAnti-guinea pig Alexa Fluor 647, donkey polyclonalJackson ImmunoResearch LaboratoriesCat # 706-605-148; RRID:AB_2340476(1:500)
Recombinant DNA reagentrAAV2/9-CAG-FLEX-GCaMP6s-WPRE-SV40AddgeneRRID:Addgene_10084; Addgene viral prep 100842-AAV95.0 × 1012 GC/ml
Recombinant DNA reagentrAAV2/rh10-hSYN-DIO-hM3D(Gq)-mCherryUniversity of North Carolina (UNC) Vector CoreRRID:Addgene_443612.0 × 1012 GC/ml
Recombinant DNA reagentrAAV2/rh10-hSYN-DIO-hM4D(Gi)-mCherryUNC Vector CoreRRID:Addgene_443622.0 × 1012 GC/ml
Recombinant DNA reagentrAAV2/9-hSYN-DIO-mCherryAddgeneRRID:Addgene_50459; Addgene viral prep 50459-AAV92.1 × 1013 GC/ml
Recombinant DNA reagentrAAV2/1-CAG-FLEX-rev-ChR2-tdTomatoAddgeneRRID:Addgene_18917; Addgene viral prep 18917-AAV16.9 × 1012 GC/ml
Recombinant DNA reagentrAAV2/9-CAG-FLEX-ArchT-GFPUNC Vector CoreRRID:Addgene_297774.7 × 1012 GC/ml
Recombinant DNA reagentrAAV2/9-CAG-FLEX-GFPUniversity of Pennsylvania (U Penn) Vector CoreRRID:Addgene_515023.3 × 1013 GC/ml
Recombinant DNA reagentrAAV2/1-CAG-FLEX-tdTomatoU Penn Vector CoreRRID:Addgene_515034.5 × 1013 GC/ml
Recombinant DNA reagentrAAV2/9-CAG-FLEX-tdTomatoU Penn Vector CoreRRID:Addgene_515034.1 × 1013 GC/ml
Recombinant DNA reagentrAAV2/8-hSYN-FLEX-TVA-Rabies B19G (TVA+)Michigan Diabetes Research Center Molecular Genetics Core, University of Michigan4 × 1012 GC/ml
Recombinant DNA reagentEnvA-∆G-Rabies-mCherryMichigan Diabetes Research Center Molecular Genetics Core, University of Michigan1 × 1010 pfu/ml
Chemical compound, drugClozapine N-oxide (CNO)Tocris BioscienceCat # 4936; PUBCHEM:135445691
Chemical compound, drugAcetic acidSigma-AldrichCat # 320099; PUBCHEM:176
Chemical compound, drugFormalinMacron Fine ChemicalsCat # 5016–02; PUBCHEM:712
Chemical compound, drugComplete Freund’s adjuvant (CFA)Sigma-AldrichCat # F5881
Chemical compound, drugMorphineNational Institute on Drug Abuse Drug Supply ProgramPUBCHEM:5288826
Software, algorithmANY-maze video tracking system v5Stoelting Co.RRID:SCR_014289
Software, algorithmDoric Neuroscience Studio v5.1Doric Lenses IncRRID:SCR_018569
Software, algorithmFIJI/ImageJ v1.52phttps://imagej.net/FijiRRID:SCR_002285
Software, algorithmPrism 8GraphPadRRID:SCR_002798
Software, algorithmMiniscope Analysis PipelineEtter, 2021
Software, algorithmCellRegSheintuch et al., 2021
Software, algorithmMATLABMathWorksRRID:SCR_001622R2019a & R2020a
OtherSnap-in Imaging Cannula Model L-VDoric Lenses IncGRIN lenses
OtherBasic Fluorescence Snap-In Microscopy System – Deep BrainDoric Lenses IncIn vivo imaging system

Further information and requests for resources and reagents should be directed to and will be fulfilled by Yeka Aponte (yeka.aponte@nih.gov).

Experimental model and subject details

Animals

All experimental protocols were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and with the approval of the National Institute on Drug Abuse and Michigan State University Animal Care and Use Committees. Male and female heterozygous PvalbCre mice (RRID:IMSR_JAX:008069; C57BL/6J background, The Jackson Laboratory, Bar Harbor, ME, USA), Slc32a1Cre mice (RRID:IMSR_JAX:028862, C57BL/6J background, The Jackson Laboratory), Slc17a6Cre mice (RRID:IMSR_JAX:028863, C57BL/6J background, The Jackson Laboratory), and LeprCre mice (RRID:IMSR_JAX:032457; C57BL/6J background, kindly provided by M.G. Myers Jr., University of Michigan Medical School, MI, USA) were used in this study. Mice were maintained at the National Institute on Drug Abuse animal facility under standard housing conditions. Up to five mice of the same sex were group housed under a 12 hr light-dark cycle at 20–24°C and 40–60% humidity with free access to water and food (PicoLab Rodent Diet 20, 5053 tablet, LabDiet/Land O’Lakes Inc, St. Louis, MO, USA). For behavior experiments, 6- to 8-week-old male and female mice (∼18–25 g) were randomly assigned to experimental groups while maintaining littermate or age-matched and gender-matched controls. Following stereotaxic surgeries, mice were individually housed.

In all experiments, biological replicates were defined as ‘parallel measurements of biologically distinct samples that capture random biological variation,’ and technical replicates were defined as ‘repeated measurements of the same sample that represent independent measures of the random noise associated with protocols or equipment’ (Blainey et al., 2014).

Surgical procedures

Request a detailed protocol

For in vivo functional imaging experiments, mice were anesthetized with isoflurane and placed onto a stereotaxic apparatus (David Kopf Instruments, Tujunga, CA, USA). After exposing the skull by a minor incision, a small hole (<1 mm diameter) was drilled unilaterally (bregma, −1.78 mm; midline, +1.38 mm) for virus injection and GRIN lens insertion. A sterile, beveled 25-gauge needle was inserted into the center of the craniotomy stopping approximately 50 μm above the dorsal-ventral coordinate for the lens implant and remaining in place for 4–5 min to create a path for the implant. Next, rAAV2/9-CAG-FLEX-GCaMP6s-WPRE-SV40 was injected offset from the center of the craniotomy (100 nl; rate: 25 nl/min; RRID:Addgene_100842; Addgene viral prep 100842-AAV9; titer: 5.0 × 1012 GC/ml) into the LH of PvalbCre mice (bregma, −1.78 mm; midline, +1.365 mm; skull surface, −5.38 mm) by a pulled glass pipette (20–30 µm tip diameter) with a micromanipulator (Narishige International USA Inc, Amityville, NY, USA) controlling the injection speed. The injection is offset to avoid damaging the tissue in the lens field of view. After injection, a 500-μm-diameter GRIN lens (Snap-in Imaging Cannula Model L-V; Doric Lenses Inc, Québec, QC, Canada) was lowered into the center of the craniotomy (bregma, −1.78 mm; midline, +1.38 mm; skull surface, −5.28 mm). Implants were affixed to the skull with C&B Metabond Quick Adhesive Cement System (Parkell, Inc, Edgewood, NY, USA). Subsequently, mice were individually housed for 3−4 weeks for post-surgical recovery and viral transduction.

For behavioral experiments, mice were anesthetized with isoflurane and placed onto a stereotaxic apparatus (David Kopf Instruments). After exposing the skull by a minor incision, small holes (<1 mm diameter) were drilled bilaterally for virus injection. For experiments targeting parvalbumin neurons in the LH (LHPV), 40 nl of an adeno-associated virus was injected bilaterally (rate: 25 nl/min) into the LH of PvalbCre mice (bregma, −1.80 mm; midline, ±1.40 mm; skull surface, −5.40 mm) or LeprCre mice (bregma, −1.50 mm; midline, ±0.90 mm; skull surface, −5.40 mm) by a pulled glass pipette (20–30 µm tip diameter) with a micromanipulator (Narishige International USA Inc) controlling the injection speed.

Viruses used for chemogenetic experiments include (1) rAAV2/rh10-hSYN-DIO-hM3D(Gq)-mCherry (RRID:Addgene_44361; UNC Vector Core viral prep; titer: 2.0 × 1012 GC/ml), (2) rAAV2/rh10-hSYN-DIO-hM4D(Gi)-mCherry (RRID:Addgene_44362; UNC Vector Core viral prep; titer: 2.0 × 1012 GC/ml), and (3) rAAV2/9-hSYN-DIO-mCherry (RRID:Addgene_50459; Addgene viral prep 50459-AAV9; titer: 2.1 × 1013 GC/ml).

Viruses used for optogenetic experiments include (1) rAAV2/1-CAG-FLEX-rev-ChR2-tdTomato (RRID:Addgene_18917; Addgene viral prep 18917-AAV1; titer: 6.9 × 1012 GC/ml), (2) rAAV2/9-CAG-FLEX-ArchT-GFP (RRID:Addgene_29777; University of North Carolina [UNC] Vector Core viral prep; titer: 4.7 × 1012 GC/ml), (3) rAAV2/9-CAG-FLEX-GFP (RRID:Addgene_51502; University of Pennsylvania [U Penn] Vector Core viral prep; titer: 3.3 × 1013 GC/ml), (4) rAAV2/1-CAG-FLEX-tdTomato (RRID:Addgene_51503; U Penn Vector Core viral prep; titer: 4.5 × 1013 GC/ml), or (5) rAAV2/9-CAG-FLEX-tdTomato (RRID:Addgene_51503; U Penn Vector Core viral prep; titer: 4.1 × 1013 GC/ml).

For somatic-targeted optogenetic experiments, optical fibers were implanted bilaterally above LHPV somas (bregma, −1.80 mm; midline, ±1.40 mm; skull surface, −5.00 mm; no angle). For experiments targeting LHPV axonal projections within the vlPAG or the LHb, optical fibers were implanted bilaterally at 10° angles above LHPV axonal projections in the vlPAG (bregma, −4.00 mm; midline, ±1.00 mm; skull surface, −2.90 mm) or LHb (bregma, −1.70 mm; midline, ±0.90 mm; skull surface, −2.90 mm). For experiments targeting LHLEPR axonal projections within the vlPAG, optical fibers were implanted bilaterally at 10° angles above LHLEPR axonal projections in the vlPAG (bregma, −4.84 mm; midline, ±0.90 mm; skull surface, −2.50 mm). The axonal projections of LHLEPR neurons in the vlPAG are more posterior than those of LHPV neurons, which is why this more posterior coordinate was used (Schiffino et al., 2019; Leinninger et al., 2009). Implants were affixed to the skull with cyanoacrylate adhesive and C&B Metabond Quick Adhesive Cement System (Parkell, Inc). Subsequently, mice were individually housed for 3−4 weeks for post-surgical recovery and viral transduction.

In vivo functional imaging

A miniature microscope with an integrated LED was used to image GCaMP6s fluorescence in LHPV neurons through an implanted GRIN lens (Basic Fluorescence Snap-In Microscopy System – Deep Brain; Doric Lenses Inc). LHPV:GCaMP6s mice underwent five imaging sessions (hot plate, cold plate, ipsilateral formalin, contralateral formalin, and no formalin). Before each imaging session, GRIN lenses were briefly cleaned with isopropanol and mice were gently restrained while the snap-in microscope was secured to the baseplate for alignment with the implanted GRIN lens. Mice were then given approximately 5 min to acclimate to the microscope and tether. Grayscale TIFF images were collected at 10 frames per second (100 ms exposure) using Doric Neuroscience Studio software version 5.1 (RRID:SCR_018569). The LED power was calibrated between 10% and 50% (0.2–1.2 mW of 458 nm blue light). At the beginning of each session, imaging was synchronized with behavioral video recordings for later alignment. Sample size estimates were derived from a previous study using miniscope recordings in the hypothalamus (Betley et al., 2015).

For the hot plate tests, mice were placed on a 51°C hot plate (IITC Life Science, Woodland Hills, CA, USA) or a room temperature black cardboard surface with similar visual and tactile properties for 10–12 trials per stimulus. Mice were removed from the hot plate when typical behavioral responses were observed (e.g., paw withdrawal or paw licking). For the cold plate tests, mice were placed on a 4°C aluminum block or a room temperature white cardboard surface for 8–9 trials per stimulus. Mice were removed from the cold plate when paw checking or withdrawal responses were observed. 1−2 min interstimulus intervals were used for these tests. For the formalin tests, mice received a 20 μl intra-plantar injection of 2% formalin (Cat # 5016–02; Macron Fine Chemicals/Avantor, Radnor, PA, USA) diluted in saline. 47 min videos were captured, and formalin was injected into one of the hindpaws at the 2 min mark. For the ‘no injection’ test, no formalin was administered. These tests were separated by at least 5 days to minimize photobleaching and inter-test effects.

Image processing

Request a detailed protocol

Image analyses were performed using MATLAB scripts available in the Miniscope Analysis pipeline (https://github.com/etterguillaume/MiniscopeAnalysis). First, images were motion-corrected using the Non-Rigid Motion Correction (NoRMCorre) package (Pnevmatikakis and Giovannucci, 2017) and downsampled spatially and temporally by factors of 3. Motion-corrected, downsampled videos were then processed using Constrained Non-negative Matrix Factorization for Endoscopic data (CNMF-E) to extract individual neural segments, denoise their signals, demix signals from nearby neurons, and deconvolve calcium transients for estimation of neuronal firing (Friedrich et al., 2017; Zhou et al., 2018; Pnevmatikakis et al., 2016).

Imaging and behavioral analysis

Request a detailed protocol

For the hot plate and cold plate tests, filtered traces were Z-score normalized and smoothed with a rolling average of 3 frames. The 30 s activity traces surrounding stimulus presentations (10 s before to 20 s after) were averaged within each stimulus to form an average peri-stimulus activity trace per neuron. Neurons were assigned into clusters for further analysis if they displayed one of the two following phenotypes: (cluster 1) the average peak Z-score amplitude was ≥1.0 and the AUC of the trace following the stimulus was positive, or (cluster 2) the average peak Z-score amplitude was ≤−1.0 and the AUC of the trace following the stimulus was negative. The remaining neurons that did not meet either of these criteria were considered non-responsive to the noxious stimuli and were not analyzed further.

For decoding analysis, average traces for each neuron were constructed for hot plate, cold plate, or neutral stimulus trials for the 10 s period following stimulus onset. Principal component analysis was used to reduce the dimensionality of the averaged traces while maintaining 99% of the variance. 80% of the resulting traces from each cluster of neurons (see above) were used to train an SVM classifier in MATLAB (built-in function) to distinguish between activity resulting from a neutral stimulus and a hot or cold plate stimulus. The resulting classifier was used to predict which stimulus generated the remaining traces, and the predictions were compared to the known stimuli labels to determine the accuracy of the classifier. This process was repeated 100 times with random subsets of training data to obtain a distribution of test accuracies. To determine the significance of the test accuracy distribution, the labels of the testing dataset were randomly shuffled 100 times. Each label permutation was compared to the predictions obtained from one of the previously trained classifiers to form a null distribution of chance accuracies. A cumulative Gaussian curve was fit to the cumulative frequency distributions (0.10 bin size) of the test and null accuracies, and the distribution means were compared using the extra sum-of-squares F-test in Prism.

For the formalin experiments, deconvolved signals were used to bin estimated Ca2+ transients for every 5 min period of the test. For statistical comparison, we averaged the number of events per 5 min period within each phase of the formalin test (0–5 min, acute; 6–15 min, interphase; 16–45 min, inflammatory).

Optical manipulations

Request a detailed protocol

Optical fiber implants were coupled to patch cords connected to lasers (Doric Lenses Inc) via rotary joints mounted over behavioral testing areas. Optical fiber implants were custom-made and assessed for output efficiency ≥80%. Laser output was controlled by Doric Neuroscience Studio software version 5.1 (RRID:SCR_018569). For photostimulation experiments, 450 nm laser diodes were used to deliver 5 ms pulses of 10–15 mW light at a frequency of 5–100 Hz. For photoinhibition experiments, 520 nm laser diodes were used to deliver 10–15 mW of constant light.

Behavioral experiments

Mice were habituated to experimenter handling for 3 days prior to experiments, and all experiments were performed during the light cycle. Mice were acclimated to behavioral rooms for at least 1 hr before experiments began. Across experimental and control groups, mice were gender-matched and age-matched or littermates. By design, sample sizes were 8–12 mice based on (Bolles and Fanselow, 1980) previous literature using similar procedures (Negus et al., 2015; Jennings et al., 2015; Jennings et al., 2013; Siemian et al., 2019; Alhadeff et al., 2018) and (Tovote et al., 2015) estimates of exclusion rates following histology. Mice were excluded from analysis if viral expression and fiber placement were not observed in at least one hemisphere after histological assessment (see Histology).

Pain-suppressed nesting assay

Request a detailed protocol

Single-housed mice were tested in their home cages, which were initially supplemented with nestlet. Mice were acclimated to the procedure room for at least 1 hr before testing and had access to food and water in their cages throughout test sessions. At the start of each test, mice were pretreated with saline or CNO (1 mg/kg, i.p.; PUBCHEM:135445691; Cat # 4936; Tocris Bioscience, Minneapolis, MN, USA). After 1 hr, the existing nest was removed from each home cage, and a new nestlet cut into six small, equal-sized pieces was placed into the home cage, distributed across zones divided by a 3 × 2 grid in the cage. The mouse was then given an i.p. injection of 0.6% acetic acid (PUBCHEM:176; Cat # 320099; Sigma-Aldrich) in saline (10 ml/kg) or saline alone (10 ml/kg) and returned to the home cage. Measurements of the number of nestlet pieces collected were taken at 10, 30, 60, and 100 min post-acetic acid injection (Negus et al., 2015) by an experimenter blinded to the treatment group. The data from the 30 min time point were presented. At least 5 days separated tests to minimize inter-test effects (Negus et al., 2015).

Formalin place conditioning

Request a detailed protocol

Place conditioning experiments were performed in a two-chamber apparatus separated by a wall with a small door that could be closed with a divider. The chambers were defined by tactile, visual, and olfactory cues. One chamber had a metal grid floor, walls decorated with tan and black alternating vertical stripes, and almond scent. The other chamber had a smooth white floor, walls decorated with white circles on a tan background, and orange scent. The front wall of each chamber remained clear, and sessions were recorded using video cameras aimed through this wall using ANY-maze software. Pilot experiments showed that mice consistently preferred the metal grid side at a rate of 60–70% per 15 min test. In comparison to using an unbiased design, this biased design permits pre-assigning groups at surgery with less potential for mismatched side preference at pretest. All mice used in this study preferred the metal grid side in the 15 min pretest on day 1, and this side was assigned for pairing with formalin treatment.

Over the next 4 days, mice received one training session per day with the center door closed and only one chamber accessible; formalin sessions were video recorded for later behavioral scoring of paw licking behavior by a blinded scorer; some videos were difficult to view the mouse to score licking behavior and were removed from this analysis (two mCherry, three hM3D, and one hM4D). All sessions were preceded by an injection of CNO (1 mg/kg, i.p.; Tocris Biosciences) to control for potential subjective effects of LHPV manipulation in the absence of inflammatory pain. On even days (sessions 2 and 4), mice received an intra-plantar injection of saline (20 μl) in the hindpaw and were immediately placed in the initially non-preferred side for 60 min. On odd days (sessions 3 and 5), mice received a 20 μl hindpaw intra-plantar injection of 2% formalin (Cat # 5016-02; Macron Fine Chemicals/Avantor, Radnor, PA, USA) diluted in saline and were placed in the initially preferred chamber for 60 min. The formalin-treated paw was different on each of the two condition sessions. On day 6, untreated mice were placed back in the testing arena with free access to both chambers and the sessions were analyzed with ANY-maze video tracking system v5 (RRID:SCR_014289; Stoelting Co., Wood Dale, IL, USA).

Thermal nociception (hot plate test)

Request a detailed protocol

A cylindrical plexiglass enclosure was placed on a 51°C hot plate (IITC Life Science). For optogenetic experiments, patch cords were connected, and mice were placed in a holding chamber for an initial 3 min period. Mice were gently transferred to the hot plate and the latency to paw withdrawal (PWLHP) was measured. A latency of 20 s was defined as complete analgesia and used as a cutoff time to avoid tissue injury. Following this measurement, mice were removed from the hot plate and photomanipulations commenced for 3 min in the holding chamber after which mice were placed back on the hot plate for a second PWLHP measurement. Photomanipulations ceased for another 3 min period in the holding chamber before a final PWLHP measurement. For frequency-response experiments, this procedure was repeated for each frequency, except only one 3 min ‘laser-OFF’ period separated photostimulation epochs. For experiments examining the effects of longer photostimulation, 50 Hz photostimulation was delivered every other second over 20 min, and PWLHP was measured at the end of the photostimulation period and at 5, 10, and 20 min post-photostimulation. For experiments examining the effects of rimonabant on photostimulation-induced antinociception, rimonabant (3 mg/kg, i.p., dissolved in a vehicle of 8% Tween-80 in saline; PUBCHEM:5360515; Cat # 9000484; Cayman Chemical, Ann Arbor, MI, USA) was administered in a volume of 10 ml/kg 30 min prior to photostimulation. For chemogenetic experiments, CNO (1 mg/kg, i.p.; PUBCHEM:135445691; Cat # 4936; Tocris Bioscience) was administered after the second PWLHP measurement and measurements were taken periodically after (0.5–72 hr).

Mechanical nociception (von Frey test)

Request a detailed protocol

Mice were habituated for 20 min in cylindrical plexiglass enclosures on a fine mesh grid floor. For optogenetic experiments, patch cords were connected, and mice were placed in a holding chamber for an initial 3 min period. Von Frey filaments ranging from 0.008 g to 4 g were used to determine paw withdrawal threshold (PWTVF), which was defined as the lowest strength filament eliciting a behavioral response in at least two out of three applications. Briefly, measurements started with the lowest strength filament, and the filament strength was increased until paw withdrawal responses reliably occurred in at least two out of three applications. This procedure was repeated for each hindpaw in three epochs as described above for PWLHP measurements: pre-photostimulation, photostimulation, and post-photostimulation.

Real-time place preference

Request a detailed protocol

RTPP sessions were performed in a standard rat cage with opaque black siding filled with a thin layer of clean rodent bedding, except for a subset of LHPV→LHb mice that were also tested in a three-chamber apparatus consisting of two identical black-walled chambers separated by a narrow hall section, and the entire apparatus was filled with a thin layer of clean rodent bedding. Patch cords were connected, and mice were placed into the chamber. Photostimulation (50 Hz) or photoinhibition was paired with one side of the chamber, which remained constant across all tests. For LHLEPR→vlPAG experiments, 20 Hz photostimulation was used (Schiffino et al., 2019). Tests lasted for 10 min (LHPV somatic manipulations) or 20 min (axonal projection manipulations). At the end of the sessions, the percentage of time spent on the laser-paired side was calculated by ANY-maze video tracking system v5 (RRID:SCR_014289; Stoelting Co.).

Persistent inflammatory pain

Request a detailed protocol

Following initial behavioral tests after stereotaxic surgery and viral transduction, CFA (Cat # F5881; Sigma-Aldrich, St. Louis, MO, USA) was diluted 1:1 in saline and injected (20 μl) into the plantar surface of one hindpaw under brief isoflurane anesthesia (Alhadeff et al., 2018). Behavioral tests resumed 5 days post-CFA.

Persistent neuropathic pain

Request a detailed protocol

Following initial behavioral tests after stereotaxic surgery and viral transduction, the SNI model was used for induction of neuropathic pain. Briefly, under isoflurane anesthesia, the tibial and common peroneal nerves were axotomized while the sural nerve was spared (Decosterd and Woolf, 2000; Suter et al., 2003). Behavioral tests resumed 5 days post-SNI.

Dose-addition analysis

Request a detailed protocol

For the experiment examining interactions between CNO and morphine, tests were conducted according to a cumulative dosing procedure, in which PWLHP measurements are taken immediately prior to i.p. drug administration, then 60 min after drug administration immediately before the next drug administration. When administered alone, CNO was tested across a dose range of 0.1–3.2 mg/kg, and morphine was tested across a dose range of 3.2–32 mg/kg. For combination tests, these dose-measurement cycles continued until near 100% maximal effect was achieved corresponding to the predetermined cutoff time of 20 s. Raw PWLHP values for CNO and morphine were transformed into percent maximum possible effect (%MPE) values according to the formula %MPE = [(post-drug PWLHP – pre-drug PWLHP) / (cutoff time – pre-drug PWLHP) × 100]. %MPEs were averaged within each group (± s.e.m.) and plotted as a function of dose. Log(ED50) values were determined from the %MPE dose-response curve via linear regression and averaged within the group to calculate the ED50 (±95% confidence interval [CIs]) for each drug, except for CNO in the mCherry control group, which did not produce 50% effect levels. Morphine was obtained from the National Institute on Drug Abuse Drug Supply Program (PUBCHEM:5288826).

To examine the antinociceptive interactions between CNO and morphine, a fixed-proportion dose-addition analysis method was used (Siemian et al., 2018; Tallarida, 2010; Negus et al., 2009). For this analysis, CNO and morphine were combined in fixed proportions (1:1, 1:3, and 3:1) and administered using the cumulative dosing procedure as described. The actual doses of the drugs in the combination were determined by the relative potencies of each drug (based on the ED50 values) in the LHPV:hM3D group. For example, the 1:1 ratio consisted of one unit of the morphine ED50 (10.31 mg/kg) for every one unit of the CNO ED50 (0.78 mg/kg). By this method, the 1:3 ratio contained 0.5 × ED50 of morphine and 1.5 × ED50 of CNO and the 3:1 ratio contained 1.5 × ED50 of morphine and 0.5 × ED50 of CNO. Fractions of these mixtures (the combined 0.125 ×, 0.25 ×, 0.5 ×, 1 ×, and 2 × ED50 values of morphine and CNO) were administered consecutively by the cumulative dosing procedure to complete one dose-effect curve test. At least 1 week separated each test to avoid the development of tolerance and inter-test effects. Furthermore, a morphine-alone dose-response curve was taken 1 week after the last combination test, which showed that the morphine ED50 had not significantly changed (mCherry mice first morphine ED50 10.51 mg/kg, second morphine ED50 10.26 mg/kg; hM3D mice first morphine ED50 10.31 mg/kg, second morphine ED50 10.45 mg/kg). The shared dose-response curves were used to calculate the ED50 of each drug within each mixture. Isobolograms plotting the ED50 values of each drug were constructed to visually represent the nature of the drug interactions as additive, infra-additive, or supra-additive (synergistic).

Dose-addition analysis was performed as described previously (Siemian et al., 2018; Tallarida, 2000). When both drugs were active in an assay, expected additive ED50 values (±95% CL) (Zadd) were calculated from the equation Zadd = fA + (1 − f)B, where A is the ED50 of morphine alone, B is the ED50 of CNO alone, and f is the fractional multiplier of A in the computation of the additive total dose (e.g., f = 0.5 when fixed ratio was 1:1). When only one drug was active (i.e., morphine in the mCherry control group), the hypothesis of additivity predicts that the inactive drug (i.e., CNO) should not contribute to the effects of the mixture, and the equation reduces to Zadd = A/ρA, where ρA is the proportion of morphine in the total drug dose. Experimental ED50 values (Zmix) were determined from the 1:3, 1:1, and 3:1 combinations and were defined as the sum of the ED50 values of both drugs in the combination. Given the within-subject experimental design, Zadd and Zmix values were analyzed with paired two-tailed Student’s t-tests to determine differences between expected and experimental ED50 values.

Morphine tolerance study

Request a detailed protocol

One week after the last morphine-alone dose-response curve, we induced morphine tolerance in LHPV:Ctrl and LHPV:hM3D mice by administering 32 mg/kg morphine (i.p.) twice per day, separated by approximately 8 hr. We measured PWLHP before and 1 hr after the first injection on day 1 and the first injection on day 4 (the seventh injection overall) to verify tolerance development. On day 5, three PWLHP measurements were taken: pre-injection, 1 hr post-CNO injection, and 1 hr post-morphine injection. On days 6–8, a combined injection of 32 mg/kg morphine and 1 mg/kg CNO was administered once per day. On day 9, the day 5 test was repeated to measure potential development of tolerance to the morphine/CNO mixture.

Histology

Request a detailed protocol

Mice were deeply anesthetized with isoflurane and transcardially perfused with 1× phosphate buffered saline (PBS) followed by 4% paraformaldehyde (PFA) in 1× PBS. Whole brains were removed and post-fixed in 4% PFA overnight at 4°C and subsequently transferred to 1× PBS for storage at 4°C until further processing. Coronal brain sections (50 µm thick) were collected in 1× PBS using a Leica VT1200 vibratome (Leica Biosystems GmBH, Wetzlar, Germany). In some instances, DsRed immunostaining was required to visualize viral transduction. Sections were blocked for 1 hr at room temperature in 1× PBS with 0.3% Triton X-100% and 3% normal goat serum. After blocking, sections were incubated with rabbit anti-DsRed antibody (1:1000 Cat # 632496/RRID:AB_10013483; Takara Bio, Inc, Mountain View, CA, USA) in block solution for 20 hr at 4°C. Tissue was then washed 4 × 10 min in 1× PBS followed by incubation in goat anti-rabbit Alexa Fluor 488 antibody (1:500 Cat # A11034/RRID:AB_2576217; Thermo Fisher Scientific, Waltham, MA, USA) in block solution for 1.5 hr at room temperature. After secondary antibody incubation, sections were washed 4 × 10 min in 1× PBS. All sections were mounted with DAPI-Fluoromount-G aqueous mounting medium (Electron Microscopy Sciences, Hatfield, PA, USA) onto Superfrost Plus glass slides (VWR International, Radnor, PA, USA). Images were taken with an AxioZoom.V16 fluorescence microscope (Carl Zeiss Microscopy LLC, Thornwood, NY, USA).

Recombinant rabies virus tracing

Request a detailed protocol

For retrograde monosynaptic tracing experiments, Slc32a1Cre and Slc17a6Cre mice were anesthetized with ketamine/xylazine (90/10 mg/kg i.p.) and placed onto a stereotaxic apparatus (David Kopf Instruments). After exposing the skull by a minor incision, a small hole (<1 mm diameter) was drilled unilaterally for helper virus injection. 40 nl of Cre-dependent AAV8/hSyn-FLEX-TVA-Rabies B19G (TVA+) was injected unilaterally (rate: 10 nl/min; titer: 4 × 1012 GC/ml) into the vlPAG (bregma, −3.90 mm; midline, ±0.2 mm; skull surface, −3.20 mm) by a 25-gauge Hamilton syringe (500 nl). 3–4 weeks later, mice were injected with 100 nl of the recombinant rabies viral vector (EnvA-∆G-Rabies-mCherry; titer: 1 × 1010 pfu/ml) at the same vlPAG coordinate. Both viruses were graciously provided by the Michigan Diabetes Research Center Molecular Genetics Core, University of Michigan. 3–4 weeks after the recombinant rabies virus injection, mice were deeply anesthetized with isoflurane and transcardially perfused with 1× PBS followed by 4% PFA in 1× PBS. Whole brains were removed and post-fixed in 4% PFA overnight at 4°C and subsequently cryoprotected by equilibration in 30% sucrose in 1× PBS at 4°C, flash frozen in isopentane on dry ice, and stored at −80°C. Tissue was embedded in Tissue-Tek O.C.T. Compound (Sakura Finetek USA, Inc, Torrance, CA, USA) for cryosectioning. Coronal brain sections (50 µm thick) were collected in 1× PBS using a Leica CM3050 S cryostat (Leica Biosystems GmBH, Wetzlar, Germany). Sample size estimates were derived from a previous study using the same methodology (Jennings et al., 2013).

For parvalbumin (PVALB) immunostaining, sections containing the hypothalamus were blocked for 1 hr at room temperature in 1× PBS with 0.3% Triton X-100% and 3% normal donkey serum. After blocking, sections were incubated with guinea pig anti-PVALB antibody (1:300 Cat # GP72; RRID:AB_2665495; Swant, Marly, Switzerland) in block solution for 16 hr at 4°C. Tissue was then washed 4 × 10 min in 1× PBS followed by incubation in donkey anti-guinea pig Alexa Fluor 488 or 647 antibody (1:500 Cat # 706-545-148/RRID:AB_2340472 or Cat # 706-605-148/RRID:AB_2340476; Jackson ImmunoResearch Laboratories, Inc, West Grove, PA, USA) in block solution for 1.5 hr at room temperature. After secondary antibody incubation, sections were stained for 5 min with 4',6-diamidino-2-phenylindole, dilactate (DAPI 1:5000; Thermo Fisher Scientific) in 1× PBS followed by 3 × 10 min washes in 1× PBS. Sections were mounted with Fluoromount-G aqueous mounting medium (Electron Microscopy Sciences) onto Superfrost Plus glass slides (VWR International). Z-stacks (30 µm) containing the LHPV region were imaged with an LSM 700 microscope using a 20× air objective (Carl Zeiss Microscopy LLC). Maximum intensity projections were manually counted using Fiji v1.52p software (RRID:SCR_002285) with the cell counter plugin (Schindelin et al., 2012). Sections were anatomically matched to ensure that the same regions were analyzed across samples. Additionally, sections containing the PAG were mounted as described, and images were taken with an AxioZoom.V16 fluorescence stereomicroscope using a 7× digital magnification to assess the injection site for mistargeted or lacking virus expression. After PAG assessment, one Slc32a1Cre and one Slc17a6Cre sample was excluded from the analysis as viral expression was not observed in the vlPAG.

Statistics

Graphs and statistics for behavioral experiments were prepared with GraphPad Prism 8 software (RRID:SCR_002798; GraphPad, La Jolla, CA, USA). All data are plotted as mean ± s.e.m., except for Ca2+ event frequency data and isobolograms, which are plotted as mean ± 95% CI, and cell counts, which are plotted in ‘part-of-whole’ format. Paired or unpaired Student’s two-tailed t-tests, one-way, two-way, or three-way mixed model ANOVAs with Bonferroni or Dunnett’s post-tests for multiple comparisons corrections were used to analyze all behavioral data, as appropriate. Mann–Whitney U-tests with Holm–Sidak correction for multiple comparisons were used to analyze Ca2+ event frequency data from the formalin tests. A chi-square test was used to compare cell counts in the rabies tracing experiment. For all statistical tests, p<0.05 was considered significant.

Data availability

All data generated or analyzed during this study are included in the manuscript and supporting files. Source data files have been provided for Figure 1.

References

    1. Cleeland CS
    2. Ryan KM
    (1994)
    Pain assessment: global use of the brief pain inventory
    Annals of the Academy of Medicine, Singapore 23:129–138.
    1. Tallarida RJ
    (2010) Combination analysis
    Advances in Experimental Medicine and Biology 678:133–137.
    https://doi.org/10.1007/978-1-4419-6306-2_17

Decision letter

  1. Peggy Mason
    Reviewing Editor; University of Chicago, United States
  2. Michael Taffe
    Senior Editor; University of California, San Diego, United States
  3. Robert Gereau
    Reviewer; Washington University in St Louis, United States
  4. Peggy Mason
    Reviewer; University of Chicago, United States
  5. Alexander C Jackson
    Reviewer; University of Connecticut, United States
  6. Gregory Corder
    Reviewer
  7. Asaf Keller
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

In this work, Siemian and co-authors investigate a cluster of parvalbumin-expressing excitatory neurons in the lateral hypothalamus and their role in modulating pain states and pain-associated behavior. Overall, this work is well-executed and convincing and will be broadly interesting to both the hypothalamic circuits and pain neuroscience communities, and will contribute to our understanding of the role of hypothalamic neurons in modulating pain responses and pain behavior.

Decision letter after peer review:

Thank you for submitting your article "An excitatory lateral hypothalamic circuit orchestrating sensory and affective pain" for consideration by eLife. Your article has been reviewed by 5 peer reviewers, including Peggy Mason as the Reviewing Editor and Reviewer #2, and the evaluation has been overseen by Michael Taffe as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Robert Gereau (Reviewer #1); Alexander C Jackson (Reviewer #3); Gregory Corder (Reviewer #4); Asaf Keller (Reviewer #5).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

– Discussion of the inhibited vs excited LH neurons.

– Responses of single neurons to multiple stimuli – coordinated or independent.

– Are there dual projections to PAG and LHab (if no data, then a thorough acknowledgment and discussion of this issue)?

– Concerns over mixed effects of Arch on synaptic transmission.

– Persistent over chronic is preferred for the time scales used here.

– Small sample sizes from low numbers of animals and small effects make the presentation of raw data points all the more desirable.

– Flesh out intro beyond to introduce the component reactions to noxious stimulation that you will discuss – nocifensive, suppression of everyday activities and the intersection with aversion that eventually is discussed with respect to both LH and LHab.

Reviewer #1:

This is an interesting study from the Aponte lab that builds on their prior work that identified lateral hypothalamic parvalbumin-expressing, glutamatergic neurons (LHPV neurons) as regulators of pain behaviors in mice. Here, the authors explore how painful stimuli impact activity of the LHPV neurons in vivo, and further examine the impact of manipulating activity of these neurons in the context of inflammatory and neuropathic pain, as well as the effects on affective aspects of pain.

Using miniscope recordings, the authors find that there are two main populations of LHPV neurons – the majority of which are activated in association noxious stimulation, but a subset that are inhibited.

Studies using chemogenetics and optogentics, all with appropriate control groups, define clearly the effects of activating these LHPV neurons, and show that the send important functional projections to the periaqueductal grey.

The manuscript misses a potentially important opportunity in defining what these different classes (pain-activated vs inhibited) neurons in the LH do, but there is still much that we learn here.

One technical issue that needs to be addressed is around the use of ArchT for inhibition of LH→PAG terminals, as ArchT can have complicated actions at terminals.

The authors identify differences in projections to the lateral habenula vs. PAG, which are interesting and suggest different functional outputs. It will be important to understand if these are unique cell populations, or just dual projections from the same population of neurons in the LH.

Overall, this represents a significant contribution to the field. I commend the authors on the thorough characterization of their behaviors and detailed exploration of the optogenetic stimulation frequency space – which is largely missing in most papers. There are a number of technical questions that should be addressed to further strengthen the manuscript, and it is worth considering whether some of the data presented here would be better reserved for a future publication to make the present manuscript more focused on the main points outlined in the manuscript title.

Comments for the authors:

1. The imaging studies in figure 1 are intriguing and provide some very interesting insights – but the presentation raises a number of questions.

a. Overall – the data show that there are LHPV neurons that seem to show increased activity to heat and cold, and some that show decreased activity. The authors call this cluster 1 and 2 respectively. This is interesting, and points to the importance of using the miniscope approach to get cellular level resolution, as it appears that simple fiber photometry may have shown no change in activity in response to these stimuli. This should be highlighted.

b. Similar trends are seen for hot plate and cold plate. If you look at the single cell level – are neurons that classify as cluster 1 for hot plate also in cluster 1 for cold plate? Presumably the same mice were tested for both assays. This is not clear.

c. The z scores show something rather interesting – there is clearly a period of suppressed activity in cluster 1 cells and increased activity in cluster 2 cells preceding contact with the plate. This is evident also in the "control" group. What do the authors think of this? The inclusion of the control groups here was absolutely critical, as it shows that there are indeed differences. Related – the post-hoc analysis states significant differences are illustrated by the bold line – but this is not easy to see. Am I correct in seeing that there is no significant differences in cluster 2 for the cold stimulus vs. control? The manuscript states significant differences, but the posthoc either does not support this or it is not clear from the bold shading of the line in Figure 1l.

d. The cluster 1 neurons are more prevalent, and the authors comment that the response magnitude is larger than those for cluster 2. This should be supported by statistical comparisons validating this claim.

e. For the formalin data – given the results from the hot and cold plate, it is difficult to know how to interpret these findings. Were these cells recorded here specifically cluster 1 cells as determined from the hot and cold plate? Again- were these the same mice tested for hot and cold plate? An analysis of these data for formalin separately for neurons defined as cluster 1 and 2 for hot/cold would be much more informative than including all together.

f. The interpretation of formalin responses as "long-term" seems unwarranted. In fact, the response to formalin is ongoing. So – this would suggest sustained responses to ongoing nociception.

g. Related – it would be of interest to analyze the formalin data as Ca2+ transients time-locked to nocifensive responses (flinching, lifting, licking). Did you collect videos of the mice that would allow this analysis? Are there also "cluster 1" and "cluster 2" type responses associated with flinching/licking?

2. The authors go on to study the impact of manipulating LHPV activity on behaviors, but sort of ignore the fact that there are these two populations of neurons with respect to activity in response to noxious stimulation. It is unfortunate that the interesting findings of the two different clusters of LHPV neurons – one inhibited by noxious stimulation and one activated – is not built upon here – but is largely ignored. The subsequent studies go about activating or inhibiting all LHPV neurons to test their effects on sensory and affective aspects of pain. The challenge with this approach, based on the imaging data, is that there is a mixture of activity patterns in response to noxious stimuli. For example – heat responses. As you are targeting LHPV neurons globally – you will be hitting both cluster 1 and cluster 2 neurons. I know that it would be a huge task to sort out what is different about these two populations – but this caveat at least needs significant discussion. It would be very interesting to be able to determine what is different about the cluster 1 and cluster 2 neurons. This is not addressed here, and I think may be beyond the scope of this study. However – this should be discussed. Why are some neurons inhibited, some activated? Do these neurons have different projection targets?

3. Activating the LHPV neurons leads to reductions in responses to painful stimuli, and reversal of pain-suppressed behaviors. Given that the majority of these neurons are activated by painful stimuli, what does this mean for the role of LHPV neurons in pain? You allude to this briefly in the discussion, but this is the major question in my mind in terms of the major conclusions we can make from this nice study. Are LHPV neurons mediating endogenous analgesia during painful stimuli? If so, then this could be mediating a DNIC type response. It would be very interesting to know if you can elicit DNIC in mice, and block this by inhibiting these LHPV neurons.

4. The use of ArchT for photoinhibition of the LHPV → vlPAG terminals is problematic given reports that activation of Arch actually produces increases in spontaneous synaptic release at some terminals. It is imperative that the authors characterize the effects of Arch activation on synaptic transmission (evoked and spontaneous) at this synapse to allow interpretation of the findings.

5. The results re: the projections to the LHb are also very interesting. Are the neurons that project to the LHb and vlPAG distinct, or do individual neurons send process to both LHb and vlPAG? A double retrograde tracing experiment would expand the scope of what we learn here significantly.

Reviewer #2:

This is a thorough, multipronged study which supports a role for parvalbumin-positive neurons in the lateral hypothalamus (LH-Pv+) in modulating the nocifensive and affective / motivational consequences of noxious stimulation and morphine suppression of the same. The authors provide evidence that both nocifensive and affective / motivational consequences are suppressed by activation of glutamatergic LH-PV+ neurons whereas activation of GABAergic LH projections (expressing the leptin receptor) result in place aversion.

Experiments also demonstrate that LH-PV+ activation combines in an additive and at one ratio in a synergistic manner with morphine and that activation of LH-PV+ neurons reverses tolerance to morphine.

The nociceptive-specific nature of the sensory responses is not convincing given that there is a comparison between stimulus and not-stimulus rather than between innocuous stimulus and noxious stimulus. The authors may argue that room temperature is a thermal stimulus and perhaps that is true, depending on the animal's skin temperature which was not measured. But a better comparison would be between a 45C or so stimulus and the 51C used. I am not suggesting that the authors add such expts rather I recommend a more careful interpretation of the data collected. This concern is strengthened by the response to ipsilateral formalin which raises the issue of whether the "responses" may be in fact correlates of efferent function involved in autonomic or state arousal. Yet, the remaining experiments mitigate this concern.

In pairing the (idiosyncratically) preferred side with CNO-inhibition, the preference for either side goes to 50%. Thus inhibition of LH-PV+ neurons takes away side preference but does NOT produce side avoidance as is stated in lines 153-54.

Please jog the gray symbols in Figure 4 supp 1, 2 laterally so they can be seen.

The LH-PV+ preference for vglut2 over vgat PAG neurons is by less than a factor of 3. Given what we know about the unimportance of absolute numbers of synapses in dictating the responses of lemniscal pathways in the thalamus, this result is interesting but with this amount of information uninterpretable. Tone this down.

Reviewer #3:

In this study, Siemian and co-authors significantly expand upon their previous work (Siemian et al., 2019) examining the role of parvalbumin-positive glutamatergic neurons in the LH (LHPV), and their projections to the vlPAG, in the modulation of pain responses. In their previous work, the authors showed that optogenetic activation or inhibition of LHPV neurons suppresses or potentiates, respectively, responses to a noxious thermal stimulus and that this effect may be mediated by an excitatory projection to the vlPAG. The current manuscript is exciting in that it describes a much more detailed investigation of this circuit, using in vivo activity monitoring and manipulation (both chemo- and optogenetic) in multiple models of pain and pain-related behavior (both acute and chronic) as well as a neuroanatomical elucidation of this circuitry. Overall, this work uses a wide-spectrum of modern tools to address the circuit-basis of LHPV modulation of pain states and pain-associated behavior. This well-executed work contributes to filling a significant gap in our understanding of the role of LH neurons in modulating pain responses and pain behavior.

In their current work, the authors start by observing in vivo calcium signals with single cell resolution using GRIN-lens endomicroscopy of LHPV neurons in PV-cre mice in several pain assays. They found that a subset of LHPV neurons were activated by an acute thermal stimulus (hotplate), a much smaller subset were inhibited and large proportion did not respond. In another set of experiments, a subset of neurons were found to be activated by an acute noxious cold stimulus (coldplate), a small subset were inhibited and large proportion did not respond. They also found that some LHPV neurons responded to formalin injection over longer time-scales. These experiments successful provide solid evidence that at least a subpopulation of LHPV neurons respond to noxious stimuli which is a key finding. Less clear is how the heterogeneity observed in these responses may be linked to the circuit manipulations in the subsequent experiments, which engage the whole population of LHPV neurons, and how specific LHPV neurons are in modulating pain pathways.

In the next series of experiments, the authors manipulate LHPV neurons using chemogenetic and optogenetic methods in PV-cre mice, in a variety of standard pain assays. They first found that chemogenetic activation of LHPV neurons suppressed pain responses, while inhibition of LHPV neurons potentiated pain responses (increases and decreases in thermal pain thresholds respectively). In an interesting assay of homecage behavior, they further found that chemogenetic activation of LHPV neurons prevented a pain-associated reduction in both nesting behavior and place aversion, which strengthens the argument that activation of LHPV neurons is antinociceptive and seems to diminish pain-associated aversion. Finally, optogenetic activation of LHPV neurons appeared to diminish thermal and inflammatory pain, consistent with their chemogenetic results. However, optogenetic activation of LHPV neurons also produced mild place avoidance in an RTPP assay, somewhat at odds with their chemogenetic results. Overall, the chemo and optogenetic manipulation experiments succeeded in showing that activation of LHPV neurons has antinociceptive effects.

Next, the authors significantly extended their previous work by further probing the role of LHPV projections to both the vlPAG and LHb. Optogenetic stimulation of LHPV inputs to the vlPAG increased both acute thermal pain and more chronic neuropathic pain thresholds while inhibition decreased it, consistent with direct somatic manipulations. Interestingly, these manipulations did not appear to affect place preference leading to their examination of LHPV projections to the LHb, activation of which was aversive, consistent with previous work showing that the broader population of LH-VGLUT2 to LHb pathway is aversive. To address whether excitatory or inhibitory vlPAG populations may be targeted by LHPV neurons, the authors used rabies tracing and found that although a very small percentage of LHPV neurons appeared to innervate the vlPAG, a greater proportion of excitatory than inhibitory neurons were targeted. Finally, the authors found that LHPV-induced antinociception enhanced and morphine-induced antinociception and rescues morphine tolerance.

Overall these data are novel and suggest that even a very tiny cluster of LH neurons has a rather outsized role to play in pain modulation. These findings would contribute to our limited understanding of both the role of hypothalamic circuits in pain modulation and the intersection of reward, aversion and pain.

Comments for the authors:

– Although the in vivo GRIN lens experiments provide evidence that a subpopulation of LHPV neurons respond to noxious stimuli, there's less clarity on the following questions:

1. Does the subpopulation of LHPV neurons activated by noxious thermal stimulation overlap with the subpopulation of neurons activated by noxious cold stimulation, ie is a specific subset of these neurons broadly tuned to respond to noxious thermal and cold stimuli or are they separate? The authors could provide more clarity on whether the cold/hot assays were done in the same animals and if not, provide some commentary on this question.

2. What is the degree of inter-animal variability in the data? Did some animals contribute more or less to the population(s) that responded to thermal/cold stimuli?

3. Given that only a relatively small subpopulation of LHPV neurons are activated by noxious stimuli, with another subpopulation inhibited, how does one interpret the chemo- and optogenetic data in which all LHPV neurons are either activated or inhibited? This should be addressed as a caveat in the discussion.

– The rabies tracing experiment is well-done and informative but is a somewhat indirect way of showing preferential innervation of different vlPAG populations. Directly addressing this is a technically difficult problem and perhaps the subject of a follow-up study but a discussion of the caveats would be helpful. One possible caveat in this experiment is that, for the sake of argument, a simply larger population of excitatory vlPAG neurons may skew the representation of retrogradely-labeled LHPV neurons despite perhaps an equal rate of connectivity with excitatory and inhibitory neurons.

– Another outstanding question is whether the same population of LHPV neurons are innervating both the vlPAG (antinociception) and the LHb (aversion) or if they are entirely different subpopulations. One way of approaching this experimentally is to inject two different conjugated CTBs, one in the vlPAG and the other in the LHb to determine if retrogradely-labeled LHPV neurons overlap or not. If this experiment is not possible, addressing this in the discussion would be helpful.

– In the Discussion, the authors refer to LHPV neurons as a whole in the context of pain modulation when their data actually implicates only a comparatively small subpopulation of these neurons in modulating pain responses, evidenced by their in vivo monitoring and rabies tracing. This language should be adjusted in the discussion given that LHPV neurons are clearly not functionally monolithic. Furthermore, a short discussion of any known heterogeneity among LHPV neurons would be useful here.

– Finally, in terms of the writing, the manuscript is overall quite well-written.

However, the introduction could be improved. The authors start with a general description of pain and pain behavior and quickly transition to the role of the lateral hypothalamus in pain. It would be useful to make that transition more fluid and more detailed given that this is rationale for studying these cells in the context of pain. This could be accomplished by a more general description of the LH and how the repertoire of behaviors associated with the LH are relevant to pain responses. For example, the section of text: "While the LH circuits controlling food intake and reward have received intense focus over the past several years, those governing nociception have gone understudied by comparison. Thus, with its diverse array of neuronal populations, uncovering genetically defined LH circuits that regulate pain may bring forth a novel therapeutic target" should be expanded upon somewhat and include some key references. At the moment, there are no references here.

– Also, the description of the role of the LH in pain in the introduction is sparse and should include a slightly more detailed description of the literature with key references. This can be included either as part of the introduction or in the discussion. Also, there are some key references missing here. An example includes a set of papers describing a role for an LH substance P projection to the PAG in mediating antinociception (Holden and Naleway, 2001; Holden et al., 2002; Pizzi and Holden, 2008; Pizzi et al., 2009)

Reviewer #4:

The miniscope calcium imaging data are excellent. For all imaging studies, summary statistics on which cells or how many cells are from which specific animal should be graphically shown somewhere. From the current display one cannot tell if the active and inhibited cells are from just one animal or are observed in all animals. It would beneficial to see several "per animal" metrics, i.e. showing the data in Figure 1 E and I not as a merge of all neurons across animals but for each individual mouse, as well as show the proportion of active and inhibited cells per subject.

1. The title states "sensory and affective pain", which I read initially as the paper would present data on the dichotomy of the sensory and emotional qualities of pain perception, as originally defined by Melzack 1968. However, the assays used (primarily reflexive tests of hypersensitivity) measure only sensory aspects of behavior. In the text though (in comparison to the title), the authors use "affect" to refer to "pain suppressed behaviors" / "pain induced co-morbidities", which I agree are partly captured by the nest building assay. The title to more accurately reflect the data, and remove "affective pain".

2. The miniscope data find two Clusters of cell-types functionally defined by being active or inhibited to noxious stimuli. This result is nice. However, it is disconnected from the remainder of the data, and as currently presented is a very elaborate, but far superior, Fos-like experiment showing that some cells are active while others are not, but it is not clear how and to what function these cell-types contribute to results presented after Figure 1. To avoid a difficult and time-consuming miniscope experiment to image projection neurons to PAG vs. habenula, I suggest a tracing-topology study to link this data to the other very important and interesting results in Figure 6 on differential results seen in the LH projections to the PAG vs habenula. For example, injection of a retrograde-AAV-FlpO recombinase into the PAG (and in a separate cohort, injection into habenula) of Pv-Cre mice with a Cre-ON/Flp-On intersectional fluorophore into the hypothalamus plus a noxious stimulus to induce FOS in these cells. Several conclusions could be drawn to compliment the miniscope functional data and the behavioral data, namely 1) are there collateral projections from LH◊PAG that also project to the habenula; 2) what proportion of these projection cells are nociceptive (FOS+); 3) are there functionally distinct cell-types (e.g. Clusters 1 and 2) that preferentially send projections to PAG or habenula (i.e. Fluorphore+ and FOS+/FOS-). This type of minor experiment might also provide additional insight to understand the two activity Clusters and how the bulk optogenetic/chemogenetic activation of LHPv vlPAG neurons is antinociceptive when this experimental design "turns on" the nociception-inhibited Cluster-2.

3. I would consider not referring to the 7-day timepoint post Spared Nerve Injury, a model of "chronic neuropathic pain". This is still an acute surgical neuropathy model, whereas most "chronic" designations should be reserved for the 3+ week timepoint, at a minimum.

4. No axonal terminals can be seen in the image of Figure 4A nor Supp Figure 3. Are there other images from these mice illustrating that this projection connects with this more posterior portion of the vlPAG? The group's prior 2019 paper shows dense LHPV innervation of the very anterior PAG (superocularmotor region), in contrast to some fiber placements in this article, which are almost 1.0 mm apart. In addition to the new images to confirm axons under the fiber tracts, I would request that images and quantification be provided for LH-PV axon densities across the anterior-posterior axis of the PAG. This will also be very helpful for the reader to link the past work with the current manuscript, as well as make sense of the choice of A-P coordinates for the fibers (-4.0 and -4.8 [Leptin-Cre study]) and for the RABV tracing experiment (which was done at the anterior -3.8 coordinate which was also used in the 2019 paper). Alternatively, performing a patch-clamp experiment (as done by the group in the 2019 paper) of Vglut2 or Vgat PAG neurons in the posterior PAG would confirm this connection, since the provided images do not show any axons in this region.

5. Even though it was stated by the authors that it created some clutter, I would still suggest to show all individual dots and lines for behavior throughout the figures (perhaps make the lines 50% transparent)

Reviewer #5:

In 2019 the Aponte lab (10.1038/s41598-019-48537-y) reported that a small cluster of lateral hypothalamic neurons that express the calcium-binding protein parvalbumin (LH-PV neurons) modulate nociception in mice. They showed that photostimulation of these neurons suppresses nociception to an acute, noxious thermal stimulus, and that photoinhibition potentiates thermal nociception. They also showed that these neurons form functional excitatory synapses on neurons in the ventrolateral periaqueductal gray (vlPAG), and that photostimulation of these axons mediates antinociception. Finally, they showed that the anti-nociceptive effect appears to occur independently of opioidergic mechanisms. Many of these findings are replicated here.

In the present study they add to these findings by demonstrating, with the use of calcium imaging from behaving mice, that the LH-PV neurons respond to noxious stimuli. They also demonstrate that projections of these neurons to vlPAG affect both sensory and affective aspects of pain, whereas projections to the habenula appear to affect only the affective/aversive components.

The conclusions of this paper are mostly supported by the data, but some detailed aspects could to be clarified (laid out below).

1. Some of the effects reported appear rather small, and some reported differences might be driven by outliers. For example, data in Figure 1 d,g,h,k,l represent changes smaller than 2 standard deviations, or less than one Z score. Data in Figure 4i suggest very small effects on paw withdrawal thresholds. A consideration of whether these small changes are functionally meaningful would be particularly useful. Differences depicted in data in Figure 1 o,p,q appear to driven by a small number of outliers; even if these data survive tests of statistical outliers, one wonders why the vast majority of experiments show no differences.

Related to this, it would be useful to know what criteria were used to ensure that parametric analyses are appropriate. It is not clear why, in some comparisons, both Bonferroni and Dunnett's multiple comparisons are used on the same datasets. And, depicting variances as confidence intervals, instead of SEM, will likely be more informative.

Sample sizes are quite small. Although a large number of neurons are depicted, they were collected from a small number (e.g. 3) mice. At the very least, showing data from individual mice (instead of pooling data from all mice) will help determine how reproducible the results are.

2. Some neurons appear to increase their activity in response to stimuli, whereas others decrease their activity (Figure 1). It would be informative if the authors discuss this intriguing finding.

3. The doses of morphine that were effective appear rather high (Figure 7). Discuss?

4. Do individual LH-PV neurons project to both vlPAG and habenula? If so, can we exclude the possibility that terminal photostimulation antidromically activated LH neurons and their unintended axonal targets?

If these are independent projections, can the authors discuss how the LH projections to vlPAG vs to habenula might be regulated or balanced during different states?

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "An excitatory lateral hypothalamic circuit orchestrating pain in mice" for further consideration by eLife. Your revised article has been evaluated by Michael Taffe as the Senior Editor and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

The revisions have exacerbated a problem in terminology. Pain is a percept. Noxious stimulation is a stimulus. Pain behavior is a package of skeletal muscle-mediated behaviors that ambiguously may include (or not) autonomic reactions. Thus, the title is problematic – you have little information on the pain percept (CPP and that is it). The bulk of your data speaks to pain behavior. And stimuli are not painful until proven so. Stimuli should be described as noxious (if they are indeed in that range).

Many of these issues converge in the sentence "As such, rodent studies searching for new pain interventions have begun to investigate ethological behaviors like nesting that are suppressed by pain (e.g., forgoing standard life activities) as well as the affective/emotional component of pain with assays of pain-induced aversion (e.g., comorbid depression) (5, 11, 12) identifying specific brain pathways capable of managing these multiple components of chronic pain and developing strategies for targeting them for translational use will advance the search for novel pain therapies."

Possible changes include: new pain interventions/therapies → new analgesic interventions, therapies; suppressed by pain → suppressed during pain behavior; assays of pain-induced aversion → assays of noxious stimulus-induced aversion; components of chronic pain → components of chronic pain behavior.

The suggestion that "the bulk activation of LHPV neurons with optogenetics or chemogenetics likely activates more than the necessary number of them (hot responders + cold responders) to evoke uniform antinociception effects in a given assay" is not clear. Please explain how this could work. Related to this, if the responses to the noxious stimulation are so easily dismissed in favor of "bulk activation," what are the implications, the import, if any, of these responses? If without import, then why are they shown?

Typo in line 274 "since-cell".

https://doi.org/10.7554/eLife.66446.sa1

Author response

Essential revisions:

– Discussion of the inhibited vs excited LH neurons.

– Responses of single neurons to multiple stimuli – coordinated or independent.

First, we decided to address the comments regarding (a) inhibited versus excited LHPV neurons and (b) responses of individual LHPV neurons to multiple stimuli (coordinated or independent) together, since they are closely related. For this, we performed cell registration across the five calcium imaging sessions (Figure 1—figure supplement 2). We found 33 neurons detected in both the hotplate and coldplate sessions in which we examined time-locked responses. Of those 33 neurons, only 6 neurons remained in a particular cluster in both sessions (e.g., cluster 1 in both the coldplate and hotplate), and an additional 2 neurons switched to the opposite cluster. Therefore, it seems likely that, in general, the responses of LHPV neurons are either not consistent over time or dependent on the type of stimulus applied. As such, we think that the bulk activation of LHPV neurons with optogenetics or chemogenetics likely activates more than the necessary number of them (hot responders + cold responders) to evoke uniform antinociception effects in a given assay, as opposed to having populations of LHPV neurons that are exclusively pronociceptive or antinociceptive, the effects of which could be potentially diluted during bulk activation of these neurons. We have added a narrative for these points in the Discussion section (Discussion, page 23 – 24, lines 459 – 476).

– Are there dual projections to PAG and LHab (if no data, then a thorough acknowledgment and discussion of this issue)?

It is still unknown whether LHPV axonal projections to their target regions follow a one-to-one or one-to-many architecture. This is certainly an important question that has yet to be determined. We thank the reviewers for suggesting various ways to address this point. However, these concerns seem to have mainly arisen from the assumption that certain LHPV neurons consistently comprise the cluster 1 or cluster 2 archetype. Our cell registration analysis now demonstrates that this is not the case. Furthermore, our behavioral data suggest that these might be independent populations since we did not observe aversive-like effects during LHPV→vlPAG stimulation or antinociception during LHPV→LHb stimulation. Future experiments will be needed to firmly draw this conclusion, combined with elucidating the functional roles of each of these potentially independent populations. Of note, we have added some significant discussion regarding this issue (Discussion, page 21 – 22, lines 426 – 433).

– Concerns over mixed effects of Arch on synaptic transmission.

We indeed acknowledge the findings of Mahn et al. 2016 on silencing thalamocortical synapses with archaerhodopsin (eArch3.0). This study is commonly cited to discourage the use of eArch3.0 at axonal projections. We are certainly aware of the biophysical limitations of optogenetic inhibition at terminals. However, during that study Mahn and colleagues concluded that halorhodopsin (eNpHR3.0) is the most suitable tool for silencing synaptic terminals even though eNpHR3.0 also shows strong light-off rebound responses. Thus, when outlining our experiments, we checked whether other rigorous studies in the hypothalamus and other brain regions used eArch3.0 as silencer. For example, Jennings et al., 2013 used eArch3.0 to inhibit BNSTVGAT neurons projecting to the lateral hypothalamus. While their validation experiments tested 5 s photoinhibition, the behavioral experiments used 10 min of constant photoinhibition. This robustly suppressed feeding behavior in food-deprived mice without affecting food intake during the light-off periods. Of note, their findings have been the precedent of many subsequent impactful primary research and review articles. Another remarkable example is a study showing the recordings of postsynaptic lateral/ventrolateral PAG (l/vlPAG) neurons while photoinhibiting ArchT-expressing dmPFC terminals over a duration of 3 min (Rozeske et al., 2018). Please notice that this was the same duration of photoinhibition that we used in our present study. Additionally, Rozeske and colleagues observed a sustained reduction in l/vlPAG post-synaptic firing rates over the duration of photoinhibition in over half of the recorded neurons (probably due to some neurons that were not connected), and this finding is strengthened by the inclusion of a channelrhodopsin (ChR2) group which showed the opposite direction of effect. In further support, their behavioral experiments showed that photostimulation and photoinhibition of the mPFC to l/vlPAG pathway produced opposing effects in a contextual fear discrimination paradigm, whereby photostimulation and photoinhibition decreased and increased freezing behavior, respectively. Collectively, these studies demonstrate that archaerhodopsin activation at axonal terminals silences synaptic transmission and evokes the predicted overall inhibitory effect on neuronal circuits in vivo and during behavior.

Once more, we would like to emphasize the biophysical constraints of photoinhibition at terminals but also cite a few other points from our own findings. First, we demonstrated that archaerhodopsin works when used for LHPV somatic inhibition, both during electrophysiological and behavioral experiments (Siemian et al., 2019). Second, archaerhodopsin and channelrhodopsin manipulations evoked opposite behavioral effects when used in either LHPV somas or their axonal terminals in the vlPAG. Furthermore, these effects mirror what we observed during chemogenetic manipulations of these neurons. Together, we hope that the cited previous studies in vivo using archaerhodopsin at axonal terminals in combination with the consistency of our current behavioral findings are sufficient to ease concerns regarding this issue. We have added a discussion for this point and cited both studies Mahn et al. 2016 and Rozeske et al., 2018 to highlight our awareness of this issue (Discussion, page 20, lines 403 – 408).

– Persistent over chronic is preferred for the time scales used here.

– Small sample sizes from low numbers of animals and small effects make the presentation of raw data points all the more desirable.

We thank the reviewers for these suggestions and understand their preference for displaying raw data points. Thus, we have changed the word “chronic” to “persistent” when discussing our experimental findings of pain models lasting less than 21 days. Additionally, for the functional imaging experiments, we have plotted the data for the hotplate and coldplate assays for each individual mouse, along with the contribution from each mouse to cluster 1 and 2 during both tests. We also added the individual data points for the vast majority of behavioral experiments.

– Flesh out intro beyond to introduce the component reactions to noxious stimulation that you will discuss – nocifensive, suppression of everyday activities and the intersection with aversion that eventually is discussed with respect to both LH and LHab.

We appreciate the reviewer’s helpful recommendations to improve our Introduction. We have expanded this section by both commenting on such components and providing more background on LHPV neurons and other key studies that examined the role of the lateral hypothalamus in pain (Introduction, page 3-4, lines 34 – 41, 46 – 51, 56 – 61).

Please see below our point-by-point responses to the individual reviewers’ comments. We sincerely appreciate the reviewers’ valuable comments and suggestions to improve the quality of our work, and we hope that these revisions will now make our work suitable for publication at eLife.

Reviewer #1:

This is an interesting study from the Aponte lab that builds on their prior work that identified lateral hypothalamic parvalbumin-expressing, glutamatergic neurons (LHPV neurons) as regulators of pain behaviors in mice. Here, the authors explore how painful stimuli impact activity of the LHPV neurons in vivo, and further examine the impact of manipulating activity of these neurons in the context of inflammatory and neuropathic pain, as well as the effects on affective aspects of pain.

Using miniscope recordings, the authors find that there are two main populations of LHPV neurons – the majority of which are activated in association noxious stimulation, but a subset that are inhibited.

Studies using chemogenetics and optogentics, all with appropriate control groups, define clearly the effects of activating these LHPV neurons, and show that the send important functional projections to the periaqueductal grey.

The manuscript misses a potentially important opportunity in defining what these different classes (pain-activated vs inhibited) neurons in the LH do, but there is still much that we learn here.

One technical issue that needs to be addressed is around the use of ArchT for inhibition of LH→PAG terminals, as ArchT can have complicated actions at terminals.

The authors identify differences in projections to the lateral habenula vs. PAG, which are interesting and suggest different functional outputs. It will be important to understand if these are unique cell populations, or just dual projections from the same population of neurons in the LH.

Overall, this represents a significant contribution to the field. I commend the authors on the thorough characterization of their behaviors and detailed exploration of the optogenetic stimulation frequency space – which is largely missing in most papers. There are a number of technical questions that should be addressed to further strengthen the manuscript, and it is worth considering whether some of the data presented here would be better reserved for a future publication to make the present manuscript more focused on the main points outlined in the manuscript title.

We appreciate the reviewer’s enthusiastic assessment of our work and helpful suggestions.

Comments for the authors:

1. The imaging studies in figure 1 are intriguing and provide some very interesting insights – but the presentation raises a number of questions.

a. Overall – the data show that there are LHPV neurons that seem to show increased activity to heat and cold, and some that show decreased activity. The authors call this cluster 1 and 2 respectively. This is interesting, and points to the importance of using the miniscope approach to get cellular level resolution, as it appears that simple fiber photometry may have shown no change in activity in response to these stimuli. This should be highlighted.

We have added a narrative in the Discussion section to emphasize the advantage of using a single-photon miniscope, which enables single-cell resolution analysis (Discussion, page 19, lines 366 – 371).

b. Similar trends are seen for hot plate and cold plate. If you look at the single cell level – are neurons that classify as cluster 1 for hot plate also in cluster 1 for cold plate? Presumably the same mice were tested for both assays. This is not clear.

We apologize for this ambiguity. The same three mice were indeed used for both assays. Per request, we used the CellReg pipeline to register cells across sessions. While a fair number of the same cells overall were detected during the hotplate and coldplate tests (n = 33), only 6 neurons remained in the same cluster between both sessions. We also observed 2 additional neurons that flipped from cluster 1 in the hotplate test to cluster 2 in the coldplate test. Thus, it appears that largely different populations of LHPV neurons are modulated by hot and cold noxious stimuli. This information has been added together with mouse-by-mouse breakdown data in Figure 1—figure supplement 2.

c. The z scores show something rather interesting – there is clearly a period of suppressed activity in cluster 1 cells and increased activity in cluster 2 cells preceding contact with the plate. This is evident also in the "control" group. What do the authors think of this? The inclusion of the control groups here was absolutely critical, as it shows that there are indeed differences. Related – the post-hoc analysis states significant differences are illustrated by the bold line – but this is not easy to see. Am I correct in seeing that there is no significant differences in cluster 2 for the cold stimulus vs. control? The manuscript states significant differences, but the posthoc either does not support this or it is not clear from the bold shading of the line in Figure 1l.

We think that the suppressed or increased activity preceding contact with the surfaces may be related to the nature of the Z-score calculation. What we show in the figures is the Z-score extracted from the entire session-long trace of these neurons, averaged across trials to create the peri-stimulus trace. A neuron that greatly increases in activity in confined epochs throughout the test (e.g., when contacting the hotplate) will by rule lower its resting Z-score baseline into negative values. The opposite would be true for neurons that presumably have some baseline amount of activity and pause for confined epochs during the test; the resting Z-score will be pushed into positive values. A resting Z-score of zero would indicate a neuron that either did not have much of a dynamic range during that session or had relatively the same amount of increases and decreases in fluorescence throughout the session. It has become popular to normalize the pre-stimulus period to zero, but we chose not to do this here to depict the traces in a less-processed form. Z-scoring the extracted fluorescent traces was important to normalize responses across neurons (some brighter than others), but to re-zero the Z-score traces would cause the neurons that had a lower resting Z-score to be overweighted in the Z-score space.

Furthermore, the fact that these pre-stimulus phenomena occurred on control trials, but then increase/decrease to around 0 suggests that perhaps there was some expectation of the mice to feel something noxious upon contact with the surface. However, despite this increase from pre-stimulus to post-stimulus during control trials, the neurons do not increase/decrease their activity to the same degree as the noxious trials.

Addressing the significant differences illustrated by bold lines for cluster 2 in Figure 1l—we apologize for this oversight as we realized that the bold segments were not shown on the submitted version of this figure. We corrected the panel in the revised version.

d. The cluster 1 neurons are more prevalent, and the authors comment that the response magnitude is larger than those for cluster 2. This should be supported by statistical comparisons validating this claim.

We greatly appreciate the reviewer’s interest in these data. After rereading our discussion on the response magnitude of each cluster, we realized that our discussion was more appropriate for the initially submitted version of this figure. For that previous version we used unsupervised k-means clustering instead of the thresholding method used in the full submission that was sent to reviewers. Since inclusion into cluster 1 or cluster 2 was defined at thresholds at equivalent absolute values on either side of 0 (e.g., ± 1 Z-score), this causes the magnitude of the clusters to be very similar. Thus, we have removed the section of the sentence comparing the magnitudes between clusters.

e. For the formalin data – given the results from the hot and cold plate, it is difficult to know how to interpret these findings. Were these cells recorded here specifically cluster 1 cells as determined from the hot and cold plate? Again, were these the same mice tested for hot and cold plate? An analysis of these data for formalin separately for neurons defined as cluster 1 and 2 for hot/cold would be much more informative than including all together.

We apologize for this ambiguity. The same three mice were used for the hotplate, coldplate, and formalin tests. The neurons reported for the formalin test are all recorded neurons, not specifically cluster 1 or 2 as determined from the hot and cold plate experiments. As stated above we found only 6 neurons that consistently classified as cluster 1 or 2 across the hot and cold plate tests, suggesting that the responses of these neurons are either dependent on the noxious stimulus or variable over time.

That being said, when we performed CellReg analysis within each mouse across all five tests (i.e., hotplate, coldplate, contralateral formalin, ipsilateral formalin, no injection) we found that neurons were predominantly only detected in 1 session, and that only 1 neuron was detected in all 5 sessions (please see the pie chart in panel g of Figure 1 —figure supplement 2). Thus, our current data again suggest that the neuronal responses are either (a) determined on an individual basis according to the stimulus, (b) variable across time due to GCaMP-expressing cell turnover, or (c) that some combination of these factors occurred. In any case, our current data do not suggest that these neurons can consistently be labeled as “cluster 1” or “cluster 2” in one test and expect similar phenotypes in a different test. Future work should examine the responses to multiple different noxious stimuli within the same test, which would help to decrease the impact of cell turnover and determine whether different neurons in fact do respond to different types of stimuli.

f. The interpretation of formalin responses as "long-term" seems unwarranted. In fact, the response to formalin is ongoing. So – this would suggest sustained responses to ongoing nociception.

We appreciate the reviewer’s attention to phrasing on this section. We have changed the Results section to remove “long-term” (page 7, lines 122 – 123) and instead indicate that the neuronal activity changes by an ongoing stimulus (page 8, line 138).

g. Related – it would be of interest to analyze the formalin data as Ca2+ transients time-locked to nocifensive responses (flinching, lifting, licking). Did you collect videos of the mice that would allow this analysis? Are there also "cluster 1" and "cluster 2" type responses associated with flinching/licking?

This is indeed a great suggestion. We recorded the contralateral formalin behavioral sessions from two out of the three mice presented here. However, we did not proceed with undertaking the suggested type of analysis since the weight of the miniscope appeared to burden the neck muscles of the mice over the duration of the 45 min formalin test. During the first few minutes of the experiment the mice licked their paws as normal untethered mice do. However, they soon began to take breaks to rest their heads, during which time they did not lick their paws or otherwise perform standard formalin-associated behaviors. As such, we did not think that these behavioral events accurately represented the level of chemical irritation the mice were experiencing. For this reason, we simplified this analysis and binned the activity into the well-defined acute/interphase/inflammatory epochs.

For the sake of speculation, it seems fair to assume that at least some neurons would show time-locked responses. However, even in the hot/cold plate tests, the main increase in activity occurs upon contact with the surface, not necessarily upon paw withdrawal. As such, LHPV neuronal activity may precede bouts of licking/flinching at times when the mouse is sensing the buildup of irritation but not yet performing the behavior. Interestingly, these would likely ‘look’ more like cluster 2 neurons when time-locked to the actual behavior, as the fluorescence would be lower post-behavior than pre-behavior. In future work we plan to optimize these types of assays for finer-grained integration of neuronal activity and spontaneous behavior.

2. The authors go on to study the impact of manipulating LHPV activity on behaviors, but sort of ignore the fact that there are these two populations of neurons with respect to activity in response to noxious stimulation. It is unfortunate that the interesting findings of the two different clusters of LHPV neurons – one inhibited by noxious stimulation and one activated – is not built upon here – but is largely ignored. The subsequent studies go about activating or inhibiting all LHPV neurons to test their effects on sensory and affective aspects of pain. The challenge with this approach, based on the imaging data, is that there is a mixture of activity patterns in response to noxious stimuli. For example – heat responses. As you are targeting LHPV neurons globally – you will be hitting both cluster 1 and cluster 2 neurons. I know that it would be a huge task to sort out what is different about these two populations – but this caveat at least needs significant discussion. It would be very interesting to be able to determine what is different about the cluster 1 and cluster 2 neurons. This is not addressed here, and I think may be beyond the scope of this study. However – this should be discussed. Why are some neurons inhibited, some activated? Do these neurons have different projection targets?

We recognize this important point and greatly appreciate the reviewer’s understanding of the challenges to dissect the differences between these two populations. Moreover, we agree that presently, this is beyond the scope of our current study. However, we would like to acknowledge that there is some precedent for observing heterogeneity (increasing/decreasing cells) upon single-cell calcium imaging, but a singular dominant behavioral phenotype driven by bulk activation of all neurons. For example, Jennings et al., 2015 showed that LHGABA neurons respond to food locations in a heterogeneous manner by either increasing or decreasing in activity (Figure 5), and further that while the neurons typically increase in activity during appetitive or consummatory behaviors (Figure 6), sizable populations were found decreasing in activity during such behaviors (Figure S7). Yet, when these neurons are bulk activated, mice eat voraciously (Figure 1-2) and when the neurons are ablated, mice eat less (Figure 3).

In the context of our study, we agree that it would be great to further subdivide our manipulations of LHPV neurons on the basis of the cluster 1 and 2 responses. However, our analysis included with this revision in fact found that the recorded LHPV neurons did not consistently remain in cluster 1 or 2 across the hot and coldplate tests, suggesting that their responses are either dependent upon the noxious stimulus or otherwise somewhat variable on an individual basis across testing. Nevertheless, future work using the same noxious stimuli over multiple sessions may indeed identify constant neurons that classify as cluster 1 or 2. In the event that neurons could be distinguished based on consistent response properties, then as stated, it would require several experiments to determine whether those would be identifiable on the basis of projection field, gene expression, tonic activity patterns, or some other factor, with the caveat that the distinguishing factor would enable us to selectively target one or both clusters for manipulation of neuronal activity using optogenetics or chemogenetics. Moreover, the small number of LHPV neurons and their restricted location within the most lateral part of the lateral hypothalamus makes such “functional dissection” endeavors more challenging than for larger, more broadly distributed, and more overtly heterogeneous neuronal populations. However, we agree that this will be an important question to address in future work. Using retrograde Cre-dependent GCaMP from different output regions to record projection-specific LHPV neurons is one method that comes to mind to elucidate whether these clusters can be segregated based on projection. We have added some additional discussion to take this single cell heterogeneity versus bulk manipulation conflict more into account (Discussion, page 23, lines 459 – 476).

3. Activating the LHPV neurons leads to reductions in responses to painful stimuli, and reversal of pain-suppressed behaviors. Given that the majority of these neurons are activated by painful stimuli, what does this mean for the role of LHPV neurons in pain? You allude to this briefly in the discussion, but this is the major question in my mind in terms of the major conclusions we can make from this nice study. Are LHPV neurons mediating endogenous analgesia during painful stimuli? If so, then this could be mediating a DNIC type response. It would be very interesting to know if you can elicit DNIC in mice, and block this by inhibiting these LHPV neurons.

Our current impression is that LHPV neurons, at least those in cluster 1, become active in response to pain, thus temporarily inhibiting it. When this ability is lost, mice become more sensitive to pain, whereas when this is bolstered, mice become less sensitive to pain. Thus, we would speculate that these neurons are a source of endogenous analgesia. It does not appear to quite be a DNIC response since activation of LHPV neurons does not cause pain (see Results of nesting experiment) or cause robust place aversion (particularly during stimulation of axonal terminals in the vlPAG, which did reduce pain responses). However, future investigations of the role of LHPV neurons in the DNIC response could certainly be warranted.

4. The use of ArchT for photoinhibition of the LHPV → vlPAG terminals is problematic given reports that activation of Arch actually produces increases in spontaneous synaptic release at some terminals. It is imperative that the authors characterize the effects of Arch activation on synaptic transmission (evoked and spontaneous) at this synapse to allow interpretation of the findings.

We indeed acknowledge the findings of Mahn et al. 2016 on silencing thalamocortical synapses with archaerhodopsin (eArch3.0). This study is commonly cited to discourage the use of eArch3.0 at axonal projections. We are certainly aware of the biophysical limitations of optogenetic inhibition at terminals. However, during that study Mahn and colleagues concluded that halorhodopsin (eNpHR3.0) is the most suitable tool for silencing synaptic terminals even though eNpHR3.0 also shows strong light-off rebound responses. Thus, when outlining our experiments, we checked whether other rigorous studies in the hypothalamus and other brain regions used eArch3.0 as silencer. For example, Jennings et al., 2013 used eArch3.0 to inhibit BNSTVGAT neurons projecting to the lateral hypothalamus. While their validation experiments tested 5 s photoinhibition, the behavioral experiments used 10 min of constant photoinhibition. This robustly suppressed feeding behavior in food-deprived mice without affecting food intake during the light-off periods. Of note, their findings have been the precedent of many subsequent impactful primary research and review articles. Another remarkable example is a study showing the recordings of postsynaptic lateral/ventrolateral PAG (l/vlPAG) neurons while photoinhibiting ArchT-expressing dmPFC terminals over a duration of 3 min (Rozeske et al., 2018). Please notice that this was the same duration of photoinhibition that we used in our present study. Additionally, Rozeske and colleagues observed a sustained reduction in l/vlPAG post-synaptic firing rates over the duration of photoinhibition in over half of the recorded neurons (probably due to some neurons that were not connected), and this finding is strengthened by the inclusion of a channelrhodopsin (ChR2) group which showed the opposite direction of effect. In further support, their behavioral experiments showed that photostimulation and photoinhibition of the mPFC to l/vlPAG pathway produced opposing effects in a contextual fear discrimination paradigm, whereby photostimulation and photoinhibition decreased and increased freezing behavior, respectively. Collectively, these studies demonstrate that archaerhodopsin activation at axonal terminals silences synaptic transmission and evokes the predicted overall inhibitory effect on neuronal circuits in vivo and during behavior.

Once more, we would like to emphasize the biophysical constraints of photoinhibition at terminals but also cite a few other points from our own findings. First, we demonstrated that archaerhodopsin works when used for LHPV somatic inhibition, both during electrophysiological and behavioral experiments (Siemian et al., 2019). Second, archaerhodopsin and channelrhodopsin manipulations evoked opposite behavioral effects when used in either LHPV somas or their axonal terminals in the vlPAG. Furthermore, these effects mirror what we observed during chemogenetic manipulations of these neurons. Together, we hope that the cited previous studies in vivo using archaerhodopsin at axonal terminals in combination with the consistency of our current behavioral findings are sufficient to ease concerns regarding this issue. We have added a discussion for this point and cited both studies Mahn et al. 2016 and Rozeske et al., 2018 to highlight our awareness of this issue (Discussion, page 20, lines 403 – 408).

5. The results re: the projections to the LHb are also very interesting. Are the neurons that project to the LHb and vlPAG distinct, or do individual neurons send process to both LHb and vlPAG? A double retrograde tracing experiment would expand the scope of what we learn here significantly.

It is still unknown whether LHPV axonal projections to their target regions follow a one-to-one or one-to-many architecture. This is certainly an important question that has yet to be determined. We thank the reviewer for suggesting a way to address this point. However, our behavioral data suggest that these might be independent populations since we did not observe aversive-like effects during LHPV→vlPAG stimulation or antinociception during LHPV→LHb stimulation. Of note, our LHPV→LHb data also demonstrates that the anticonception evoked by activating the LHPV→vlPAG pathway was not due to antidromic stimulation effects. Future experiments will be needed to determine that these are indeed independent populations of LHPV neurons. We have added some significant discussion regarding this issue (Discussion, page 21 – 22, lines 426 – 433).

Reviewer #2:

This is a thorough, multipronged study which supports a role for parvalbumin-positive neurons in the lateral hypothalamus (LH-Pv+) in modulating the nocifensive and affective / motivational consequences of noxious stimulation and morphine suppression of the same. The authors provide evidence that both nocifensive and affective / motivational consequences are suppressed by activation of glutamatergic LH-PV+ neurons whereas activation of GABAergic LH projections (expressing the leptin receptor) result in place aversion.

Experiments also demonstrate that LH-PV+ activation combines in an additive and at one ratio in a synergistic manner with morphine and that activation of LH-PV+ neurons reverses tolerance to morphine.

The nociceptive-specific nature of the sensory responses is not convincing given that there is a comparison between stimulus and not-stimulus rather than between innocuous stimulus and noxious stimulus. The authors may argue that room temperature is a thermal stimulus and perhaps that is true, depending on the animal's skin temperature which was not measured. But a better comparison would be between a 45C or so stimulus and the 51C used. I am not suggesting that the authors add such expts rather I recommend a more careful interpretation of the data collected. This concern is strengthened by the response to ipsilateral formalin which raises the issue of whether the "responses" may be in fact correlates of efferent function involved in autonomic or state arousal. Yet, the remaining experiments mitigate this concern.

We thank the reviewer for these suggestions and agree that temperature gradient analysis could be used in future studies to determine whether LHPV neurons are technically temperature- or pain-sensing neurons or signaling arousal states. Thus, we have adjusted our wording on the Results section for the functional imaging data by removing several instances of the word “noxious”.

In pairing the (idiosyncratically) preferred side with CNO-inhibition, the preference for either side goes to 50%. Thus inhibition of LH-PV+ neurons takes away side preference but does not produce side avoidance as is stated in lines 153-54.

We appreciate this distinction and understand that “avoidance” of one side technically indicates the outright preference (>50% time) for the other side, and not a lowered occupancy relative to baseline which is how we had used the term. Thus, the wording in this section has been rearranged to discuss the results in terms of losing preference for the initially preferred side (Results, page 10, lines 176 – 180).

Please jog the gray symbols in Figure 4 supp 1, 2 laterally so they can be seen.

As suggested, we have reformatted the symbols on these figures for better visualization of individual data points.

The LH-PV+ preference for vglut2 over vgat PAG neurons is by less than a factor of 3. Given what we know about the unimportance of absolute numbers of synapses in dictating the responses of lemniscal pathways in the thalamus, this result is interesting but with this amount of information uninterpretable. Tone this down.

Determining how LHPV neurons regulate vlPAG microcircuitry is indeed a subject for our future studies. In our previous work and current study, we extensively discussed previous experiments in mice using selective chemogenetic manipulation of neuronal activity in the vlPAG. Such work demonstrated that glutamatergic or GABAergic neurons play opposing roles in nociception and defensive behaviors (Samineni et al., 2017; Tovote et al., 2016). Thus, we would speculate that glutamatergic LHPV neurons modulate nociceptive processing by excitatory control of glutamatergic neurons in the vlPAG to attenuate nociception. Our findings that more vlPAGVGLUT2+ neurons (15%) than vlPAGVGAT (7%) neurons are synaptically targeted by LHPV neurons further support this idea. Alternatively, LHPV neurons may function by activating inhibitory interneurons in the vlPAG that provide local inhibitory control of vlPAGVGAT neurons to suppress nociception. Future experiments will use techniques such as ChR2-assisted circuit mapping (CRACM) from LHPV-ChR2+ axonal projections onto postsynaptic vlPAG neurons followed by single-cell RT-qPCR analysis to elucidate how LHPV neurons regulate vlPAG microcircuitry. We have expanded this narrative in the revised version of the Discussion (Discussion, page 21, lines 419 – 423).

Reviewer #3:

In this study, Siemian and co-authors significantly expand upon their previous work (Siemian et al., 2019) examining the role of parvalbumin-positive glutamatergic neurons in the LH (LHPV), and their projections to the vlPAG, in the modulation of pain responses. In their previous work, the authors showed that optogenetic activation or inhibition of LHPV neurons suppresses or potentiates, respectively, responses to a noxious thermal stimulus and that this effect may be mediated by an excitatory projection to the vlPAG. The current manuscript is exciting in that it describes a much more detailed investigation of this circuit, using in vivo activity monitoring and manipulation (both chemo- and optogenetic) in multiple models of pain and pain-related behavior (both acute and chronic) as well as a neuroanatomical elucidation of this circuitry. Overall, this work uses a wide-spectrum of modern tools to address the circuit-basis of LHPV modulation of pain states and pain-associated behavior. This well-executed work contributes to filling a significant gap in our understanding of the role of LH neurons in modulating pain responses and pain behavior.

In their current work, the authors start by observing in vivo calcium signals with single cell resolution using GRIN-lens endomicroscopy of LHPV neurons in PV-cre mice in several pain assays. They found that a subset of LHPV neurons were activated by an acute thermal stimulus (hotplate), a much smaller subset were inhibited and large proportion did not respond. In another set of experiments, a subset of neurons were found to be activated by an acute noxious cold stimulus (coldplate), a small subset were inhibited and large proportion did not respond. They also found that some LHPV neurons responded to formalin injection over longer time-scales. These experiments successful provide solid evidence that at least a subpopulation of LHPV neurons respond to noxious stimuli which is a key finding. Less clear is how the heterogeneity observed in these responses may be linked to the circuit manipulations in the subsequent experiments, which engage the whole population of LHPV neurons, and how specific LHPV neurons are in modulating pain pathways.

In the next series of experiments, the authors manipulate LHPV neurons using chemogenetic and optogenetic methods in PV-cre mice, in a variety of standard pain assays. They first found that chemogenetic activation of LHPV neurons suppressed pain responses, while inhibition of LHPV neurons potentiated pain responses (increases and decreases in thermal pain thresholds respectively). In an interesting assay of homecage behavior, they further found that chemogenetic activation of LHPV neurons prevented a pain-associated reduction in both nesting behavior and place aversion, which strengthens the argument that activation of LHPV neurons is antinociceptive and seems to diminish pain-associated aversion. Finally, optogenetic activation of LHPV neurons appeared to diminish thermal and inflammatory pain, consistent with their chemogenetic results. However, optogenetic activation of LHPV neurons also produced mild place avoidance in an RTPP assay, somewhat at odds with their chemogenetic results. Overall, the chemo and optogenetic manipulation experiments succeeded in showing that activation of LHPV neurons has antinociceptive effects.

Next, the authors significantly extended their previous work by further probing the role of LHPV projections to both the vlPAG and LHb. Optogenetic stimulation of LHPV inputs to the vlPAG increased both acute thermal pain and more chronic neuropathic pain thresholds while inhibition decreased it, consistent with direct somatic manipulations. Interestingly, these manipulations did not appear to affect place preference leading to their examination of LHPV projections to the LHb, activation of which was aversive, consistent with previous work showing that the broader population of LH-VGLUT2 to LHb pathway is aversive. To address whether excitatory or inhibitory vlPAG populations may be targeted by LHPV neurons, the authors used rabies tracing and found that although a very small percentage of LHPV neurons appeared to innervate the vlPAG, a greater proportion of excitatory than inhibitory neurons were targeted. Finally, the authors found that LHPV-induced antinociception enhanced and morphine-induced antinociception and rescues morphine tolerance.

Overall these data are novel and suggest that even a very tiny cluster of LH neurons has a rather outsized role to play in pain modulation. These findings would contribute to our limited understanding of both the role of hypothalamic circuits in pain modulation and the intersection of reward, aversion and pain.

Comments for the authors:

– Although the in vivo GRIN lens experiments provide evidence that a subpopulation of LHPV neurons respond to noxious stimuli, there's less clarity on the following questions:

1. Does the subpopulation of LHPV neurons activated by noxious thermal stimulation overlap with the subpopulation of neurons activated by noxious cold stimulation, ie is a specific subset of these neurons broadly tuned to respond to noxious thermal and cold stimuli or are they separate? The authors could provide more clarity on whether the cold/hot assays were done in the same animals and if not, provide some commentary on this question.

We apologize for this ambiguity. The same three mice were indeed used for both assays. For better clarification, we used the CellReg pipeline to register cells across the hotplate and coldplate sessions. While a fair number of the same cells overall were detected during the hotplate and coldplate tests (n = 33), only 6 neurons remained in the same cluster between both sessions. We also observed 2 additional neurons that flipped from cluster 1 in the hotplate test to cluster 2 in the coldplate test. Thus, it appears that largely different populations of LHPV neurons are modulated by hot and cold noxious stimuli. This information has been added together with mouse-by-mouse breakdown data in Figure 1—figure supplement 2.

2. What is the degree of inter-animal variability in the data? Did some animals contribute more or less to the population(s) that responded to thermal/cold stimuli?

We appreciate the reviewer’s interest in these data. Thus, in combination with the response above, we have added individual heatmaps depicting neuronal activity per mouse over these two tests, along with pie charts showing the contribution of each mouse to each response cluster (Figure 1 —figure supplement 2).

3. Given that only a relatively small subpopulation of LHPV neurons are activated by noxious stimuli, with another subpopulation inhibited, how does one interpret the chemo- and optogenetic data in which all LHPV neurons are either activated or inhibited? This should be addressed as a caveat in the discussion.

We acknowledge this important point brought up by the reviewer. However, we would like to emphasize that there is some precedent for observing heterogeneity (increasing/decreasing cells) upon single-cell calcium imaging, but a singular dominant behavioral phenotype driven by bulk activation of all neurons. For example, Jennings et al., 2015 showed that LHGABA neurons respond to food locations in a heterogeneous manner by either increasing or decreasing in activity (Figure 5), and further that while the neurons typically increase in activity during appetitive or consummatory behaviors (Figure 6), sizable populations were found decreasing in activity during such behaviors (Figure S7). Yet, when these neurons are bulk activated, mice eat voraciously (Figure 1-2) and when the neurons are ablated, mice eat less (Figure 3).

While a fair number of the same cells overall were detected during the hotplate and coldplate tests (n = 33), only 6 neurons remained in the same cluster between both sessions. We also observed 2 additional neurons that flipped from cluster 1 in the hotplate test to cluster 2 in the coldplate test. Thus, it appears that largely different populations of LHPV neurons are modulated by hot and cold noxious stimuli. This information has been added together with mouse-by-mouse breakdown data in Figure 1 —figure supplement 2.

– The rabies tracing experiment is well-done and informative but is a somewhat indirect way of showing preferential innervation of different vlPAG populations. Directly addressing this is a technically difficult problem and perhaps the subject of a follow-up study but a discussion of the caveats would be helpful. One possible caveat in this experiment is that, for the sake of argument, a simply larger population of excitatory vlPAG neurons may skew the representation of retrogradely-labeled LHPV neurons despite perhaps an equal rate of connectivity with excitatory and inhibitory neurons.

We sincerely appreciate the reviewer’s understanding of the technical challenges to determine the preferential targeting of LHPV axonal projections onto neurons within the vlPAG. We agree that determining how LHPV neurons regulate vlPAG microcircuitry will expand our understanding of how activation of this LHPV→vlPAG pathway modulates nociceptive responses to noxious stimuli. This is indeed a subject for our future studies which will require the use of techniques such as ChR2-assisted circuit mapping (CRACM) from LHPV-ChR2+ axonal projections onto postsynaptic vlPAG neurons followed by single-cell RT-qPCR analysis. As suggested, we have expanded this narrative in the revised version of the Discussion (Discussion, page 21, lines 419 – 423).

– Another outstanding question is whether the same population of LHPV neurons are innervating both the vlPAG (antinociception) and the LHb (aversion) or if they are entirely different subpopulations. One way of approaching this experimentally is to inject two different conjugated CTBs, one in the vlPAG and the other in the LHb to determine if retrogradely-labeled LHPV neurons overlap or not. If this experiment is not possible, addressing this in the discussion would be helpful.

It is still unknown whether LHPV axonal projections to their target regions follow a one-to-one or one-to-many architecture. This is certainly an important question that has yet to be determined. We thank the reviewer for suggesting a way to address this point. We have tried the CTB approach on these neurons without success. However, we would like to highlight that our behavioral data suggest that these might be independent populations since we did not observe aversive-like effects during LHPV→vlPAG stimulation or antinociception during LHPV→LHb stimulation. Of note, our LHPV→LHb data also demonstrate that the antinociception evoked by activating the LHPV→vlPAG pathway was not due to antidromic stimulation effects. Future experiments will be needed to determine that these are indeed independent populations of LHPV neurons. We have added some significant discussion regarding this issue (Discussion, page 21-22, lines 426 – 433).

– In the Discussion, the authors refer to LHPV neurons as a whole in the context of pain modulation when their data actually implicates only a comparatively small subpopulation of these neurons in modulating pain responses, evidenced by their in vivo monitoring and rabies tracing. This language should be adjusted in the discussion given that LHPV neurons are clearly not functionally monolithic. Furthermore, a short discussion of any known heterogeneity among LHPV neurons would be useful here.

We apologize again for this ambiguity. As stated above, we used the CellReg pipeline to register cells across sessions and found that cluster 1 and cluster 2 neurons were variable across sessions. This information has been added together with mouse-by-mouse breakdown data in Figure 1—figure supplement 2. Thus, this argues against constant fractions of LHPV neurons that can always be considered the antinociceptive or pronociceptive clusters, respectively. We would speculate that the response of each individual LHPV neuron may be determined by several factors including the stimulus intensity or modality.

– Finally, in terms of the writing, the manuscript is overall quite well-written.

However, the introduction could be improved. The authors start with a general description of pain and pain behavior and quickly transition to the role of the lateral hypothalamus in pain. It would be useful to make that transition more fluid and more detailed given that this is rationale for studying these cells in the context of pain. This could be accomplished by a more general description of the LH and how the repertoire of behaviors associated with the LH are relevant to pain responses. For example, the section of text: "While the LH circuits controlling food intake and reward have received intense focus over the past several years, those governing nociception have gone understudied by comparison. Thus, with its diverse array of neuronal populations, uncovering genetically defined LH circuits that regulate pain may bring forth a novel therapeutic target" should be expanded upon somewhat and include some key references. At the moment, there are no references here.

We appreciate the reviewer’s helpful recommendations to improve our Introduction. We have added several references surrounding that sentence and have introduced some prior work regarding LHPV neurons to segue from the introduction of lateral hypothalamic circuits into the summary of the current study (pages 3 – 4, lines 46 – 61).

– Also, the description of the role of the LH in pain in the introduction is sparse and should include a slightly more detailed description of the literature with key references. This can be included either as part of the introduction or in the discussion. Also, there are some key references missing here. An example includes a set of papers describing a role for an LH substance P projection to the PAG in mediating antinociception (Holden and Naleway, 2001; Holden et al., 2002; Pizzi and Holden, 2008; Pizzi et al., 2009).

As suggested, we have expanded this part and added these citations to the Introduction.

Reviewer #4:

The miniscope calcium imaging data are excellent. For all imaging studies, summary statistics on which cells or how many cells are from which specific animal should be graphically shown somewhere. From the current display one cannot tell if the active and inhibited cells are from just one animal or are observed in all animals. It would beneficial to see several "per animal" metrics, i.e. showing the data in Figure 1 E and I not as a merge of all neurons across animals but for each individual mouse, as well as show the proportion of active and inhibited cells per subject.

1. The title states "sensory and affective pain", which I read initially as the paper would present data on the dichotomy of the sensory and emotional qualities of pain perception, as originally defined by Melzack 1968. However, the assays used (primarily reflexive tests of hypersensitivity) measure only sensory aspects of behavior. In the text though (in comparison to the title), the authors use "affect" to refer to "pain suppressed behaviors" / "pain induced co-morbidities", which I agree are partly captured by the nest building assay. The title to more accurately reflect the data, and remove "affective pain".

We understand the reviewer’s concern for our use of these terms. However, in addition to the nest building assay we would also suggest that the effects of LHPV neuronal activation on formalin-induced changes in place preference also capture changes in affective pain; this has been interpreted as such by others in previous work (e.g., Alhadeff et al., 2018 Cell, Figure 4: “Hunger attenuates negative affective components of pain”) even though the same manipulation had a primary effect of reducing formalin-induced paw licking (measure of sensory pain; Figure 1: “Hunger attenuates response to inflammatory pain).

We acknowledge that several studies have demonstrated that affective pain can be selectively manipulated without having effects on sensory pain per se (e.g., Johansen et al., 2001; Corder et al., 2019). Yet, our work attempts to emphasize that simply treating sensory pain neither implies that (a) affective/motivational effects of pain are also improved nor that (b) normal behaviors are resumed. Thus, we wanted to highlight that we incorporated the nest building and place preference experiments to specifically examine these issues, but we acknowledge that LHPV neuronal activation likely reduces sensory pain first, which then has secondary effects on affective pain. Clearly, pain is generally multifaceted and its different components interact substantially to evoke a physically painful, emotionally unpleasant experience. However, even some clinical studies have shown that certain manipulations can induce changes in sensory but not affective pain descriptors (Schilder et al., 2018). Thus, we think that to postulate that the effects of LHPV neurons are specific for sensory pain would also not be technically accurate, given that the assays of affective pain did yield positive effects. To avoid confusion, we have changed the title to “An excitatory lateral hypothalamic circuit orchestrating pain in mice.”

2. The miniscope data find two Clusters of cell-types functionally defined by being active or inhibited to noxious stimuli. This result is nice. However, it is disconnected from the remainder of the data, and as currently presented is a very elaborate, but far superior, Fos-like experiment showing that some cells are active while others are not, but it is not clear how and to what function these cell-types contribute to results presented after Figure 1. To avoid a difficult and time-consuming miniscope experiment to image projection neurons to PAG vs. habenula, I suggest a tracing-topology study to link this data to the other very important and interesting results in Figure 6 on differential results seen in the LH projections to the PAG vs habenula. For example, injection of a retrograde-AAV-FlpO recombinase into the PAG (and in a separate cohort, injection into habenula) of Pv-Cre mice with a Cre-ON/Flp-On intersectional fluorophore into the hypothalamus plus a noxious stimulus to induce FOS in these cells. Several conclusions could be drawn to compliment the miniscope functional data and the behavioral data, namely 1) are there collateral projections from LH◊PAG that also project to the habenula; 2) what proportion of these projection cells are nociceptive (FOS+); 3) are there functionally distinct cell-types (e.g. Clusters 1 and 2) that preferentially send projections to PAG or habenula (i.e. Fluorphore+ and FOS+/FOS-). This type of minor experiment might also provide additional insight to understand the two activity Clusters and how the bulk optogenetic/chemogenetic activation of LHPv vlPAG neurons is antinociceptive when this experimental design "turns on" the nociception-inhibited Cluster-2.

We agree that this experiment would yield several interesting findings. However, it is based on the idea that cluster 1 or 2 neurons remain constant over time and across stimuli. For our revision, we performed cell registration across the five calcium imaging sessions (Figure 1—figure supplement 2). We found 33 neurons detected in both the hotplate and coldplate sessions in which we examined time-locked responses. Of those 33 neurons, only 6 neurons remained in a particular cluster in both sessions (e.g., cluster 1 in both the coldplate and hotplate), and an additional 2 neurons switched to the opposite cluster. Therefore, it seems likely that, in general, the responses of LHPV neurons are either not consistent over time or dependent on the type of stimulus applied. Thus, FOS-induced cells would likely be different depending on the applied stimulus, which further limits the interpretability of the proposed experiment if cells cannot rigidly be classified into these cluster designations.

3. I would consider not referring to the 7-day timepoint post Spared Nerve Injury, a model of "chronic neuropathic pain". This is still an acute surgical neuropathy model, whereas most "chronic" designations should be reserved for the 3+ week timepoint, at a minimum.

We appreciate the reviewer’s attention to phrasing on this part. As suggested, we changed these occurrences to “persistent”.

4. No axonal terminals can be seen in the image of Figure 4A nor Supp Figure 3. Are there other images from these mice illustrating that this projection connects with this more posterior portion of the vlPAG? The group's prior 2019 paper shows dense LHPv innervation of the very anterior PAG (superocularmotor region), in contrast to some fiber placements in this article, which are almost 1.0 mm apart. In addition to the new images to confirm axons under the fiber tracts, I would request that images and quantification be provided for LH-PV axon densities across the anterior-posterior axis of the PAG. This will also be very helpful for the reader to link the past work with the current manuscript, as well as make sense of the choice of A-P coordinates for the fibers (-4.0 and -4.8 [Leptin-Cre study]) and for the RABV tracing experiment (which was done at the anterior -3.8 coordinate which was also used in the 2019 paper). Alternatively, performing a patch-clamp experiment (as done by the group in the 2019 paper) of Vglut2 or Vgat PAG neurons in the posterior PAG would confirm this connection, since the provided images do not show any axons in this region.

We apologize for this ambiguity. We wanted to depict a general example for fiber placement given that we have four different groups (i.e., ArchT + GFP control and ChR2 + GFP control). For better clarification, we have included images for LHPV axonal projections in the vlPAG in the revised version.

Related to the vlPAG coordinates for the PvalbCre mice studies, the discrepancy between vlPAG targets of AP −3.87 in the previous study, AP −3.9 for RABV, and AP −4.0 for behavior in the current study was a trivial adjustment in our opinion (≤ 130 µm difference; smaller than standard virus spread). We made this adjustment in order to try to get the optical fibers to be closer to the center of the terminal field in the PAG. The LHPV terminal field is quite longitudinally dispersed. Therefore, the difference of ∼0.1 mm will not make any difference and is not a basis for anterior/posterior delineation of the LHPV terminal field. The requested axonal density study has fortunately already been done by Marco Celio’s group in this publication [Celio et al., 2013 J Comp Neurol. “Efferent connections of the parvalbumin-positive (PV1) nucleus in the lateral hypothalamus of rodents”]. Note especially Figure 9a-d (where axons are observed between AP −3.9 and AP −4.24) and the schematic in Figure 11. Therefore, the AP −4.0 mm coordinate is justified to use.

For the LHLEPR→PAG experiments, we implanted the optical fibers at (AP −4.8 mm) because LHLEPR neurons project to a more posterior area of the PAG compared to the LHPV axonal projections. Of note, we and others previously mapped the axonal projections of LHLEPR neurons (Schiffino et al., 2019; Leinninger et al., 2009). We have included this narrative in the Results and Methods sections on the revised version for better clarification (Results, page 14, line 273; Methods, page 31, lines 556 – 558).

5. Even though it was stated by the authors that it created some clutter, I would still suggest to show all individual dots and lines for behavior throughout the figures (perhaps make the lines 50% transparent)

We have reformatted figures to show individual data points except for (i) the 5-min binned formalin paw licking graph, since the individual data points are plotted in the next panel, (ii) the morphine dose-response curves which contain 5 lines that overlap significantly already, and (iii) the isobolograms which are plotted as mean ± 95% CI.

Reviewer #5:

In 2019 the Aponte lab (10.1038/s41598-019-48537-y) reported that a small cluster of lateral hypothalamic neurons that express the calcium-binding protein parvalbumin (LH-PV neurons) modulate nociception in mice. They showed that photostimulation of these neurons suppresses nociception to an acute, noxious thermal stimulus, and that photoinhibition potentiates thermal nociception. They also showed that these neurons form functional excitatory synapses on neurons in the ventrolateral periaqueductal gray (vlPAG), and that photostimulation of these axons mediates antinociception. Finally, they showed that the anti-nociceptive effect appears to occur independently of opioidergic mechanisms. Many of these findings are replicated here.

In the present study they add to these findings by demonstrating, with the use of calcium imaging from behaving mice, that the LH-PV neurons respond to noxious stimuli. They also demonstrate that projections of these neurons to vlPAG affect both sensory and affective aspects of pain, whereas projections to the habenula appear to affect only the affective/aversive components.

The conclusions of this paper are mostly supported by the data, but some detailed aspects could to be clarified (laid out below).

1. Some of the effects reported appear rather small, and some reported differences might be driven by outliers. For example, data in Figure 1 d,g,h,k,l represent changes smaller than 2 standard deviations, or less than one Z score. Data in Figure 4i suggest very small effects on paw withdrawal thresholds. A consideration of whether these small changes are functionally meaningful would be particularly useful. Differences depicted in data in Figure 1 o,p,q appear to driven by a small number of outliers; even if these data survive tests of statistical outliers, one wonders why the vast majority of experiments show no differences.

We respectfully disagree with the statement “the vast majority of experiments show no difference.”

For Figure 1d, these sample traces depict some relatively large (2-3 SD) transients occurring after hotplate contacts, comparable to previous publications showing individual representative traces (see Jennings et al., 2015 Figure 4i).

For Figure 1g,h,k,l, please note that one criterion used to select cells into these clusters were that they achieved ± 1 Z-score, on average across trials, at some point following the stimulus (note the range of the red/blue values in the heatmaps plotting individual neuron activity, range ± 2.5 Z-score). When these traces are averaged within each cluster to plot the activity traces, since the time of the maximal Z-score was asynchronous across neurons, this has the effect of creating a constant Z-score value lower than the absolute maximum of any of the neurons during their time of highest activity.

For Figure 4i, we specifically acknowledged this small effect in the original text – “Due to the modest effects of LHPV→vlPAG activation on mechanical thresholds, we predicted that this pathway may be more effective during inflammatory than neuropathic pain conditions.”

Related to this, it would be useful to know what criteria were used to ensure that parametric analyses are appropriate. It is not clear why, in some comparisons, both Bonferroni and Dunnett's multiple comparisons are used on the same datasets. And, depicting variances as confidence intervals, instead of SEM, will likely be more informative.

We understand the reviewer’s concerns. Our original analysis used a two-way ANOVA, with the repeated measure factor of treatment (i.e., the same cell was tracked across phases of the formalin test). The QQ plot of residuals appeared mostly linear, with a few points at one tail indicating perhaps a slight left skew. Since a non-parametric version of a mixed-model (or even unmatched) two-way ANOVA is unavailable in our statistical software, we proceeded with this report since we felt the matching values over the test was a vital factor and more appropriate than switching to independent tests.

Nonetheless, we have conducted non-parametric analyses, comparing each phase of these formalin test data to control in independent tests (i.e., contralateral vs no injection and ipsilateral vs no injection) with Mann-Whitney non-parametric tests corrected for multiple comparisons with the Holm-Sidak method and found that the results from our two-way ANOVA remain. While contralateral formalin induced significant increases in calcium transients over no injection in the acute (p = 0.048), interphase (p = 0.0078) and inflammatory (p = 0.048) phases, ipsilateral formalin did not. We have substituted these comparisons in the revised version. Thus, the inclusion of the Bonferroni comparison has been removed from that analysis. It was a feature of a previous version of Prism to conduct post-hoc tests between overall group data but that appears to have been recently removed. It was not a critical piece of information regarding the interpretation of these data, and the Dunnett’s post-tests, which were more important, are now essentially replaced by the Mann-Whitney tests.

Of note, we have changed the error bars of the formalin calcium transient dataset to 95% confidence intervals.

Sample sizes are quite small. Although a large number of neurons are depicted, they were collected from a small number (e.g. 3) mice. At the very least, showing data from individual mice (instead of pooling data from all mice) will help determine how reproducible the results are.

The results from the individual mice are now shown in the new supplemental figure (Figure 1 —figure supplement 1), along with the contributions of each mouse to the cluster 1 and cluster 2 designations.

2. Some neurons appear to increase their activity in response to stimuli, whereas others decrease their activity (Figure 1). It would be informative if the authors discuss this intriguing finding.

We found 33 neurons detected in both the hotplate and coldplate sessions in which we examined time-locked responses. Of those 33 neurons, only 6 neurons remained in a particular cluster in both sessions (e.g., cluster 1 in both the coldplate and hotplate), and an additional 2 neurons switched to the opposite cluster. Therefore, it seems likely that, in general, the responses of LHPV neurons are either not consistent over time or dependent on the type of stimulus applied. We have now commented on this finding in the discussion (Discussion, page 23 – 24, lines 459 – 476).

3. The doses of morphine that were effective appear rather high (Figure 7). Discuss?

The doses of morphine used in the current study (i.e., 3.2 – 32 mg/kg) for morphine-alone treatments, are a common range reported throughout the mice literature in assays of acute thermal pain. Of note, we used up to 32 mg/kg as it was necessary for us to obtain 100% MPE to accurately calculate the ED50 values.

1. Miller et al., 2011 Psychopharmacology (Berl) “Effects of morphine on pain-elicited and pain-suppressed behavior in CB1 knockout and wildtype mice” find a near-identical dose-response curve for i.p. morphine in the hotplate test (see Figure 1) as we report in the present study.

2. Neelakantan et al., 2015 Behav Pharmacol “Distinct interactions of cannabidiol and morphine in three nociceptive behavioral models in mice” find a near-identical dose-response curve for i.p. morphine in the hotplate test (see Figure 2g) as we report in the present study.

3. Stone et al., 2014 PLoS One Morphine and clonidine combination therapy improves therapeutic window in mice: synergy in antinociceptive but not in sedative or cardiovascular effects” report < 50% MPE at 10 mg/kg i.p. in mice for the tail-flick assay.

4. Do individual LH-PV neurons project to both vlPAG and habenula? If so, can we exclude the possibility that terminal photostimulation antidromically activated LH neurons and their unintended axonal targets?

If these are independent projections, can the authors discuss how the LH projections to vlPAG vs to habenula might be regulated or balanced during different states?

It is still unknown whether LHPV axonal projections to their target regions follow a one-to-one or one-to-many architecture. This is certainly an important question that has yet to be determined. We thank the reviewers for suggesting various ways to address this point. However, these concerns seem to have mainly arisen from the assumption that certain LHPV neurons consistently comprise the cluster 1 or cluster 2 archetype. Our cell registration analysis now demonstrates that this is not the case. Furthermore, our behavioral data suggest that these might be independent populations since we did not observe aversive-like effects during LHPV→vlPAG stimulation or antinociception during LHPV→LHb stimulation. Of note, we have added some significant discussion regarding this issue (Discussion, page 21 – 22, lines 426 – 433).

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

The revisions have exacerbated a problem in terminology. Pain is a percept. Noxious stimulation is a stimulus. Pain behavior is a package of skeletal muscle-mediated behaviors that ambiguously may include (or not) autonomic reactions. Thus, the title is problematic – you have little information on the pain percept (CPP and that is it). The bulk of your data speaks to pain behavior. And stimuli are not painful until proven so. Stimuli should be described as noxious (if they are indeed in that range).

Many of these issues converge in the sentence "As such, rodent studies searching for new pain interventions have begun to investigate ethological behaviors like nesting that are suppressed by pain (e.g., forgoing standard life activities) as well as the affective/emotional component of pain with assays of pain-induced aversion (e.g., comorbid depression) (5, 11, 12) identifying specific brain pathways capable of managing these multiple components of chronic pain and developing strategies for targeting them for translational use will advance the search for novel pain therapies."

Possible changes include: new pain interventions/therapies → new analgesic interventions therapies; suppressed by pain → suppressed during pain behavior; assays of pain-induced aversion → assays of noxious stimulus-induced aversion; components of chronic pain → components of chronic pain behavior.

We appreciate the importance of semantics in this context and in the preclinical literature at large. We have adjusted the wording throughout the manuscript as suggested above when discussing the results from this study in mice. However, in cases where we allude to future potential treatments or discuss conditions in humans, we have left the word “pain” intact. In addition, we have changed the title to “An excitatory lateral hypothalamic circuit orchestrating pain behaviors in mice.”

The suggestion that "the bulk activation of LHPV neurons with optogenetics or chemogenetics likely activates more than the necessary number of them (hot responders + cold responders) to evoke uniform antinociception effects in a given assay" is not clear. Please explain how this could work. Related to this, if the responses to the noxious stimulation are so easily dismissed in favor of "bulk activation," what are the implications, the import, if any, of these responses? If without import, then why are they shown?

We have revised this section of the discussion as follows:

“Therefore, it seems likely that, in general, the responses of LHPV neurons are either not consistent over time or dependent on the type of stimulus applied. […] However, further work will be needed to elucidate how noxious stimuli are responded to and encoded by LHPV neuronal activity.” (Discussion, page 24)

These findings are important for at least three reasons. First, they demonstrate that responding during noxious stimulation is part of the “natural repertoire” of LHPV neurons. Thus, activating them via optogenetics or chemogenetics likely does not induce some primary effect totally unrelated to nociception (e.g., fear) which then has a secondary effect of suppressing nociception. Rather, suppressing nociception is likely directly affected via these manipulations. Second, they indicate that different types of stimuli engage different individual LHPV neurons. As such, future investigations may link specific LHPV neurons to specific noxious stimuli for targeted antinociception to that specific stimulus. For example, in an arthritis patient, reducing mechanical hypersensitivity (maladaptive) without affecting thermal sensing (normal) could be an ideal way to selectively manage pathological pain. Third, they replicate the broad phenomenon throughout the single-cell literature that even genetically similar neurons do not generally respond in a homogenous manner. We elaborated on the divergence of LHVGAT neuronal responses to food-oriented behaviors in the discussion and rebuttal already. Similarly, although arcuate hypothalamic AGRP neurons are commonly discussed to uniformly decrease in neuronal activity upon access to food (when using fiber photometry; Su et al., 2017; Beutler et al., 2020; Mazzone et al., 2020), the only manuscript that directly measured AGRP neuronal activity using in vivo extracellular electrophysiology showed that ∼30% of AGRP neurons do not show a decrease in activity upon presentation of food/feeding (Mandelblat-Cerf et al., 2015), a finding also somewhat visible when using a single-photon miniscope to record AGRP neuronal activity (Betley et al., 2015). Likewise, neurons in the MnPO that control fluid homeostasis respond in a heterogeneous manner to oral fluid access or intragastric infusion (Zimmerman et al., 2019). Thus, we do not think our findings here on the responses of LHPV neurons are at all discordant with the general findings of how neuronal populations respond to certain stimuli.

https://doi.org/10.7554/eLife.66446.sa2

Article and author information

Author details

  1. Justin N Siemian

    Neuronal Circuits and Behavior Unit, National Institute on Drug Abuse Intramural Research Program, National Institutes of Health, Baltimore, United States
    Contribution
    Conceptualization, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3337-7333
  2. Miguel A Arenivar

    Neuronal Circuits and Behavior Unit, National Institute on Drug Abuse Intramural Research Program, National Institutes of Health, Baltimore, United States
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Sarah Sarsfield

    Neuronal Circuits and Behavior Unit, National Institute on Drug Abuse Intramural Research Program, National Institutes of Health, Baltimore, United States
    Contribution
    Resources, Formal analysis, Investigation, Visualization, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0527-8154
  4. Cara B Borja

    Neuronal Circuits and Behavior Unit, National Institute on Drug Abuse Intramural Research Program, National Institutes of Health, Baltimore, United States
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  5. Lydia J Erbaugh

    Neuronal Circuits and Behavior Unit, National Institute on Drug Abuse Intramural Research Program, National Institutes of Health, Baltimore, United States
    Contribution
    Software, Formal analysis, Writing - review and editing
    Competing interests
    No competing interests declared
  6. Andrew L Eagle

    Department of Physiology, Michigan State University, East Lansing, United States
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  7. Alfred J Robison

    Department of Physiology, Michigan State University, East Lansing, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
  8. Gina Leinninger

    1. Department of Physiology, Michigan State University, East Lansing, United States
    2. Institute for Integrative Toxicology at Michigan State University, East Lansing, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Project administration, Writing - review and editing
    Competing interests
    No competing interests declared
  9. Yeka Aponte

    1. Neuronal Circuits and Behavior Unit, National Institute on Drug Abuse Intramural Research Program, National Institutes of Health, Baltimore, United States
    2. The Solomon H. Snyder Department of Neuroscience, Johns Hopkins University School of Medicine, Baltimore, United States
    Contribution
    Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Validation, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    yeka.aponte@nih.gov
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5967-2579

Funding

National Institute on Drug Abuse (Intramural Research Program)

  • Justin N Siemian
  • Miguel A Arenivar
  • Sarah Sarsfield
  • Cara B Borja
  • Lydia J Erbaugh
  • Yeka Aponte

National Institute of Diabetes and Digestive and Kidney Diseases (P30-DK020572)

  • Andrew L Eagle
  • Alfred J Robison
  • Gina Leinninger

National Institute of Diabetes and Digestive and Kidney Diseases (RO1-DK103808)

  • Gina Leinninger

National Institute of Mental Health (R01-111604)

  • Andrew L Eagle
  • Alfred J Robison

National Institute on Drug Abuse (R01-040621)

  • Andrew L Eagle
  • Alfred J Robison

Eunice Kennedy Shriver National Institute of Child Health and Human Development (R01-072968)

  • Andrew L Eagle
  • Alfred J Robison

National Institute of Neurological Disorders and Stroke (R01-085171)

  • Andrew L Eagle
  • Alfred J Robison

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors acknowledge with gratitude C Lupica for discussions and comments on the manuscript, MG Myers Jr for kindly providing the LeprCre mice, BT Laing for performing SNI surgeries, T Larson for assistance with imaging, the NIDA IRP Histology Core, in particular L Shen and C Mejias-Aponte, for technical assistance with histology, and NIDA IRP Visual Media, in particular A Russell and L Brick, for brain slice drawings. Mouse clip art was adapted from Openclipart.org (Creative Commons CC0). Modified rabies tracing vectors were graciously provided by the Michigan Diabetes Research Core, funded by NIH P30-DK020572. AJ Robison and AL Eagle are supported by NIMH R01-111604, NIDA R01-040621, NICHD R01-072968, and NINDS R01-085171. GM Leinninger is supported by NIDDK RO1-DK103808. Y Aponte is supported by the National Institute on Drug Abuse Intramural Research Program (NIDA IRP), U.S. National Institutes of Health (NIH).

Ethics

Animal experimentation: All experimental protocols were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and with the approval of the National Institute on Drug Abuse and Michigan State University Animal Care and Use Committees. All of the animals were handled according to approved institutional animal care and use committee protocols (NIDA 19-CNRB-116, 19-CNRB-127, and 20-CNRB-132; MSU 201900103). Surgeries were performed under either isoflurane or ketamine/xylazine anesthesia, and every effort was made to minimize suffering.

Senior Editor

  1. Michael Taffe, University of California, San Diego, United States

Reviewing Editor

  1. Peggy Mason, University of Chicago, United States

Reviewers

  1. Robert Gereau, Washington University in St Louis, United States
  2. Peggy Mason, University of Chicago, United States
  3. Alexander C Jackson, University of Connecticut, United States
  4. Gregory Corder
  5. Asaf Keller

Publication history

  1. Received: January 11, 2021
  2. Accepted: May 12, 2021
  3. Version of Record published: May 27, 2021 (version 1)

Copyright

This is an open-access article, free of all copyright, and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CC0 public domain dedication.

Metrics

  • 864
    Page views
  • 92
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Neuroscience
    Xiaoxuan Jia et al.
    Research Article

    Temporal continuity of object identity is a feature of natural visual input, and is potentially exploited -- in an unsupervised manner -- by the ventral visual stream to build the neural representation in inferior temporal (IT) cortex. Here we investigated whether plasticity of individual IT neurons underlies human core-object-recognition behavioral changes induced with unsupervised visual experience. We built a single-neuron plasticity model combined with a previously established IT population-to-recognition-behavior linking model to predict human learning effects. We found that our model, after constrained by neurophysiological data, largely predicted the mean direction, magnitude and time course of human performance changes. We also found a previously unreported dependency of the observed human performance change on the initial task difficulty. This result adds support to the hypothesis that tolerant core object recognition in human and non-human primates is instructed -- at least in part -- by naturally occurring unsupervised temporal contiguity experience.

    1. Neuroscience
    Nick Taubert et al.
    Research Article

    Dynamic facial expressions are crucial for communication in primates. Due to the difficulty to control shape and dynamics of facial expressions across species, it is unknown how species-specific facial expressions are perceptually encoded and interact with the representation of facial shape. While popular neural network models predict a joint encoding of facial shape and dynamics, the neuromuscular control of faces evolved more slowly than facial shape, suggesting a separate encoding. To investigate these alternative hypotheses, we developed photo-realistic human and monkey heads that were animated with motion capture data from monkeys and humans. Exact control of expression dynamics was accomplished by a Bayesian machine-learning technique. Consistent with our hypothesis, we found that human observers learned cross-species expressions very quickly, where face dynamics was represented largely independently of facial shape. This result supports the co-evolution of the visual processing and motor control of facial expressions, while it challenges appearance-based neural network theories of dynamic expression recognition.