1. Neuroscience

The cytokine GDF15 signals through a population of brainstem cholecystokinin neurons to mediate anorectic signalling

  1. Amy A Worth
  2. Rosemary Shoop
  3. Katie Tye
  4. Claire H Feetham
  5. Giuseppe D'Agostino
  6. Garron T Dodd
  7. Frank Reimann
  8. Fiona M Gribble
  9. Emily C Beebe
  10. James D Dunbar
  11. Jesline T Alexander-Chacko
  12. Dana K Sindelar
  13. Tamer Coskun
  14. Paul J Emmerson
  15. Simon M Luckman  Is a corresponding author
  1. Faculty of Biology, Medicine and Health, The University of Manchester, United Kingdom
  2. Rowett Institute, University of Aberdeen, United Kingdom
  3. School of Biomedical Sciences, The University of Melbourne, Australia
  4. Institute of Metabolic Science, Addenbrooke’s Hospital, United Kingdom
  5. Lilly Research Laboratories, Eli Lilly & Company, United States
https://doi.org/10.7554/eLife.55164
Share this article
Cite this article
  1. Amy A Worth
  2. Rosemary Shoop
  3. Katie Tye
  4. Claire H Feetham
  5. Giuseppe D'Agostino
  6. Garron T Dodd
  7. Frank Reimann
  8. Fiona M Gribble
  9. Emily C Beebe
  10. James D Dunbar
  11. Jesline T Alexander-Chacko
  12. Dana K Sindelar
  13. Tamer Coskun
  14. Paul J Emmerson
  15. Simon M Luckman
(2020)
The cytokine GDF15 signals through a population of brainstem cholecystokinin neurons to mediate anorectic signalling
eLife 9:e55164.
https://doi.org/10.7554/eLife.55164
  2. Download BibTeX
  3. Download .RIS

Abstract

The cytokine, GDF15, is produced in pathological states which cause cellular stress, including cancer. When over expressed, it causes dramatic weight reduction, suggesting a role in disease-related anorexia. Here, we demonstrate that the GDF15 receptor, GFRAL, is located in a subset of cholecystokinin neurons which span the area postrema and the nucleus of the tractus solitarius of the mouse. GDF15 activates GFRALAP/NTS neurons and supports conditioned taste and place aversions, while the anorexia it causes can be blocked by a monoclonal antibody directed at GFRAL or by disrupting CCK neuronal signalling. The cancer-therapeutic drug, cisplatin, induces the release of GDF15 and activates GFRALAP/NTS neurons, as well as causing significant reductions in food intake and body weight in mice. These metabolic effects of cisplatin are abolished by pre-treatment with the GFRAL monoclonal antibody. Our results suggest that GFRAL neutralising antibodies or antagonists may provide a co-treatment opportunity for patients undergoing chemotherapy.

Introduction

The cytokine, GDF15 (a member of the TGF-β cytokine family, also known as MIC-1 and NAG-1), is expressed in several tissues throughout the body and circulates in the bloodstream of healthy humans (Bootcov et al., 1997; Tsai et al., 2018; Patel et al., 2019). Plasma levels increase dramatically in a number of pathological states associated with cellular stress, including cancers, cardiac failure, chronic kidney disease, infection and obesity (Patel et al., 2019; Welsh et al., 2003; Kempf et al., 2007; Ho et al., 2013; Bauskin et al., 2006; Luan et al., 2019). Furthermore, over expression of GDF15 causes a dramatic reduction in food intake and weight loss (Tsai et al., 2018; Johnen et al., 2007; Macia et al., 2012; Chrysovergis et al., 2014; Xiong et al., 2017). Together, this has led to the supposition that GDF15 does not have a normal, physiological role but is, instead, secreted as an adaptive response to disease (Patel et al., 2019; Luan et al., 2019). This said, the transgenic knock out of GDF15 from the germline results in obesity, which could be interpreted that the cytokine has an alternative physiological function to regulate body weight (Tsai et al., 2013; Low et al., 2017; Tran et al., 2018).

The GDF15 receptor, GFRAL (GDNF-family receptor α-like) is located exclusively in a small population of cells in the in the area postrema (AP) and nucleus of the tractus solitarius (NTS) of the mouse dorsomedial medulla oblongata, (Mullican et al., 2017; Yang et al., 2017; Emmerson et al., 2017; Hsu et al., 2017) a brainstem region containing a number of characterised neurons that previously have been linked with appetite regulation (Luckman, 1992; Rinaman et al., 1993; Larsen et al., 1997; Lawrence et al., 2000; Luckman and Lawrence, 2003; Ellacott et al., 2006; D'Agostino et al., 2016; Roman et al., 2016; Frikke-Schmidt et al., 2019). Administration of recombinant GDF15 reduces food intake, but not in GFRAL knock-out mice (Mullican et al., 2017; Yang et al., 2017; Emmerson et al., 2017; Hsu et al., 2017). It also induces the cellular activation marker, Fos protein, in GFRAL-positive cells in the AP/NTS, and in putative downstream targets in the pons and amygdala, (Xiong et al., 2017; Hsu et al., 2017; Frikke-Schmidt et al., 2019), whereas selective surgical lesioning of the AP/NTS blocks the anorexic effects of the peptide (Tsai et al., 2014). The absolute identity of the primary responsive neurons has not been determined. Although a small number of GFRAL-positive neurons contain immunoreactivity for the catecholaminergic marker, tyrosine hydroxylase (TH), (Yang et al., 2017) relatively few TH-positive neurons are activated by exogenous GDF15 (Tsai et al., 2014). We have extended these investigations and found that the highest proportion of GFRAL-positive neurons contain the neuropeptide transmitter, cholecystokinin (CCK). We demonstrate that GFRAL cells are a sub-population of CCK neurons, which respond to administration of GDF15 or the cancer therapeutic drug, cisplatin, but not to other anorectic signals. Additionally, the effect of GDF15 to inhibit food intake is abrogated by the targeted deletion of CCK-containing neurons in the AP/NTS or by pre-administration of a CCK receptor antagonist. A single injection of GDF15 causes marked conditioned taste and place aversions, suggesting a strong negative affect, as well as activating downstream pathways previously described as mediating anorexia. Lastly, since we can block GDF15- or cisplatin-induced anorexia by neutralising the GFRAL receptor with a selective monoclonal antibody, we suggest that the anorectic responses to disease and, potentially, their therapeutic treatment may be mediated by this distinct signalling pathway.

Results

GFRAL is localised to CCK-positive neurons in the AP/NTS

Mice expressing Cre recombinase under the control of neuropeptide genes, were crossed with a reporter mouse expressing enhanced Yellow Fluorescent Protein in a Cre-dependent fashion (Rosa26-eYFP), so that the double mutants expressed eYFP in discrete populations of CCK (Taniguchi et al., 2011), glutamate (VGlut2; gene Slc17a6), preproglucagon (PPG; gene Gcg) (Parker et al., 2012) or prolactin-releasing peptide (PrRP; gene Prlh) (Dodd et al., 2014) neurons in the AP/NTS. Using antibodies against GFRAL, eYFP and TH, we were able to characterise putative, GDF15-sensitive cells. In the mouse, GFRAL was expressed in a continuous grouping of cells in the AP, extending into the medial region of the NTS (Figure 1A), as noted previously by others (Mullican et al., 2017; Yang et al., 2017; Emmerson et al., 2017; Hsu et al., 2017). Although dense within the AP and more sparsely distributed in the NTS, there were similar numbers of GFRAL neurons in the two structures (20 ± 1 per section in both the AP and the NTS). 60% of GFRAL-immunoreactive cells in the AP co-localised with CckCre::eYFP, though this proportion was 31% in the NTS (Figure 1B and summary in Supplementary file 1). We confirmed this pattern using RNAScope in situ hybridisation histology, which provided slightly higher values of 69% and 35% overlap in the AP and NTS, respectively (Figure 1—figure supplement 1A and Supplementary file 1). Thus, overall, the majority of GFRAL neurons are CCKergic. By comparison, dual immunohistochemistry revealed that 27% of GFRAL neurons in the AP and 45% in the NTS contained TH (Figure 1B). Overall, approximately, 15% of GFRAL neurons contained both CCK and TH (Figure 1—figure supplement 1B). At least 54% GFRAL neurons in the AP co-localised with Slc17a6Cre::eYFP, which fits with the consensus that CCK neurons in this brain region are glutamatergic (Figure 1—figure supplement 1C). Finally, GFRAL cells were distinct from other NTS populations which contain either PrRP or PPG (Figure 1B), which themselves either form a separate sub-population of TH neurons (Dodd and Luckman, 2013) or overlap with other CCK cells (Garfield et al., 2012; Figure 1—figure supplement 1D), respectively. In conclusion, GFRAL cells appear to form a distinct subset of CCK neurons, a proportion of which also contain TH.

Figure 1 with 1 supplement see all
Download asset Open asset
GFRAL-positive neurons in the AP and NTS co-localise with CCK.

(A) Schematic describing the distribution of GFRAL-immunoreactive cell bodies in the AP and dorsal NTS at different rostrocaudal levels relative to bregma. (B) Dual-fluorescence labelling for GFRAL (magenta) with TH or eYFP (staining using antibody raised against green fluorescent protein) in three reporter mice, CckCre::eYFP, PrlhCre::eYFP or GcgCre::eYFP. GFRAL co-localised with CCK and TH, but not PrRP or PPG (the latter being located more caudal to the majority of GFRAL neurons). White arrows in higher magnification inset indicate co-labelled cells. AP (area postrema), cc (central canal), DMX (dorsal motor nucleus of the tenth cranial nerve, vagus), NTS (nucleus of the tractus solitarius).

GDF15 produces anorexia and a negative affective valence

A single injection of GDF15 (2–8 nmol/kg, subcutaneous; s.c.) at lights-out produced a significant and dose-dependent decrease in normal, night-time feeding within 2 hr (hr) of administration (Figure 2A). Cumulative food intake had mostly recovered by 24 hr. This is within the range used by others who report dose-dependent effects of single, systemic injections of GDF15 in mice (Patel et al., 2019; Mullican et al., 2017). Doses of 4–8 nmol/kg were required to see a significant effect on fast-induced, day-time feeding (Figure 2B), and were the doses used in later experiments. Administration of GDF15 was associated with a negative affective valence, since a single injection supported a strong conditioned taste aversion (CTA) when paired with sucrose (see also Patel et al., 2019), and a conditioned place aversion (CPA) in mice (Figure 2C and D). In addition, pica behaviour, in which kaolin clay is consumed to remedy gastric malaise, was used as a measure of sickness given that rodents cannot vomit. Administration of GDF15 to rats on 3 consecutive days induced pica behaviour – a proxy for sickness behaviour – to an extent comparable to that seen following injection of the nausea-inducing toxin, lithium chloride (LiCl; Figure 2—figure supplement 1).

Figure 2 with 1 supplement see all
Download asset Open asset
GDF15 produces anorexia and a negative affective valence.

(A) Subcutaneous administration of GDF15, just before ‘lights out,’ decreased normal, night-time feeding (n = 6 per group; *p<0.05, **p<0.01, ***p<0.001, compared with 0 nmol/kg group; two-way ANOVA followed by Tukey’s multiple comparison test). (B) GDF15 also decreased fast-induced, day-time re-feeding (n = 5–6 per group; **p<0.01, ***p<0.001, compared with 0 nmol/kg group). (C) GDF15 supported a conditioned taste aversion in mice when paired with sucrose as the conditioned stimulus. Data show two-bottle fluid intake 24 hr following a single conditioning to GDF15 (n = 6 per group; ***p<0.001, water versus sucrose intake for saline- and GDF15-treated groups; two-way ANOVA followed by Sidak’s multiple comparison test). (D) GDF15 supported a conditioned place aversion in mice. Mice showed a preference for one side of the arena measured as time spent (seconds) in preferred side. During conditioning, mice received an injection of GDF15 on their preferred (dark) side and saline on their non-preferred side. On the test day, the mice displayed a decreased preference for the side on which they received GDF15 (n = 12; **p<0.01, time spent in preferred side; paired t-test).

Together, these and recently published data (Borner et al., 2020a; Borner et al., 2020b) suggest that GDF15 is probably not a natural satiety factor, but exerts a pathophysiological action to cause anorexia. This conclusion is supported by the findings that circulating GDF15 levels do not correlate with meal times in humans, (Patel et al., 2019; Tsai et al., 2015) and that GDF15 knock out in mice does not result in significant changes in normal chow intake (Tsai et al., 2013; Tran et al., 2018).

GDF15 activates CCKAP/NTS neurons

Next, we determined the identity of GDF15-activated neurons by carrying out Fos-activity mapping in CckCre::eYFP, PrlhCre::eYFP, GcgCre::eYFP or PomceGFP mice. A single, low-anorectic dose of GDF15 activated GFRAL+ve/CCK and TH neurons in the AP/NTS, as well as GFRAL-ve neurons at the same level of both the AP and NTS (Figure 3A; Figure 3—figure supplement 1A and B). Neither PrRP, PPG nor POMC neurons are activated significantly by GDF15 (Figure 3—figure supplement 1B). Thus, a very obvious group of GFRAL-ve cells in the medial NTS, that are activated by GDF15, remain unidentified (Figure 3A, arrow head). GFRAL cells are not activated by natural satiety signals acting after meal intake (Figure 3—figure supplement 1C); that is, following fast-induced re-feeding, the infusion of lipid directly into the stomach (300 µl Intralipid by gavage) or a low-dose of satiety-inducing, systemic CCK (6 μg/kg body weight, intraperitoneal; i.p.). Perhaps surprisingly, only a small number are activated by LiCl (128 mg/kg body weight, i.p.). This contrasts with the general activation profile of CCK neurons in the AP/NTS, which respond to different anorectic stimuli and which underlines that they are probably a mixed population (D'Agostino et al., 2016; Roman et al., 2017).

Figure 3 with 1 supplement see all
Download asset Open asset
GDF15 activates GFRAL+ve/CCK neurons in the AP/NTS.

(A) Fluorescence photomicrographs showing Fos expression (magenta) in GFRAL and CCK-positive neurons (green) in the AP/NTS following a minimal effective dose of GDF15. For triple labelling, see Figure 3—figure supplement 1A. The percentage of activated GFRAL-immunopositive or CckCre::eYFP neurons is presented on the right (n = 6–7 per group). White arrows in higher magnification insets indicate co-labelled cells. Note that GDF15 administration activated a group of cells in the medial NTS which are GFRAL-ve and CCK-ve (large arrow head). (B) Dual-label immunofluorescence for Fos and downstream neuronal targets. CGRP neurons were visualised by injecting CalcaCre mice with AAV-DIO-mCherry (Fos green; CGRP magenta). In the other pictures, PKC-δ+ or CrhCre::eYFP cells are coloured green. Quantification is provided below the relevant photomicrographs (CalcaCren = 4 per group; CrhCren = 6–7 per group; PKC-δ+n = 6–7 per group). (C) GFRAL neurons, which were activated by GDF15, project directly to the parabrachial nucleus, as demonstrated using Fluoro-Gold retrotracing. White arrows in higher magnification insets indicate triple-labelled cells. aca (anterior part of the anterior commissure), AP (area postrema), BLA (basolateral amygdala), ovBNST (bed nucleus of the stria terminalis, oval sub-nucleus), cc (central canal), CeA (central nucleus of the amygdala), DMX (dorsal motor nucleus of the tenth cranial nerve, vagus), ic (internal capsule), LV (lateral ventricle), NTS (nucleus of the tractus solitarius), PBN (parabrachial nucleus), PVH (paraventricular nucleus of the hypothalamus), scp (superior cerebellar peduncle), 3V (third ventricle). *p<0.05, **p<0.01, ***p<0.001; unpaired t-test.

In terms of potential downstream mediators of the GDF15 signal, in addition to the non-GFRAL cells in the NTS and AP, significant increases in Fos staining were recorded in the lateral parabrachial nucleus of the pons (PBN), the paraventricular nucleus of the hypothalamus (PVH), the oval sub-nucleus of the bed nucleus of the stria terminalis (ovBNST) and in the central nucleus of the amygdala (CeA; Figure 3—figure supplement 1D). Using CalcaCre::eYFP or CrhCre::eYFP mice, we show that some neurons in the PBN activated by GDF15 express calcitonin gene-related peptide (CGRP), while many in the PVH, but not the CeA and ovBNST, express corticotrophin-releasing hormone (CRH; Figure 3B and Figure 3—figure supplement 1E). Instead, within the ovBNST and CeA, a large proportion of Fos-positive cells contained PKC-δ immunoreactivity (Figure 3B). CCKNTS neurons project directly to the PBN, including to cells containing CGRP (Roman et al., 2016; Roman et al., 2017); and, CCK1 receptors in the PBN have been proposed to modulate information flow from gut to brain (Mercer and Beart, 2004; Saleh et al., 1997). CGRPPBN neurons respond to a number of anorectic signals and may act as a point of convergence for different signalling pathways, themselves projecting forward to the CeA and elsewhere (Wu et al., 2012; Carter et al., 2013). Therefore, we used dual-fluorescence RNAScope to demonstrate the expression of Cckr1 in the lateral PBN, but found the receptor mRNA in relatively few CGRP (Calca mRNA-expressing) cells (Figure 3—figure supplement 1F). Thus, CGRP neurons are unlikely to be the only target in the PBN for GFRAL neurons.

By repeating our Fos experiment but in mice previously injected with the retrograde tracer, Fluoro-Gold, into the lateral PBN, we demonstrate that both GFRAL and CCK neurons activated by GDF15 project directly to the PBN (Figure 3C). Likewise, CCKNTS neurons also send direct projections to the PVH (D'Agostino et al., 2016). We confirmed this projection pattern using retrograde tracing, however, we showed that almost no GDF15-activated, GFRAL+ve or GFRAL-ve AP/NTS cells project directly to the PVH or to the ovBNST (Figure 3—figure supplement 1G and H). The most parsimonious conclusion is that GFRAL cells activated by GDF15 project directly to the PBN, which then activates downstream targets in the CeA, ovBNST and PVH. GFRAL cells may also synapse locally to activate other neuronal populations, including cells in the medial NTS. A very small number of these synaptically activated cells contain CCK or TH, and because almost none contain either PrRP, PPG or POMC, they may represent another distinct NTS phenotype. These GDF15-activated cells do not project to either the PBN or the ovBNST, but a few do project to the PVH. The others may represent local interneurons or potentially be responding to descending pathways.

Blocking CCK signalling attenuates the anorexia caused by GDF15

To confirm the importance of GFRAL/CCKAP/NTS neurons in mediating the anorectic effects of GDF15, we used a recombinant adeno-associated virus (AAV) expressing a Cre-dependent designer pro-caspase and its activator, the Tobacco Etch Protease (flex-taCasp3-TEVp) to commit CCKAP/NTS neurons to cell-autonomous apoptosis. CckCre::eYFP mice were injected into the AP/NTS with the AAV-caspase or an AAV expressing mCherry to control for viral load and transduction efficiency. Post hoc examination confirmed that the designer caspase achieved an effective ablation of CCKAP/NTS neurons, assessed by immunostaining for eYFP and also GFRAL (Figure 4A and Figure 4—figure supplement 1). Over the 10-week period following viral injection, there was no significant difference in body weight between groups (Figure 4—figure supplement 1). However, whereas mice transduced with control AAV-mCherry responded to GDF15 with a significant decrease in night-time food intake, those bearing cell-specific ablation of CCKAP/NTS neurons showed an abrogated response (Figure 4B). We further injected GDF15 into mice which had been pre-treated with the CCK receptor antagonist, devazepide. Compared with mice receiving a vehicle control injection, those receiving devazepide displayed an attenuated anorectic response to a subsequent single injection of GDF15 (Figure 4C). At early time points, the anorectic effect of GDF15 was reduced by approximately half.

Figure 4 with 1 supplement see all
Download asset Open asset
GDF15-induced anorexia is dependent on CCK signalling.

(A) Injection of AAV-caspase into the AP and dorsal NTS of CckCre::eYFP mice caused a reduction in the number of eYFP cells as assessed by immunohistochemistry (n = 7 per group; **p<0.01, unpaired t-test). (B) CckCre::eYFP mice transduced with control AAV displayed a significant decrease in food intake following GDF15 administration, while those transduced with AAV-caspase showed reduced anorexia (n = 7 per group; *p<0.05; two-way ANOVA, followed by a post hoc Tukey test). (C) Pre-administration of the CCK receptor antagonist, devazepide, attenuated the anorectic response to GDF15 (n = 6 per group; *p<0.05, **p<0.01, ***p<0.001; two-way ANOVA, followed by a post hoc Tukey test).

GFRAL receptor blockade with a monoclonal antibody alleviates the adverse side effects of the cancer therapeutic drug, cisplatin

Finally, the platinum-based therapeutic drug, cisplatin, causes a long-term reduction in food intake, which can have a major contribution to mortality in cancer patients treated with the drug (Dasari and Tchounwou, 2014). There is recent evidence that cisplatin acts through the NTS → CGRPPBN axis in both rats (Alhadeff et al., 2015; Alhadeff et al., 2017) and mice, (Hsu et al., 2017) and the latter study indicates that GFRAL knock-out mice are protected against the anorexic effects of cisplatin. In our hands, a single injection of cisplatin (4 mg/kg, i.p.) at the beginning of the dark phase, led to a reduction in both food intake and body weight which lasted for three days (Figure 5A and B). In the same animals, there was a significant increase in circulating GDF15, which also lasted for between 2 and 3 days (Figure 5C). After 1 day, GDF15 levels were 45 pg/ml and 270 pg/ml in vehicle- and cisplatin-treated mice, respectively. Food intake and weight loss at 2 days correlated directly with the plasma level of GDF15 (Figure 5—figure supplement 1). In a separate experiment, we found also that cisplatin at the same dose activated GFRALAP/NTS neurons, as assessed by Fos immunoreactivity (Figure 5D).

Figure 5 with 1 supplement see all
Download asset Open asset
The anorectic action of the cancer therapeutic drug, cisplatin, is blocked by inhibition of signalling through GFRAL.

(A) A single dose of cisplatin reduced food intake and (B) body weight over the following 3 days (n = 6 per time point; *p<0.05, **p<0.01, ***p<0.001; two-way ANOVA, followed by a post hoc Tukey test). (C) This corresponded with an increase in circulating GDF15 (n = 6 per time point, ***p<0.001, unpaired t-test) and (D) induction of Fos (magenta) in immunopositive GFRAL neurons (green) at 24 hr after administration (n = 5–6 per group; ***p<0.001, unpaired t-test). (E) Three injections of GDF15, on days 2–4, led to a decrease in cumulative food intake and (F) body weight (n = 6 per group; *p<0.05, **p<0.01; repeated measures ANOVA, followed by a post hoc Tukey test, control Ab + GDF15 versus all other groups). The actions of GDF15 were blocked completely by pre-administration of a monoclonal antibody against GFRAL (10 mg/kg) on day 1. The GFRAL mAb had no effect on food intake or body weight by itself. For full data set, using different concentrations of GFRAL mAb, see Figure 5—figure supplement 1B and C. (G) Pre-administration of 10 mg/kg GFRAL mAb the day before, completely blocked the reduction of food intake and (H) body weight caused by cisplatin (n = 5–6 per group; *p<0.05, **p<0.01, ***p<0.001; repeated measures ANOVA, followed by a post hoc Tukey test).

We have shown previously that weight loss induced in rats by a long-lasting recombinant GDF15 protein can be blocked by the single, subcutaneous administration of a selective monoclonal antibody raised against its receptor (GFRAL mAb) (Emmerson et al., 2017). Further, here we show that the reduction in food intake and body weight in mice, caused by three daily doses of native GDF15, is blocked by pre-administration of 10 mg/kg of the GFRAL mAb (Figure 5E and F). Importantly, this high dose of GFRAL mAb which effectively blocks GDF15 signalling, did not affect either food intake or body weight when injected alone, signifying that GFRAL is unlikely to have a role in either satiety signalling or normal energy balance (for full mAb dose-response data see Figure 5—figure supplement 1). Having determined an effective dose of the GFRAL mAb in mice, this was then used to investigate the role of GDF15/GFRAL signalling in mediating the effects of cisplatin. The GFRAL mAb was administered 1 day before cisplatin and, as before, by itself did not affect either food intake (Figure 5G) or body weight (Figure 5H). However, pre-administration of the GFRAL mAb completely blocked the anorectic action of cisplatin (Figure 5G) and prevented the cisplatin-induced body weight loss (Figure 5H).

Discussion

The recent identification of the receptor for GDF15 and its localisation to a small population of cells in the dorsomedial medulla oblongata, an area involved in gut-brain signalling, ignited interest in the cytokine as having a potential role in body-weight regulation (Mullican et al., 2017; Yang et al., 2017; Emmerson et al., 2017; Hsu et al., 2017). Over expression of GDF15 leads to large reductions in both food intake and body weight (Tsai et al., 2018; Johnen et al., 2007; Macia et al., 2012; Chrysovergis et al., 2014; Xiong et al., 2017), whilst the obese phenotype reported for both the GDF15 and GFRAL knock-out mice is supporting evidence for a homeostatic role in normal body-weight regulation and, perhaps, satiety signalling (Tsai et al., 2013; Low et al., 2017; Tran et al., 2018; Mullican et al., 2017; Hsu et al., 2017). Careful examination of these null animals also provides results which would not be expected if this were the case. Thus, increases in food intake are either sex specific or only clearly apparent in mice fed high-energy diet. However, one of the more obvious phenotypes is a significant reduction in overall locomotor activity, which might explain a major part of the obesity (Tsai et al., 2013; Tran et al., 2018). Furthermore, the measurement of plasma levels of GDF15 in humans or mice does not correlate with meal times or nutritional status, which again would argue against it being a circulating satiety factor (Patel et al., 2019; Tsai et al., 2015). By contrast, GDF15 levels are greatly enhanced in a number of disease states, including cancer, cardiovascular disease and chronic kidney disease, pointing towards a more obvious role as a non-physiological anorectic signal (Patel et al., 2019; Welsh et al., 2003; Kempf et al., 2007; Ho et al., 2013; Bauskin et al., 2006; Luan et al., 2019). In fact, the only non-chronic disease situation described, in which GDF15 is up regulated, is hyperemesis gravidarum associated with intense nausea and vomiting in early pregnancy (Fejzo et al., 2018; Petry et al., 2018). Thus, while we will not rule out a physiological role for GDF15, most of the literature seems to point towards it having a more significant role in sickness behaviour. This is strongly supported by our results that show that systemic administration of GDF15 causes robust conditioned taste aversion, conditioned place aversion and pica behaviour, and that blocking endogenous GFRAL signalling with a selective monoclonal antibody does not affect baseline food intake or body weight.

GFRAL-positive neurons span the AP and the inner border with the NTS, a region synonymous with responses to toxins and other nausea-inducing agents, and which lacks a blood-brain barrier (Miller and Leslie, 1994). The AP and NTS are also part of the dorsal vagal complex, which responds to gut-brain signalling. GFRAL is not expressed in the gut or in the vagus nerve, and since GDF15 is still effective at inducing anorexia in vagotomised rats, (Yang et al., 2017) the simplest conclusion is that circulating GDF15 acts directly on GFRAL expressed in the AP/NTS. We and others have argued for the existence of specific neuron types in this region of the brainstem which respond selectively to different sensory modalities (for example satiety versus nausea) in order to reduce food intake (Luckman and Lawrence, 2003; Kreisler et al., 2014). Thus, it is important to identify the phenotype of GFRAL cells. Although evidence has been provided that GFRAL is expressed in TH-positive catecholaminergic neurons, until now this has not been accurately quantified (Yang et al., 2017; Tsai et al., 2014). Here, using a variety of models, we conclude that the major, identifiable population of GFRAL neurons contain the neuropeptidergic transmitter, CCK. GFRAL-positive neurons respond to GDF15, but not with many other stimuli, and the action of GDF15 can be abrogated either by genetically ablating CCK neurons selectively in this region or by blocking CCK transmission. Thus, we propose that the primary target for GDF15 is a distinct population of GFRAL/CCK neurons which span the AP/NTS to engage well-characterised circuitry involved in anorexia and conditioned aversion (Wu et al., 2012; Carter et al., 2013; Chen et al., 2018). We have found that CCK neurons activated by GDF15 project directly to the PBN, and that other downstream targets, in the PVH, CeA and ovBNST are also involved. There is ample evidence for anorectic signals to utilise parallel downstream pathways, but with convergence at specific nodes which we demonstrate are activated by GDF15, including CGRP neurons in the PBN (D'Agostino et al., 2016; Roman et al., 2016; Roman et al., 2017) and/or PKC-δ+ neurons in both the CeA and ovBNST (Cai et al., 2014; Wang et al., 2019). There is still much to learn about these pathways, not least because we have shown that GFRAL neurons are not activated by the archetypal nausea-inducing agent LiCl, nor does GDF15 activate PPG (here) or GLP-1 receptor neurons (Welsh et al., 2003; Frikke-Schmidt et al., 2019), both capable of transmitting nauseous signals. Recently, bacterial or viral infections have been associated with secretion of GDF15 and an adaptive response in order to increase pathogen tolerance (Luan et al., 2019). In this ground-breaking paper, it was demonstrated that GDF15/GFRAL signalling increases triglyceride production by the liver in order to protect tissues, which are dependent on triglycerides for fuel, from metabolic damage due to inflammation. Also, the loss of appetite and weight caused by metformin in diet-induced obese mice is dependent on GDF15/GFRAL signalling, as is the accompanying increase in insulin sensitivity (Coll et al., 2020).

It is interesting to note that increased GDF15 has been measured in the circulation of human cancer subjects (Welsh et al., 2003; Bauskin et al., 2006; Brown et al., 2003) and, recently an NTS → CGRPPBN → CeA/ovBNST axis has been implicated in mediating the anorexia associated with cancer models in mice (Campos et al., 2017). Likewise, work in rats has demonstrated that this pathway appears also to be activated by the platinum-based, cancer therapeutic drug, cisplatin (Alhadeff et al., 2015; Alhadeff et al., 2017). Although in neither case has the primary brainstem neuron been identified (Hsu et al., 2017), it is reported that weight loss caused by cisplatin is reduced in GFRAL knock-out mice, and we now show that this is true also if wild-type mice are pre-treated with a neutralising GFRAL antibody. Although it is yet to be verified, the possibility exists that both a disease state (cancer) and the treatment (cisplatin) may exacerbate anorexia through the same brainstem pathway. If this is the case, then either GFRAL neutralising antibodies or GFRAL antagonists may provide a possible co-treatment opportunity for patients suffering with cancer-related anorexia/cachexia. The caveat to this is that the secretion of GDF15 during cancer, as it appears so for inflammatory infections, is presumably an adaptive response, so blocking GDF15/GFRAL signalling may worsen the disease and other symptoms. GDF15 has been located in different tissues when they become cancerous (Welsh et al., 2003; Buckhaults et al., 2003). If GDF15 has an adaptive systemic effect, then it may be possible to bypass this and, instead, selectively target the central pathways downstream of brainstem GFRAL.

Materials and methods

Key resources table
Reagent type
(species) or resource		DesignationSource or referenceIdentifiersAdditional
information
Genetic reagent (Mus musculus)C57Bl/6J
(Mouse, male)		EnvigoStock #057MGI:2164189
Genetic reagent (Mus musculus)C57Bl/6NHsd
(Mouse, male)		EnvigoStock #044MGI:2161078
Genetic reagent
(M. musculus)		C57Bl/6J
(Mouse, male)		Charles RiverStock #632MGI:3028467
Genetic reagent
(M. musculus)		C57Bl/6J
(Mouse, male)		Janvier labsN/AMGI:2670020
Genetic reagent
(M. musculus)		PomceGFP
(Mouse, male)		Jackson LaboratoriesStock #: 009593MGI:3851684
Genetic reagent
(M. musculus)		Cckires-Cre
(Mouse, male)		Jackson LaboratoriesStock #: 012706MGI:5014249
Genetic reagent
(M. musculus)		Crhires-Cre
(Mouse, male)		Jackson LaboratoriesStock #: 012704MGI:4452101
Genetic reagent
(M. musculus)		Slc17a6ires-Cre
(Mouse, male)		Jackson LaboratoriesStock #: 016963MGI:5300532
Genetic reagent
(M. musculus)		Rosa26-loxSTOPlox-eYFP (Mouse, male)Jackson LaboratoriesStock #: 006148MGI:3621481
Genetic reagent
(M. musculus)		GcgiCre
(Mouse, male)		Parker et al., 2012 PMID:22638549N/AMGI:5432481
Genetic reagent
(M. musculus)		Prlhires-Cre
(Mouse, male)		Dodd et al., 2014 PMID:25176149N/AMGI:5634277
Genetic reagent
(M. musculus)		CalcaCre
(Mouse, male)		Carter et al., 2013 PMID:24121436N/AMGI:5559692
Genetic reagent (Rattus norvegicus)Sprague Dawley (Rat, male)EnvigoStock #: SD-002N/A
Antibodyanti-cFos (Rabbit polyclonal)Santa CruzCat.# SC52
RRID:AB_2106783		Primary antibody (1:500) IHC
Antibodyanti-DS Red (Goat polyclonal)Santa CruzCat.# 33353
RRID:AB_639924		Primary antibody (1:500) IHC
Antibodyanti-GFP (Chicken polyclonal)AbcamCat.# 13970
RRID:AB_300798		Primary antibody (1:2000) IHC
Antibodyanti-GFRAL
(Sheep polyclonal)		ThermofisherCat.# PA5-47769
RRID:AB_2607220		Primary antibody (1:200) IHC
Antibodyanti-GLP1 (Rabbit polyclonal)PenLabsCat.#. T-4363
RRID:AB_518978		Primary antibody (1:2000) IHC
Antibodyanti-PKCδ (Mouse monoclonal)BD BiosciencesCat.#. 610398
RRID:AB_397781		Primary antibody (1:500) IHC
Antibodyanti-TH (Rabbit polyclonal)AbCamCat.# AB112
RRID:AB_297840		Primary antibody (1:2000) IHC
Antibodyanti-TH (Sheep polyclonal)MilliporeCat.# AB1542
RRID:AB_90755		Primary antibody (1:1000) IHC
Antibodyanti-chicken, Alexa Fluor 488 (Donkey polyclonal)Jackson ImmunoResearchCat.# 703-545-155
RRID:AB_2340375		Secondary antibody (1:1000) IHC
Antibodyanti-mouse, Alexa Fluor 594 (Donkey polyclonal)Jackson ImmunoResearchCat.# 715-585-150
RRID:AB_2340854		Secondary antibody (1:1000) IHC
Antibodyanti-rabbit, Alexa Fluor 350 (Donkey polyclonal)Molecular ProbesCat.# A10039
RRID:AB_2534015		Secondary antibody (1:1000) IHC
Antibodyanti-Sheep, Alexa Fluor 350 (Donkey polyclonal)Molecular ProbesCat.# A21097
RRID:AB_10376162		Secondary antibody (1:1000) IHC
Antibodyanti-sheep, Alexa Fluor 594 (Donkey polyclonal)Molecular ProbesCat.# A11016
RRID:AB_10562537		Secondary antibody (1:1000) IHC
AntibodyAnti-GFRAL (Mouse monoclonal)Emmerson et al., 2017 PMID:28846098mIgG1 GFRAL 8A2Subcutaneous injection
(0–10 mg/kg)
Recombinant DNA reagentAAV8-hSyn-DIO-mCherryDr Bryan Roth
Addgene		Cat.# 50459-AAV8N/A
Recombinant DNA reagentAAV5-flex-taCasp3-TEVpDr Nirao Shah
University of North Carolina Vector Core		N/APMID:23663785
Sequence-based reagentGfralAdvanced Cell DiagnosticsCat.# 417021-C3RNAScope mRNA probe
Sequence-based reagentCckAdvanced Cell DiagnosticsCat.# 402271-C1RNAScope mRNA probe
Sequence-based reagentCckr1Advanced Cell DiagnosticsCat.# 313751-C1RNAScope mRNA probe
Sequence-based reagentCalcaAdvanced Cell DiagnosticsCat.# 420361-C2RNAScope mRNA probe
Peptide, recombinant proteinGDF15R and D SystemsCat.# 9279-GD(4 nmol/kg)
Peptide, recombinant proteinStreptavadin 488Jackson ImmunoResearchCat.# 016-540-084
RRID:AB_2337249		(1:1000) IHC
Peptide, recombinant proteinStreptavadin 594Jackson ImmunoResearchCat.# 016-580-084
RRID:AB_2337250		(1:1000) IHC
Commercial assay or kitMouse/rat GDF15 ELISAR and D SystemsCat.# MGD-150N/A
Commercial assay or kitRNAscope Multiplex Fluorescent AssayAdvanced Cell DiagnosticsCat # 323100
Chemical compound, drugDevazepideTocris BioscienceCat.# 2304(1 mg/kg)
Chemical compound, drugLithium chloride (LiCl)SigmaCat # L9650 (mouse)
Cat.# 73036 (rat)		Mouse (128 mg/kg)
Rat (128 mg/kg)
Chemical compound, drugHydroxystibamidine (Fluoro-Gold)Invitrogen, ThermofisherCat.# H228454% in H2O
Chemical compound, drugCisplatinSigma AldrichCat.# PHR1624(4 mg/kg)
Software, algorithmPrismGraphPadRRID:SCR_002798Version 7
Software, algorithmFijiImageJRRID:SCR_002285Version 2.0.0-rc-69/1.52 p
Software, algorithmMicromanagerImageJRRID:SCR_016865Version 1.4.23
Software, algorithmSmartPanlab, Harvard Biosciences/Biochrom LtdRRID:SCR_002852Version 3.0

Animals

Non-transgenic C57Bl/6 mice were obtained from Charles River (Manston, Kent, UK), Envigo (Huntingdon, UK and Indianapolis IN) or Janvier Labs (Le Genest-Saint-Isle, France). PomceGFP, Cckires-Cre, Crhires-Cre (Taniguchi et al., 2011), Slc17a6ires-Cre and Rosa26-loxSTOPlox-eYFP were all purchased from Jackson Laboratories (stock numbers 009593, 012706, 012704, 016963 and 006148, respectively; Bar Harbor, ME). We have described the generation of the PPG (GcgCre) and PrRP (Prlhires-Cre) mice (Parker et al., 2012; Dodd et al., 2014). Calcaires-Cre mice (Carter et al., 2013) were a kind gift from Prof Richard Palmiter (Howard Hughes Medical Institute, University of Washington).

Drugs and viruses

Request a detailed protocol

Recombinant human GDF15 was either made in house as described (Emmerson et al., 2017) or purchased from R and D Systems (Abingdon, UK). GDF15 was initially dissolved in 15 mM HCl, neutralised and diluted in saline. Devazepide was purchased from Tocris Bioscience (Bristol, UK) and dissolved in 40% DMSO. Cisplatin (Sigma-Aldrich, Gillingham, UK) was dissolved directly in saline.

Fluoro-Gold (hydroxystilbamidine, 4% w/v solution in water; Invitrogen, ThermoFisher, MA) was injected into mice anaesthetised with isoflurane (2–3% in oxygen) and placed in a stereotaxic frame. The skull was exposed and holes drilled at the site of injection. Fluoro-Gold was delivered unilaterally via a glass micropipette affixed to a Nanoject II Auto Nanoliter Injector (Drummond Scientific Company, PA) using co-ordinates as determined in the Mouse Brain Atlas: (Paxinos and Franklin, 2004) elPBN, −4.9 mm A/P, −1.4 mm M/L, −3.8 mm D/V from bregma (12 nl); ovBNST, +0.3 mm A/P, −1.0 mm M/L, −4.5 mm D/V (18 nl); PVH, - 0.7 mm A/P; −0.3 mm M/L; −5.5 mm D/V (18 nl). All animals were left to recover for 2 weeks to allow axonal transport before being transcardially perfused (see below).

Viral injections into the AP/NTS were performed as described previously with minor modifications (D'Agostino et al., 2016; D'Agostino et al., 2018). Briefly, 9- to 11-week-old male mice were anaesthetised with a mixture of ketamine and xylazine dissolved in saline (80 and 10 mg/kg, respectively; 10 ml/kg i.p.). Mice were placed in a stereotaxic frame, an incision was made at the level of the cisterna magna, and neck muscles were carefully retracted. Following dura incision, the obex served as reference point for injections with a glass micropipette. AP/NTS coordinates were approximatively 0.2 mm A/P, 0 and ±0.2 mm M/L, −0.2 mm D/V from obex. About 150 nl of virus were delivered during each of the three microinjections. Animals were administered analgesia (5 mg/kg Carprofen, s.c.) for 2 days post-operatively and given a minimum of 14 days recovery before night-time feeding measurement. AAV5-mCherry and AAV5-flex-taCasp3-TEVp were obtained from Addgene (Watertown, MA) and the University of North Carolina Vector Core (Chapel Hill, NC), respectively.

Tissue preparation and histology

Request a detailed protocol

For all immunohistochemical experiments, animals were deeply anaesthetised with 4% isoflurane in oxygen and transcardially perfused with heparinsed saline (20,000U per litre in 0.9% NaCl) followed by 4% paraformaldehyde in 0.1 M phosphate buffer. Brains were dissected and post-fixed overnight at 4°C and then cryoprotected in 30% sucrose. Brains were cut into 30-µm-thick coronal sections using a freezing sledge microtome (Bright 8000, Cambridge, UK) and either processed immediately or stored in cryoprotectant solution at −20°C.

Immunohistochemistry was performed on free-floating sections at room temperature unless stated otherwise. All antibodies are listed in the Key Resource Table. Brain sections were washed in 0.2% Triton X-100 in 0.1 M phosphate buffer and blocked in 5% normal serum for 1 hr, before being incubated in primary antibody (made up in to 1% normal serum) overnight at 4°C. The next day, sections were washed again and then incubated in secondary antibody for 2 hr. Sections were washed and, where biotinylated secondary antibodies were used, incubated for a further hour in streptavidin-conjugated fluorophores diluted in phosphate buffer. Finally, sections were washed in water, mounted onto glass slides, air-dried overnight and coverslipped with ProLong Gold (Thermo Fisher Scientific, MA). Sections were visualised on a Zeiss Axiomanager.D2 upright microscope (Zeiss, Oberkochen, Germany) and images captured using a Coolsnap HQ1 camera (Photometics, AZ) through Micromanager software v1.4.23 (https://imagej.net/Micro-Manager). Specific band pass filter sets for DAPI, FITC and Texas Red were used to prevent bleed through from one channel to the next. All images were processed and analysed in Fiji ImageJ (https://fiji.sc/).

Eight-week-old C57BL/6J mice (n = 3) were anaesthetised by CO2, decapitated, and the brains removed and snap frozen on crushed dry ice. Four or five 10-μm-thick tissue sections at the level of the AP or PBN were collected for RNA in situ hybridisation histology for Gfral (cat#417021-C3) and Cck (cat# 402271-C1) or Cckr1 (cat#313751-C1) and Calca (cat#420361-C2), respectively. mRNA was detected using RNAscope Multiplex Fluorescent Assay reagent kits (Advanced Cell Diagnostics, Inc, Newark, CA), according to the manufacturer’s instructions, at Gubra (Hørsholm, Denmark). Slides were counter stained with DAPI to identify cellular nuclei. Slides were scanned under a 20X objective in an Olympus VS120 Fluorescent scanner.

Feeding and body-weight studies

Request a detailed protocol

For GDF15 night-time feeding experiments, food was removed from the animals for 2 hr before lights out. At lights out, mice were administered GDF15 subcutaneously and food was returned at the same time. Food intake was recorded at 0, 1, 2, 4, and 24 hr after injection of GDF15. Devazepide (1 mg/kg, i.p) was administered 45 min before GDF15. For the fast-refeeding experiment, food was removed from the mice at lights out on the night before the experiment. After a 16.5 hr fast, mice were administered GDF15 and food returned. Food intake was measured at the same time points (n = 5–7 per group).

C57Bl/6J male mice (aged 10 = 18 weeks; n = 6 for each group) were administered a single, intraperitoneal dose of saline or cisplatin (4 mg/kg). Food intake and body weight were monitored daily. At 24, 48, and 72 hr, n = 6 mice from each group were sacrificed with CO2 inhalation and blood was collected with cardiac puncture. Plasma was taken with Approtinin and DPP4 inhibitors. GDF15 levels are measured using a mouse specific ELISA, according to manufacturer’s directions (R and D Systems, MGD150). A dose-finding experiment for GFRAL mAb (mIgG1 GFRAL 8A2, Lilly Indianapolis, USA) was performed. Briefly, 1 day prior to GDF15 administration at 4 nmol/kg for 3 consecutive days, the mice were subcutaneously dosed once either with control antibody (mIgG1 antibody, Lilly Indianapolis) at 10 mg/kg or ascending doses of antibody at 0.3, 1, 3 and 10 mg/kg of GFRAL mAb. Daily food intake and body weight were measured for three days. In a different experiment, control antibody or GFRAL mAb (both 10 mg/kg) were subcutaneously dosed once 1 day prior to intraperitoneal dose of cisplatin at 4 mg/kg.

Conditioned taste and place aversion tests, pica behaviour

Request a detailed protocol

For conditioned taste aversion, C57Bl/6J mice were housed in cages that permitted ad libitum access to water from two bottles, side-by-side, for at least 1 week before the experiment. On day 1 of the study, animals were water deprived overnight for 16.5 hr. The following morning (day 2), water-deprived animals were provided with two bottles of a novel 15% sucrose solution (dissolved in drinking water) for 30 min. At the end of the 30 min sucrose exposure, animals received an s.c. injection of either saline or GDF15 (4 nmol/kg, 4 ml/kg; n = 6 per group). Two water bottles were returned immediately and mice had unlimited access to water for one night. On day 3, mice were again water deprived overnight. On day 4, water-deprived animals were provided with one bottle of 15% sucrose and one bottle of water for a period of 24 hr. Volumes of sucrose and water intake were measured at 2 hr and 24 hr and used to calculate sucrose preference (sucrose intake/total fluid intake * 100). Food was available ad libitum throughout the study. The positioning of the sucrose and water bottles (left or right) was randomised within treatment groups.

Conditioned place aversion was performed using an apparatus composed of two chambers with distinct visual and tactile qualities, connected by a brightly lit corridor (Harvard Biosciences/Biochrom Ltd., Cambridge, UK). The darker chamber consisted of a rough black floor and black spotted walls, whereas the lighter chamber consisted of a smooth grey floor and grey striped walls. Time spent in each chamber was monitored by video cameras mounted directly above the apparatus, connected to a computer running tracking software (Smart v3.0, Panlab, Harvard Biosciences/Biochrom Ltd.). All procedures were carried out between three and five hours after lights on. On day 1, C57Bl/6J mice (n = 12) were given free access to the full apparatus and allowed to freely explore both chambers for 30 min. Their initial pre-test preference was calculated from the time spent in each chamber. A biased design was used, whereby GDF15 was associated with the most-preferred chamber, which was the darker chamber for all mice. On days 2 and 3, a conditioning session was performed, where mice were restricted to one chamber following administration of either GDF15 (4 nmol/kg, injected s.c. on day 2) or saline (0.9% NaCl injected s.c. on day 3). Following each injection, mice were returned to their home cage for 10 min and then placed in the relevant chamber for 30 min. On day 4, a test session was performed in identical fashion to day 1, and their post-conditioning preference was calculated from the time spent in each chamber. Food and water was available ad libitum throughout the study, except for when mice were in the conditioning apparatus.

To measure pica behaviour, male Sprague-Dawley rats (Envigo, Indianapolis, IN) were acclimated with kaolin pellets available ad libitum in a hopper placed continuously in the home cage. Rats were assigned randomly to three groups (n = 9–10 per group) and treated on 3 consecutive days with vehicle (acetate buffer, pH 5.5, 1 ml/kg, s.c.), human recombinant GDF15 (0.2 mg/kg, s.c.) or LiCl (0.3 M in water, 1% body weight, equivalent to 128 mg/kg, i.p.). Body weight, chow and kaolin intake were determined daily.

Statistics

Statistical analyses were performed using Prism 7 (Graphpad Software, La Jolla, CA). Data were analysed using t-test, one-way ANOVA, two-way or repeated measures ANOVA with post hoc comparisons. When appropriate, non-parametric equivalents were used. N represents independent biological replicates. No statistical methods were used to predetermine sample sizes. Sample size was computed based on pilot data and published literature. Data are presented as mean ± SEM and statistical significance was set at p<0.05.

Data availability

All data generated or analysed during this study are included in the manuscript and supporting files.

References

    1. Brown DA
    2. Ward RL
    3. Buckhaults P
    4. Liu T
    5. Romans KE
    6. Hawkins NJ
    7. Bauskin AR
    8. Kinzler KW
    9. Vogelstein B
    10. Breit SN
    (2003)
    MIC-1 serum level and genotype: associations with progress and prognosis of colorectal carcinoma
    Clinical Cancer Research 9:2642–2650.
    1. Buckhaults P
    2. Zhang Z
    3. Chen YC
    4. Wang TL
    5. St Croix B
    6. Saha S
    7. Bardelli A
    8. Morin PJ
    9. Polyak K
    10. Hruban RH
    11. Velculescu VE
    12. Shih I
    (2003)
    Identifying tumor origin using a gene expression-based classification map
    Cancer Research 63:4144–4149.
  42. Book
    1. Paxinos G
    2. Franklin KBJ
    (2004)
    The Mouse Brain in Stereotaxic Coordinates
    Elsevier Academic Press.

Decision letter

  1. Richard D Palmiter
    Reviewing Editor; Howard Hughes Medical Institute, University of Washington, United States
  2. Catherine Dulac
    Senior Editor; Harvard University, United States
  3. Richard D Palmiter
    Reviewer; Howard Hughes Medical Institute, University of Washington, United States
  4. Marcelo O Dietrich
    Reviewer; Yale University School of Medicine, United States

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

Acceptance summary:

Anorexia is a common response to illness or during cancer treatment and it can have serious consequences. The authors of this paper reveal that the cytokine GDF-15 is produced in response to the cancer drug, cisplatin and that it induces anorexia by activating cells in the hindbrain that express cholecystokinin and the GDF-15 receptor. Blockade of GDF-15 receptor or CCK signaling ameliorates the anorexic effect of cisplatin.

Decision letter after peer review:

Thank you for submitting your article "The cytokine GDF15 signals through a population of brainstem cholecystokinin neurons to mediate anorectic signalling" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Richard D Palmiter as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Catherine Dulac as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Marcelo O Dietrich (Reviewer #3).

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

The role of GDF15 and its receptor GFRAL in signaling peripheral threats is an exciting advance in the field of interoceptive signaling. Understanding how the GFRAL-expressing neurons in the AP/NTS signal to other brain regions is important. Characterizing the signaling molecules made by GFRAL neurons and their relevant post-synaptic targets is critical to this endeavor.

The reviewers are pleased with the overall thrust of the experiments described in this paper. The manuscript includes some important observations, but many details are missing (see below). Consequently, we are not sure whether your team can address these issues in a timely manner (2 months). If not, you may prefer to withdraw the paper now and submit it as a new paper when you have more definitive data.

The major concerns that need to be addressed are as follows:

1) The molecular identity of only about half of the GFRAL neurons was determined. What is the molecular identity of the other half? What fast transmitters do GFRAL neurons use, if any? Single-cell RNA-Seq is one way to answer these questions but is probably not feasible in 2 months unless the authors have already initiated this approach.

2) Addressing the identity of GFRAL post-synaptic neurons that mediate anorexia or other phenotypes is one of the goals of this study. GDF15 activates Fos in parabrachial CGRP neurons, in agreement with previous results, suggesting that GFRAL (CCK subtype) neurons may directly innervate CGRP neurons; however, experiments should be included that reveal whether the signaling is direct or indirect. Is the projection to the PBN the only one that mediates anorexia in response to GFRAL neuron activation?

3) The observation that CCK neurons and CCK itself are an important contributor to the anorexic phenotype is an interesting observation. If the GFRAL/CCK neurons are glutamatergic, it is surprising that CCK is so important. Experiments showing the CCK action in the PBN is important and determining which CCK receptors are involved are needed to solidify this interesting observation.

4) Does CCK ablation in AP/NTS have long-term effects on food intake or body weight. Perhaps this experiment has already been published. If so, provide a reference.

5) The conditioned place aversion experiment needs clarification. The authors say that testing was the same as day one, i.e. 30-min (1800-s) session. But Figure 2D shows only ~500 s on the preferred side which is less than half of the total time. Pairing GDF15 with the preferred side decreased that time to about 250 seconds suggesting that it is aversive. One problem is that "CPA score" is never defined.

6) The number of mice used for each experiment should be included in the figure legends.

The reviewers also have specific suggestions for improvement of this paper which can be found in the individual comments.

Reviewer #1:

The authors of this paper provide data demonstrating that GDF15 induces anorexia and nausea-like behavior in mice/rats and suggest that GDF15 is not a physiological regulator of satiety but rather a sickness signal. They also show that cisplatin stimulates GDF15 expression and the anorexia effect of this cancer-treatment drug depends on GFRAL. Many of the experiments are similar to those recently published by some of the same authors in Cell Metabolism (2020, 31:1-12).

1) The strength of this paper is that the authors begin to characterize the signaling molecules expressed by GFRAL neurons and the neural circuits that they activate. Using a candidate-gene approach they provide evidence that about half of the GFRAL neurons express CCK and/or TH, but the remaining neurons remain unidentified. What fraction of the GFRAL neurons are glutamatergic? Using RNA-Seq of GFRAL-expressing neurons could identify the remaining GFRAL neurons and help elucidate whether GFRAL represents one or multiple populations of neurons.

2) The authors show that treatment with GDF15 induces Fos expression in CGRP neurons in the parabrachial nucleus (PBN) and this induction depends on CCK-expressing neurons (ablation study) and CCK (antagonist study). The implication of these studies is that CCK/GFRAL neurons project directly to CGRP-expressing neurons in the PBN and that CCK plays a significant role in the anorexia phenotype. It is a little surprising that CCK has such a large effect considering that the GFRAL/CCK neurons are probably also glutamatergic and may express other neuropeptides. Authors should determine whether CGRP neurons in the PBN express CCK receptors and whether they are type 1 and/or type 2 receptors. Demonstrating that CCK signaling in PBN is sufficient for the anorexia phenotype is also important.

Reviewer #2:

The manuscript by Beebe and colleagues investigates various aspects of GDF15 signaling and its receptor GFRAL within the brainstem that extends many previous findings about this cytokine system. This current study reports five main findings: (1) GFRAL-positive neurons in the AP and NTS partially overlap with CCK. (2) GDF produces negative affect, as measured by conditioned taste aversion, conditioned place aversion, and increased pica behavior. (3) GDF15 activates CCK neurons (as measured by Fos expression), as well as several downstream brain areas including the PBN and PVN. (4) Ablating CCK neurons in the brainstem or injecting mice with a global CCK antagonist attenuates the anorexigenic effects of GDF15. (5) GFRAL receptor blockade with a monoclonal antibody attenuates the adverse side effects of a cancer therapeutic drug, cisplatin.

Taken together, these results extend the findings of previous studies that reported the anorexigenic effects of GDF15 and the expression patterns of GFRAL in the brainstem. Many of the experiments in the study seem well done, however, some concerns dampen enthusiasm for this study.

1) The quantitative results of the Fos studies reported in Figure 1 and Supplementary Figure 1 are not sufficient to justify the conclusion that "GFRAL-positive neurons in the AP and NTS co-localize with CCK." Indeed, the only quantification provided indicates that less than half (45%) of GFRAL cells co-localize with CCK. Could the authors provide a Table with exact measurements across many animals to fully report the quantitative details of these experiments? If 45% of GFRAL cells co-localize with CCK, then why does the Venn diagram in Supplementary Figure 1 imply that much more than 45% (looks like 2/3) co-localize? Is this Venn diagram based on data, or only a rough approximation? What is the percentage of CCK neurons that colocalize with GFRAL? The number of details missing in this analysis calls into question any major conclusion. Certainly, the claim that "GFRAL is localized to CCK-positive neurons in the AP/NTS," as highlighted in the text, is not justified by the data.

2) Why is the Pica behavior performed in rats whereas all other experiments are performed in mice? Switching model organisms in the middle of a study seems odd. Wouldn't this experiment be relatively easy to perform in mice?

3) The title of one of the Results sections… "The effects of GDF15 are mediated by CCK neurons" is not actually demonstrated in this section. The section that follows reports the effects of GDF15 administration on Fos staining. One cannot conclude from these studies that the effects of GDF15 are mediated by CCK neurons. Ironically, this could be the conclusion of the following section, which demonstrates the necessity of CCK neural activity for GFF15-mediated anorexia.

4) Why do the authors study a loss of function of CCK neurons using two methods in Figure 4 (the Tobacco Etch Protease and CCK receptor antagonist), and then, when studying the effects of cisplatin, use a completely different method in Figure 5 (monoclonal antibody against GFRAL) to study GFRAL neurons instead of CCK neurons? This study is sometimes about CCK-expressing neurons and sometimes about GFRAL-expressing neurons, and some consistency across experiments would lend to better overall conclusions.

5) Does the ablation of CCK-expressing cells in the AP/NTS cause long-term changes in food intake behavior? The results in Figure 4 imply that ablation of these cells does not alter food intake, however, they prevent the anorexia caused by GDF15 administration. Could the authors make a larger conclusion about the fact that there is no result of CCK neuron ablation in the absence of GDF15 administration?

6) The histology results shown in Figure 5D are too small to discern.

Reviewer #3:

The paper provides novel insight into the population of brainstem neurons that express the receptor GFRAL for the cytokine GDF15 that is produced in pathological states and leads to anorexia and weight loss. In addition, they show data that suggest the effect of GDF15 to reduce food intake is in part mediated by CCK expressing neurons in the NTS and AP and by CCK signaling. Moreover, they show that GDF15 evoke conditioned taste and food aversion. These are important observations in a field of research that is rapidly expanding. Some concerns remain that would need to be addressed.

1) The animal models that have Cre-recombinase can have leaky Cre-expression. At the minimum, the authors should comment on the specificity of the Cre lines used for labeling the neuropeptidergic neurons studied.

2) The number of animals used in each experiment seems to be missing from all figures/experiments.

3) Given the importance of the anatomical characterizations of GFRAL neurons in the brainstem, it would be useful to include quantification of cells in Figure 1.

4) "Together these data suggest that GDF15 is not a natural satiety factor…". At the moment this statement is not very convincing, and the experiments were not designed to generate this conclusion.

5) Conditioned place aversion (Figure 2D). I might have missed something here, which could be clarified in the text, however it seems that in both pre-test and test conditions animals still prefer the side of the chamber in which they received GDF15, instead of switching the preference to the saline side. Can the decrease in preference be considered as a place aversion, even though the animals still prefer the same side of the chamber?

6) Figure 5G: the reduction in food intake by cisplatin in the presence of control Ab seems to be small and only significant at day 2 (?) Clarify statistics and N. In Supplementary Figure 5A and B would be helpful to see the correlation between GDF15 and cumulative food intake and %BW in control mice injected saline.

https://doi.org/10.7554/eLife.55164.sa1

Author response

The major concerns that need to be addressed are as follows:

1) The molecular identity of only about half of the GFRAL neurons was determined. What is the molecular identity of the other half? What fast transmitters do GFRAL neurons use, if any? Single-cell RNA-Seq is one way to answer these questions but is probably not feasible in 2 months unless the authors have already initiated this approach.

We were able to carry out two experiments before complete lockdown. We carried out some RNAScope in situ hybridisation, to further analyse the relationship between GFRAL and CCK. This showed that 69 ± 2 % of GFRAL neurons in the area postrema contain Cck mRNA, though the proportion of double-labelled cells is lower in the NTS (35 ± 4 %). As 46 ± 2 % of CCK neurons in the AP contain GFRAL, this represents a significant population. In addition, using a Slc17a6Cre::eYFP mouse we were able to show that GFRAL immunoreactive cells co-express VGlut. This fits with the consensus that CCK neurons in the brainstem are glutamatergic. Thus, we feel that this is ample evidence to confirm that GFRAL cells are predominantly CCK-ergic and have reworded the manuscript accordingly.

2) Addressing the identity of GFRAL post-synaptic neurons that mediate anorexia or other phenotypes is one of the goals of this study. GDF15 activates Fos in parabrachial CGRP neurons, in agreement with previous results, suggesting that GFRAL (CCK subtype) neurons may directly innervate CGRP neurons; however, experiments should be included that reveal whether the signaling is direct or indirect. Is the projection to the PBN the only one that mediates anorexia in response to GFRAL neuron activation?

We do not have a Gfral-cre mouse available. However, Roman et al., 2016, have demonstrated, using channel rhodopsin-assisted mapping, that CCKNTS neurons project directly to CGRPPBN cells. We will cite this paper explicitly on this point. Bearing in mind the limitations of tracing techniques, we have shown that the majority of CCK/GFRALNTS and CCK/GFRALAP neurons, which are activated by GDF15, project to the PBN and not to other identified, presumably transynaptically activated areas. Furthermore, when we ablate these PBN-projecting CCK neurons, the anorectic response to GDF15 is lost. Importantly, 59% of CCK neurons activated by GDF15 are in the AP, whereas only 11% are in the NTS. Unfortunately, it will be difficult to separate accurately the NTS and AP GFRAL populations with AAV injections, since they span the nuclei. In the manuscript, we have also tried to highlight the population of unidentified, GDF15-activated neurons in the NTS. These too may be dependent on CCK input.

It may be possible to establish the role of CGRPPBN neurons in the response to GDF15, using Calca-cre mice, that can be injected with a cre-dependent AAV-caspase. However, due to Covid lockdown, we cannot initiate this experiment. We have tried to be clear that, while some of the neurons activated in the PBN contain Calca-cre, not all of them do. Therefore, we are not convinced that we would necessarily block the whole effect of GDF15 by inactivating CGRP neurons (see point below also). We have re-written the paragraph to make it clear that some of the activated neurons in the PBN contain CGRP, and so these are one downstream target.

3) The observation that CCK neurons and CCK itself are an important contributor to the anorexic phenotype is an interesting observation. If the GFRAL/CCK neurons are glutamatergic, it is surprising that CCK is so important. Experiments showing the CCK action in the PBN is important and determining which CCK receptors are involved are needed to solidify this interesting observation.

This relates to the point above. We are reliant on the Calca-cre mouse injected with AAV-mCherry for identification of CGRPPBN neurons. Although we, and some others, have attempted immunostaining for native CGRP and CCK1R, the results are not reliable and we would not trust interpretation based on this methodology. Thus, we also have performed dual label in situ hybridisation histology for Calca and Cckr1, which clearly shows a small proportion of CGRP neurons contain Cckr1 mRNA.

Roman et al., 2016, showed that the immediate electrical response to stimulation of CCKNTS neurons can be blocked with glutamate antagonists. Although establishing that input, it does not rule out an effect of CCK on the functioning of CGRPPBN or other PBN cells. Our results demonstrate that about half of the acute effects of GDF15 are dependent on CCK receptor signalling. We have rewritten the section to incorporate our new findings and to put the size of the effect of the antagonist into context. We agree that the effect of the CCK antagonist is significant, and we cannot discount an important role for synaptically activated cells in the NTS (a region which also contains CCKR1). We are keen to investigate this signalling pathway further, but cannot because of Covid restrictions.

4) Does CCK ablation in AP/NTS have long-term effects on food intake or body weight. Perhaps this experiment has already been published. If so, provide a reference.

To our knowledge, the long-term effects of ablating CCK neurons has not been published. Across the time frame we have studied in our mice, we have not seen any obvious changes in body weight. This has now been included in Figure 4—figure supplement 1.

5) The conditioned place aversion experiment needs clarification. The authors say that testing was the same as day one, i.e. 30-min (1800-s) session. But Figure 2D shows only ~500 s on the preferred side which is less than half of the total time. Pairing GDF15 with the preferred side decreased that time to about 250 seconds suggesting that it is aversive. One problem is that "CPA score" is never defined.

Thank you for this observation. We have not made this point clear, but there is an adjoining “corridor” between the two chambers, permitting “unforced choice.” This is considered good practice in any form of place conditioning, as described by Prus et al., (2009; PMID: 21204336). As the mice are free to explore three areas, the figures do not appear to add up. Thus, we have re-graphed the results to show the absolute time spent in the GDF15-paired chamber, which we hope is more clear. We will clarify this in the text and removed reference to the CPA score. Please see a fuller explanation in response to the reviewer.

6) The number of mice used for each experiment should be included in the figure legends.

Numbers of mice will be included in each figure legend.

The reviewers also have specific suggestions for improvement of this paper which can be found in the individual comments.

Reviewer #1:

The authors of this paper provide data demonstrating that GDF15 induces anorexia and nausea-like behavior in mice/rats and suggest that GDF15 is not a physiological regulator of satiety but rather a sickness signal. They also show that cisplatin stimulates GDF15 expression and the anorexia effect of this cancer-treatment drug depends on GFRAL. Many of the experiments are similar to those recently published by some of the same authors in Cell Metabolism (2020, 31:1-12).

This is not our paper, and it does not show the same data. However, it has been cited.

1) The strength of this paper is that the authors begin to characterize the signaling molecules expressed by GFRAL neurons and the neural circuits that they activate. Using a candidate-gene approach they provide evidence that about half of the GFRAL neurons express CCK and/or TH, but the remaining neurons remain unidentified. What fraction of the GFRAL neurons are glutamatergic? Using RNA-Seq of GFRAL-expressing neurons could identify the remaining GFRAL neurons and help elucidate whether GFRAL represents one or multiple populations of neurons.

As described above, we have now provided additional data to show GFRAL neurons contain VGlut. Importantly, RNAScope shows that the majority of Gfral cells contain Cck, and so they are the predominant phenotype. We do not have the opportunity, resources or time to carry out RNASeq analysis.

2) The authors show that treatment with GDF15 induces Fos expression in CGRP neurons in the parabrachial nucleus (PBN) and this induction depends on CCK-expressing neurons (ablation study) and CCK (antagonist study). The implication of these studies is that CCK/GFRAL neurons project directly to CGRP-expressing neurons in the PBN and that CCK plays a significant role in the anorexia phenotype. It is a little surprising that CCK has such a large effect considering that the GFRAL/CCK neurons are probably also glutamatergic and may express other neuropeptides. Authors should determine whether CGRP neurons in the PBN express CCK receptors and whether they are type 1 and/or type 2 receptors. Demonstrating that CCK signaling in PBN is sufficient for the anorexia phenotype is also important.

We have now shown that a proportion of CGRPPBN neurons contain Cckr1 mRNA and have stated clearly that there may be other GFRAL targets in the PBN. We have pointed out that about half of the acute anorectic effect of GDF15 is blocked by the CCK receptor antagonist. Importantly, we reiterate that GFRAL neurons appear to activate unidentified neurons in the both the PBN and in the NTS.

Reviewer #2:

[…] 1) The quantitative results of the Fos studies reported in Figure 1 and Supplementary Figure 1 are not sufficient to justify the conclusion that "GFRAL-positive neurons in the AP and NTS co-localize with CCK." Indeed, the only quantification provided indicates that less than half (45%) of GFRAL cells co-localize with CCK. Could the authors provide a Table with exact measurements across many animals to fully report the quantitative details of these experiments?

We now have provided the exact measurements in a supplementary table (Supplementary file 1). This includes new data from the RNAScope study, which indicates a higher co-localisation.

If 45% of GFRAL cells co-localize with CCK, then why does the Venn diagram in Supplementary Figure 1 imply that much more than 45% (looks like 2/3) co-localize? Is this Venn diagram based on data, or only a rough approximation? What is the percentage of CCK neurons that colocalize with GFRAL? The number of details missing in this analysis calls into question any major conclusion. Certainly, the claim that "GFRAL is localized to CCK-positive neurons in the AP/NTS," as highlighted in the text, is not justified by the data.

We have now dealt with this by carrying out dual-label RNAScope (see comments to editors). This clearly shows that a majority of Gfral-containing neurons also contain Cck mRNA. Likewise, 46 % of Cck neurons in the AP contain Gfral. These data, along with additional quantification of co-localisation of VGlut and TH in GFRAL neurons has also been included in the supplementary table (Supplementary file 1). Although we had stated that the Venn diagram was only a rough representation, we agree that it does not portray our results accurately, so we have now omitted it from the manuscript.

2) Why is the Pica behavior performed in rats whereas all other experiments are performed in mice? Switching model organisms in the middle of a study seems odd. Wouldn't this experiment be relatively easy to perform in mice?

Pica behaviour is notoriously difficult to perform in mice as they have a tendency to gnaw at kaolin and spread the powder liberally around their cages, making accurate weighing almost impossible. We think that adding data from another species adds value, but we decided only to include the figure as supplementary material.

3) The title of one of the Results sections "The effects of GDF15 are mediated by CCK neurons" is not actually demonstrated in this section. The section that follows reports the effects of GDF15 administration on Fos staining. One cannot conclude from these studies that the effects of GDF15 are mediated by CCK neurons. Ironically, this could be the conclusion of the following section, which demonstrates the necessity of CCK neural activity for GFF15-mediated anorexia.

Strictly true, but as we have now shown GFRAL is expressed mostly in CCK neurons, it is a fair assumption. That said, we have altered the sub-section title to “GDF15 activates CCKAP/NTS neurons.”

4) Why do the authors study a loss of function of CCK neurons using two methods in Figure 4 (the Tobacco Etch Protease and CCK receptor antagonist), and then, when studying the effects of cisplatin, use a completely different method in Figure 5 (monoclonal antibody against GFRAL) to study GFRAL neurons instead of CCK neurons? This study is sometimes about CCK-expressing neurons and sometimes about GFRAL-expressing neurons, and some consistency across experiments would lend to better overall conclusions.

With respect, the reviewer will not have seen our additional results, which shows the high overlap between GFRAL and CCK neurons. Thus, we are then trying to study the relevance of that cell mediator in different ways. We believe that providing evidence using different methods can add strength to the argument. Caspase will kill CckCre neurons, importantly including GFRAL-expressing CCK neurons, as we have verified in post mortem cell counts (Figure 4—figure supplement 1). CCK neurons in the brainstem are glutamatergic and may contain other transmitters. Therefore, we wanted to assess the role of CCK itself (as opposed to other transmitters) in mediating the effects of GDF15, and so used a receptor antagonist. The last section of the paper addresses a separate question of whether the GFRAL receptor can be targeted in translational research. The GFRAL receptor antibody is an excellent tool to provide useful data, as an antagonist is not available.

5) Does the ablation of CCK-expressing cells in the AP/NTS cause long-term changes in food intake behavior? The results in Figure 4 imply that ablation of these cells does not alter food intake, however, they prevent the anorexia caused by GDF15 administration. Could the authors make a larger conclusion about the fact that there is no result of CCK neuron ablation in the absence of GDF15 administration?

To our knowledge, the long-term effects of ablating CCK neurons has not been published. Across the time frame we have studied in our mice, we have not seen any obvious changes in body weight. This has now been included as Figure 4—figure supplement 1. We are not in a position to speculate whether CCK neurons have a homeostatic role, as our experiments have not been designed specifically to answer such a question.

6) The histology results shown in Figure 5D are too small to discern.

We have remade this figure.

Reviewer #3:

[…] 1) The animal models that have Cre-recombinase can have leaky Cre-expression. At the minimum, the authors should comment on the specificity of the Cre lines used for labeling the neuropeptidergic neurons studied.

References are provided in the Materials and methods section for all the mouse lines used, which demonstrate that each produces a true representation of endogenous gene expression. It is known that both the CckCre and CalcaCre lines may report on more cells than express the gene in adult mice; suggesting that more cells may express significant levels of these genes during development. This is a fair point regarding the initial section, where we used a candidate-gene approach and transgenic mouse lines. However, we have now followed this up with dual-label RNAScope in adult mice, which shows a good consensus. For the experiment using CalcaCre, there is not an issue, since adult mice were injected with a viral reporter.

2) The number of animals used in each experiment seems to be missing from all figures/experiments.

Numbers have been included

3) Given the importance of the anatomical characterizations of GFRAL neurons in the brainstem, it would be useful to include quantification of cells in Figure 1.

Quantification has been added in the new supplementary table (Supplementary file 1).

4) "Together these data suggest that GDF15 is not a natural satiety factor…". At the moment this statement is not very convincing, and the experiments were not designed to generate this conclusion.

We have added reference to the recent publications and tempered the sentence by saying “Together, these and recently published data (Borner et al., 2020 two papers) suggest that GDF15 is probably not a satiety factor.” The following sentence in our manuscript also provides another four references which contain additional supporting evidence for this statement. In this manuscript we go on to show that GFRAL neurons are only activated by stimuli which cause gastric malaise, and not by stimuli that induce normal satiety. We also show that treatment of mice with the GFRAL monoclonal antibody itself does not affect feeding or body weight, but does block anorexia caused by a drug that causes gastric malaise.

5) Conditioned place aversion (Figure 2D). I might have missed something here, which could be clarified in the text, however it seems that in both pre-test and test conditions animals still prefer the side of the chamber in which they received GDF15, instead of switching the preference to the saline side. Can the decrease in preference be considered as a place aversion, even though the animals still prefer the same side of the chamber?

Thank you for this observation, as we can see how confusing this is. It highlights the issue of using a CPA “score” to denote aversion. As done by others, CPA was calculated by [time in chamber A] minus [time in chamber B] and, therefore, represents the difference in time spent between chambers, not the absolute time spent in each chamber. This calculation is often used in CPP/CPA experiments to account for the inherent preference of mice to one-side of a two-chamber apparatus and normalises this preference when there is a difference in separate treatment groups (which is not an issue here). Our study is further complicated by the fact that we have an adjoining corridor between two chambers, permitting unforced choice (as described by Prus and colleagues, PMID 21204336), but giving the mice three areas to explore.

We accept that the CPA score is misleading and perhaps not appropriate here. Therefore, we have changed Figure 2C to depict time spent in the GDF15-paired chamber and have described explicitly both our apparatus and analysis in the Materials and methods section.

6) Figure 5G: the reduction in food intake by cisplatin in the presence of control Ab seems to be small and only significant at day 2 (?) Clarify statistics and N. In Supplementary Figure 5A and B would be helpful to see the correlation between GDF15 and cumulative food intake and %BW in control mice injected saline.

Thank you for this comment. We agree that the result for day 1 looks small; this is because the first injection of cisplatin is given on day 2. We have altered the axis in the Figure 5G to reflect this (i.e. daily not cumulative food intake). Stats and N are now clear in the legend. We have included the correlation analysis for mice injected with control in Figure 5—figure supplement 1.

https://doi.org/10.7554/eLife.55164.sa2

Article and author information

Author details

  1. Amy A Worth

    Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom
    Contribution
    Conceptualization, Data curation, Formal analysis, 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:" 0000-0003-2573-7140

  2. Rosemary Shoop

    Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom
    Contribution
    Data curation, Formal analysis, Investigation, Visualization
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-3617-4358

  3. Katie Tye

    Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom
    Contribution
    Formal analysis, Investigation
    Competing interests
    No competing interests declared

  4. Claire H Feetham

    Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom
    Contribution
    Investigation
    Competing interests
    No competing interests declared

  5. Giuseppe D'Agostino

    1. Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom
    2. Rowett Institute, University of Aberdeen, Aberdeen, United Kingdom
    Contribution
    Data curation, Formal analysis, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared

  6. Garron T Dodd

    School of Biomedical Sciences, The University of Melbourne, Victoria, Australia
    Contribution
    Resources, Formal analysis, Investigation
    Competing interests
    No competing interests declared

  7. Frank Reimann

    Institute of Metabolic Science, Addenbrooke’s Hospital, Cambridge, United Kingdom
    Contribution
    Resources
    Competing interests
    No competing interests declared

  8. Fiona M Gribble

    Institute of Metabolic Science, Addenbrooke’s Hospital, Cambridge, United Kingdom
    Contribution
    Resources
    Competing interests
    No competing interests declared

  9. Emily C Beebe

    Lilly Research Laboratories, Eli Lilly & Company, Indianapolis, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    Paid employee of Eli Lilly.

  10. James D Dunbar

    Lilly Research Laboratories, Eli Lilly & Company, Indianapolis, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    Paid employee of Eli Lilly.

  11. Jesline T Alexander-Chacko

    Lilly Research Laboratories, Eli Lilly & Company, Indianapolis, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    Paid employee of Eli Lilly.

  12. Dana K Sindelar

    Lilly Research Laboratories, Eli Lilly & Company, Indianapolis, United States
    Contribution
    Formal analysis, Investigation
    Competing interests
    Paid employee of Eli Lilly.

  13. Tamer Coskun

    Lilly Research Laboratories, Eli Lilly & Company, Indianapolis, United States
    Contribution
    Resources, Formal analysis, Investigation, Methodology, Writing - review and editing
    Competing interests
    Paid employee of Eli Lilly.

  14. Paul J Emmerson

    Lilly Research Laboratories, Eli Lilly & Company, Indianapolis, United States
    Contribution
    Resources, Formal analysis, Investigation, Methodology, Writing - review and editing
    Competing interests
    Paid employee of Eli Lilly.

  15. Simon M Luckman

    Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    simon.luckman@manchester.ac.uk
    Competing interests
    BB/S008098/1 is a BBSRC Industrial Partnership Award between SML and Eli Lilly
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5318-5473

Funding

Biotechnology and Biological Sciences Research Council (BB/M001067/1)

  • Simon M Luckman

Biotechnology and Biological Sciences Research Council (BB/L021129/1)

  • Simon M Luckman

Medical Research Council (MR/R002991/1)

  • Simon M Luckman

Medical Research Council (MR/P009824/2)

  • Giuseppe D'Agostino

Biotechnology and Biological Sciences Research Council (BB/S008098/1)

  • Simon M Luckman

University of Manchester (PhD Scholarship)

  • Rosemary Shoop

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

Acknowledgements

This work was funded through BBSRC and MRC grants to SML (BB/M001067/1; BB/L021129/1; BB/S008098/1; MR/R002991/1). RS was funded for part of this project by the award of a University of Manchester PhD Scholarship. GD’A is funded by an MRC Career Development Award (MR/P009824/2).

Ethics

Animal experimentation: All procedures were conducted in accordance with either: the United Kingdom Animals (Scientific Procedures) Act, 1986 (ASPA) and approved by the local animal welfare ethical review body (AWERB); the Eli Lilly Institutional Animal Care and Use Committee (IACUC) in accordance with the National Institutes of Health Guide for Care and Use of Laboratory Animals; or the University of Melbourne Animal Ethics Committee (1914919) and conformed to National Health & 8 Medical Research Council (Australia) guidelines regarding the care and use of experimental animals. Additional guidance from the UK National Centre for 3R's (NC3Rs) was followed where applicable.

Senior Editor

  1. Catherine Dulac, Harvard University, United States

Reviewing Editor

  1. Richard D Palmiter, Howard Hughes Medical Institute, University of Washington, United States

Reviewers

  1. Richard D Palmiter, Howard Hughes Medical Institute, University of Washington, United States
  2. Marcelo O Dietrich, Yale University School of Medicine, United States

Version history

  1. Received: January 14, 2020
  2. Accepted: July 26, 2020
  3. Accepted Manuscript published: July 29, 2020 (version 1)
  4. Version of Record published: August 6, 2020 (version 2)

Copyright

© 2020, Worth et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 4,466
    Page views
  • 592
    Downloads
  • 41
    Citations

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

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)

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

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

  1. Amy A Worth
  2. Rosemary Shoop
  3. Katie Tye
  4. Claire H Feetham
  5. Giuseppe D'Agostino
  6. Garron T Dodd
  7. Frank Reimann
  8. Fiona M Gribble
  9. Emily C Beebe
  10. James D Dunbar
  11. Jesline T Alexander-Chacko
  12. Dana K Sindelar
  13. Tamer Coskun
  14. Paul J Emmerson
  15. Simon M Luckman
(2020)
The cytokine GDF15 signals through a population of brainstem cholecystokinin neurons to mediate anorectic signalling
eLife 9:e55164.
https://doi.org/10.7554/eLife.55164

Categories and tags

Research organisms

Further reading

    1. Genetics and Genomics
    2. Neuroscience

    Translation of dipeptide repeat proteins in C9ORF72 ALS/FTD through unique and redundant AUG initiation codons

    Yoshifumi Sonobe, Soojin Lee ... Paschalis Kratsios
    Research Article Updated

    A hexanucleotide repeat expansion in C9ORF72 is the most common genetic cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). A hallmark of ALS/FTD pathology is the presence of dipeptide repeat (DPR) proteins, produced from both sense GGGGCC (poly-GA, poly-GP, poly-GR) and antisense CCCCGG (poly-PR, poly-PG, poly-PA) transcripts. Translation of sense DPRs, such as poly-GA and poly-GR, depends on non-canonical (non-AUG) initiation codons. Here, we provide evidence for canonical AUG-dependent translation of two antisense DPRs, poly-PR and poly-PG. A single AUG is required for synthesis of poly-PR, one of the most toxic DPRs. Unexpectedly, we found redundancy between three AUG codons necessary for poly-PG translation. Further, the eukaryotic translation initiation factor 2D (EIF2D), which was previously implicated in sense DPR synthesis, is not required for AUG-dependent poly-PR or poly-PG translation, suggesting that distinct translation initiation factors control DPR synthesis from sense and antisense transcripts. Our findings on DPR synthesis from the C9ORF72 locus may be broadly applicable to many other nucleotide repeat expansion disorders.

    1. Cell Biology
    2. Neuroscience

    Removal of extracellular human amyloid beta aggregates by extracellular proteases in C. elegans

    Elisabeth Jongsma, Anita Goyala ... Collin Yvès Ewald
    Research Article Updated

    The amyloid beta (Aβ) plaques found in Alzheimer’s disease (AD) patients’ brains contain collagens and are embedded extracellularly. Several collagens have been proposed to influence Aβ aggregate formation, yet their role in clearance is unknown. To investigate the potential role of collagens in forming and clearance of extracellular aggregates in vivo, we created a transgenic Caenorhabditis elegans strain that expresses and secretes human Aβ1-42. This secreted Aβ forms aggregates in two distinct places within the extracellular matrix. In a screen for extracellular human Aβ aggregation regulators, we identified different collagens to ameliorate or potentiate Aβ aggregation. We show that a disintegrin and metalloprotease a disintegrin and metalloprotease 2 (ADM-2), an ortholog of ADAM9, reduces the load of extracellular Aβ aggregates. ADM-2 is required and sufficient to remove the extracellular Aβ aggregates. Thus, we provide in vivo evidence of collagens essential for aggregate formation and metalloprotease participating in extracellular Aβ aggregate removal.

    1. Computational and Systems Biology
    2. Neuroscience

    Task-dependent optimal representations for cerebellar learning

    Marjorie Xie, Samuel P Muscinelli ... Ashok Litwin-Kumar
    Research Article Updated

    The cerebellar granule cell layer has inspired numerous theoretical models of neural representations that support learned behaviors, beginning with the work of Marr and Albus. In these models, granule cells form a sparse, combinatorial encoding of diverse sensorimotor inputs. Such sparse representations are optimal for learning to discriminate random stimuli. However, recent observations of dense, low-dimensional activity across granule cells have called into question the role of sparse coding in these neurons. Here, we generalize theories of cerebellar learning to determine the optimal granule cell representation for tasks beyond random stimulus discrimination, including continuous input-output transformations as required for smooth motor control. We show that for such tasks, the optimal granule cell representation is substantially denser than predicted by classical theories. Our results provide a general theory of learning in cerebellum-like systems and suggest that optimal cerebellar representations are task-dependent.