Introduction

In mammals, lactation is accompanied by a period of infertility. This adaptive change establishes appropriate birth spacing to enable maternal metabolic resources to be directed towards caring for the new-born offspring, rather than supporting another pregnancy ¹. Lactational infertility is characterized by a lactation-induced suppression of pulsatile luteinizing hormone (LH) secretion, and the temporary loss of the reproductive cycle (in rodents this is exhibited as an extended period of diestrus or anestrus) ^2–5. Lactation is also characterised by chronically elevated levels of the anterior pituitary hormone prolactin, which is essential for milk production and promotes adaptive changes in maternal physiology and behaviour ^{1, 2, 4, 5}. Despite hyperprolactinaemia being a well-recognized cause of infertility, the specific role that prolactin plays in lactational infertility, as distinct from other suckling- or metabolic-related cues, is currently unclear ^{4, 6}.

Recent in vivo studies have confirmed that kisspeptin neurons in the arcuate nucleus of the hypothalamus are responsible for the periodic release of gonadotrophin-releasing hormone (GnRH) and subsequent pulsatile luteinising hormone (LH) secretion that drives reproductive function ^7–11. Studies using GCaMP6 fibre photometry in conscious mice have demonstrated that the arcuate kisspeptin neuronal population exhibits episodes of increased intracellular calcium levels coincident with, and immediately preceding, each pulse of LH secretion in intact and gonadectomised male and female mice ^{7, 8 9}. Miniscope investigation showed that individual kisspeptin neurons within the arcuate population act in a coordinated, synchronised, and episodic manner ^{10, 11}. Loss of pulsatile LH secretion during lactation and consequent lactational infertility may be caused by the loss of kisspeptin-mediated stimulation of GnRH secretion ^12–18. Kisspeptin expression is markedly suppressed in lactation ^{12, 16} and when exogenously stimulated, kisspeptin neurons are unable to activate GnRH neurons during lactation, likely due to a lack of kisspeptin synthesis ¹³.

It is well established that hyperprolactinemia causes infertility, and thus, the elevated prolactin present in lactation seems a likely candidate to be involved in suppressing fertility during lactation. Prolactin administration acutely suppresses LH secretion ¹⁹, and chronic exposure to elevated prolactin reduces Kiss1 mRNA expression in the arcuate nucleus ^{17, 20, 21}. In lactating mice, suppressing endogenous prolactin secretion shortens the period of infertility ²², suggesting that prolactin is important for maintaining the suppression of pulsatile LH secretion during lactation. Such a role for prolactin is controversial ^{4, 6, 23–27}, however, with studies in a number of species suggesting that the neural stimulation of suckling may be more important in maintaining lactational infertility ^{28, 29}. However, it has previously been difficult to disentangle the specific role of prolactin, as suckling, prolactin, and milk production are so tightly linked that manipulating one ultimately impacts the others, making it difficult to determine the contribution of any one element. Here, using a conditional deletion strategy, we have blocked prolactin action in the brain leaving suckling, lactation, and maternal behaviour intact. Using GCaMP fibre photometry techniques, we have also documented arcuate kisspeptin neuron activity across pregnancy and lactation transitions in the same mice and established that prolactin directly acts on these neurons to suppress fertility in lactation.

Materials and Methods

Animals

All experiments were performed using adult female mice on a C57BL/6J background (8-20 weeks of age). Mice were housed under controlled temperature (22 °C ± 2°C) and lighting (12-hour light/12-hour dark schedule, with lights on at 0600 hours) with ad libitum access to food and water (Teklad Global 18% Protein Rodent Diet 2918; Envigo, Huntingdon, United Kingdom). Daily body weight was recorded and vaginal cytology was used to monitor the estrous cycle stage. All experiments were carried out with approval from the University of Otago Animal Welfare and Ethics Committee.

Mice were mated with male wild-type C57BL/6J mice (presence of sperm plug = day 1 pregnancy). The first day a litter was seen was counted as day 1 of lactation and maternal mice were left undisturbed till day 3 of lactation, when vaginal monitoring would resume and litter size was normalised to 6 pups per animal, unless otherwise stated.

To monitor pulsatile secretion of LH, serial tail tip blood sampling and measurement of LH by ELISA was undertaken as reported previously ^{19, 30, 31}. As novel exposure and restraint stress has been shown to suppress pulsatile LH secretion ³², all mice were habituated to the tail tip blood sampling procedure in a gentle restraint device (soft cardboard tube) or hand, for at least 3 weeks prior to experimentation ³³. Sequential whole blood samples (4μl) were collected in 6 minute intervals for 3 hours between 0900 and 1200 hours unless otherwise stated. Samples were immediately diluted in 48ul 0.01M PBS/0.05% Tween 20, and frozen on dry ice before being stored at −20°C for subsequent LH measurement.

Effect of neuron-specific deletion of the prolactin receptor gene on the maintenance of lactational infertility

To investigate whether prolactin action in the brain is required for lactational infertility, neuron-specific Prlr knockout mice (Prlr^lox/lox/Camk2a^Cre) and their respective Cre-negative controls (Prlr^lox/lox) were generated, as previously described ³⁴. We have previously shown that while Prlr^lox/lox/Camk2a^Cre mice do not have a complete Prlr deletion in the forebrain, there are areas of extensive deletion (as measured by reduced prolactin-induced pSTAT5), such as the arcuate nucleus and ventromedial nucleus of the hypothalamus, and areas where Prlr is reduced by about 50% such as the medial pre-optic area ^{34, 35}. RNAscope in-situ hybridization was done to confirm knockdown. Briefly, intact diestrous mice and 14 day OVX mice (all aged 8-16 weeks) were perfused with 2% PFA to enable visualisation of kisspeptin cell bodies in both the RP3V and ARC regions (as kisspeptin cell bodies are only visible in the RP3V of intact mice and in the ARC of OVX mice, due to estradiol regulation³⁶). Brain sections (14μm-thick) were prepared, thaw mounted onto superfrost-plus microscope slides and then stored at −80°C. RNAscope in-situ hybridization was performed using the RNAscope 2.5 High definition Duplex Detection kit – chromogenic (Advanced Cell Diagnostics, Hayward, CA) largely in accordance with manufacturer’s instruction. The channel 1 Prlr probe was custom designed to pick up only the long form of the prolactin receptor. It was designed to transcript NM_011169.5 with a target sequence spanning nucleotides 1107-2147 (Ref: 588621; Advanced Cell Diagnostics, Hayward, CA). The channel 2 Kiss1 probe was custom designed to transcript NM_178260.3 with a target sequence spanning nucleotides 5 to 485 (Ref: 500141-C2; Advanced Cell Diagnostics, Hayward, CA). Sections were thawed at 55°C, postfixed for 3 minutes in 2% PFA, washed in 0.01M PBS for 5 minutes, and endogenous peroxidases were blocked with a hydrogen peroxidase solution for 10 minutes. Tissue was washed in distilled water (dH2O) (3x 2 minutes), then immersed in 100% ethanol briefly, air dried for 5 minutes, and a hydrophobic barrier was applied. Tissue was permeabilized with RNAscope protease plus for 30 minutes at 40°C. Sections were washed (2x 2 minutes) and were hybridized with the Prlr and Kiss1 probes (1:300 dilution, Prlr:Kiss1) or negative control probe (Cat#320751; Advanced Cell Diagnostics, Hayward, CA) at 40°C for 2 hours. Amplification (Amp 1-6) was performed in accordance with the manufacturer’s instructions. Sections were then hybridized with a Fast-RED (1:60, Fast-RED B:Fast-RED A) for 10 minutes at room temperature, before undergoing further amplification steps (Amp 7-10) in accordance to manufacturer’s instructions. The final positive hybridization was detected by incubation with the secondary detection reagents (1:50, Fast-GREEN B:Fast-GREEN) for 10 minutes at room temperature. Sections were washed, counterstained with haematoxylin (25% Gills), dried at 60°C for 20 minutes, and cover-slipped with VectaMount (Vector laboratories, H-5000) before imaging as previously described. Quantification of the proportion of kisspeptin neurons co-expressing Prlr mRNA was undertaken in FIJI software (National Institute of Health, Bethesda, Maryland, USA) following image acquisition. The total number of Kiss1-expressing cells and the total number of these that showed Prlr mRNA expression were counted. Prlr^lox/lox/Camk2a^Cre mice showed a significant decrease in the percentage of Kiss1- expressing cells co-expressing Prlr compared to controls in both the RP3V (p = <0.0001) and arcuate nucleus (p = 0.0009) (unpaired two-tailed t tests, Supplementary Figure 1A-D). The Prlr^lox/lox/Camk2a^Cre mice are hyperprolactinaemic due to impaired negative feedback of prolactin on hypothalamic dopamine neurons ³⁴ and therefore show disrupted estrous cycles (showing recurrent pseudopregnancy-like cycles with long periods of diestrus of approximately 14 days between estrus stages). However, these mice are able to become pregnant and have normal pregnancies. All mice were given a 250μl subcutaneous injection of bromocriptine (5mg/kg, 5% ethanol/saline; Tocris Bioscience Cat#0427) prior to being mated. This treatment was designed to reinstate an estrous cycle in Prlr^lox/lox/Camk2a^Cre mice. Bromocriptine is an agonist for the type 2 dopamine receptor and inhibits prolactin secretion from the pituitary gland ³⁷, thereby terminating the pseudopregnancy-like state and bringing the mice into proestrus the following day. Following treatment, all mice were then housed with a stud male.

For Prlr^lox/lox/Camk2a^Cre mice, estrous cycles were monitored from day 3 of lactation until the first day of diestrus following a day of estrus (proestrus and estrus had to be observed prior to transcardial perfusion on the first day of diestrus). Brains were collected following transcardial perfusion for assessment of kisspeptin immunoreactivity. For every lactating Prlr^lox/lox/Camk2a^Cre mouse (n = 8), the brain of a Prlr^lox/lox control mouse (n = 8) of the equivalent day (±1) of lactation was also collected. A group of non-lactating (NL) mice of both genotypes (n = 5-6) was also perfused for immunohistochemistry on diestrus.

To evaluate pulsatile LH secretion in early lactation (prior to the return of estrous cycles) and to determine whether progesterone played any role in regulating pulsatile LH secretion in lactation, additional groups of lactating Prlr^lox/lox/Camk2a^Cre and Prlr^lox/lox control mice were generated and treated with either the progesterone receptor antagonist, mifepristone (4mg/kg in sesame oil, s.c.; AK Scientific Inc Cat#J10622), or vehicle (n = 7-8 per group) on the morning of day 4 of lactation and on day 5 of lactation, 30 minutes prior to blood sampling that day. This dose was selected as it was found to be sufficient to cause termination of pregnancy in wild-type C57BL/6J mice (p = 0.0072, Chi-squared test, Supplementary Figure 2A; pilot study) and neither vehicle nor mifepristone treatment had an effect on litter weight gain (interaction of time x genotype & treatment p = 0.5322, two-way repeated measures ANOVA, Supplementary Figure 2B).

Measurement of LH concentrations

An established sandwich ELISA method was used to determine LH concentration in diluted whole blood samples collected from mice ^{30, 31}. Briefly, a 96-well high plate was incubated with bovine monoclonal antibody (LHβ518b7, 1:1000 in 1xPBS; Dr. L. Sibley, UC Davis, CA, USA) for 16 h at 4°C. Following incubation of standards, controls and experimental samples for 2 hours, plates were incubated in rabbit polyclonal LH antibody (AFP240580Rb; 1:10,000; National Hormone and Pituitary Program, NIH) for 90 minutes, followed by incubation with polyclonal goat anti-rabbit IgG/HRP antibody (1:1000; DAKO Cytomation) for 90 minutes. Finally, plates were incubated in OPD (o-phenylenediamine capsules; Sigma-Aldrich Cat#P7288) for 30 minutes. A standard curve for the detection of LH concentration was generated using serial dilutions of mouse LH-reference preparation peptide (National Hormone and Pituitary Program, NIH). Luteinizing hormone levels were read using a standard absorbance plate reader (SpectraMax ABS Plus; Molecular Devices) at 490nm and 630nm wavelengths.

PULSAR Otago was used to define LH pulses ³⁸. Parameters used; Smoothing 0.7, Peak split 2.5, Level of detection 0.04, Amplitude distance 3, Assay variability 0, 2.5, 3.3, G(1)=3.5, G(2)=2.6, G(3)=1.9, G(4)=1.5, G(6)=1.2. Mean LH levels were calculated by averaging all LH levels collected during the experiment. The assay had a sensitivity of 0.04ng/ml to 4ng/ml, with an intra-assay coefficient of variation of 4.40% and an inter-assay coefficient of variation of 8.29%. See supplementary information, figure 1, for all individual LH profiles.

Figure 1.

Assessment of kisspeptin expression

Perfusion and fixation of tissue

Mice were anaesthetised with sodium pentobarbital (15mg/mL) and transcardially perfused with 4% paraformaldehyde. Brains were removed, postfixed in the same solution, and cryoprotected overnight in 30% sucrose before being frozen at −80°C. Two sets of 30μm thick coronal brain sections were cut using a sliding microtome, from Bregma 1.10mm to −2.80mm. Brain sections were kept in cryoprotectant solution (pH = 7.6) at −20°C until immunohistochemistry was performed.

Immunohistochemistry

Immunohistochemistry for kisspeptin in the RP3V and arcuate nucleus was performed as previously described ³⁹. Briefly, sections were incubated in polyclonal rabbit anti-kisspeptin primary antibody (AC 566, 1:10,000; gift from A. Caraty, Institut National de la Recherche Agronomique, Paris, France) for 48 hours at 4°C. Sections were then incubated with biotinylated goat anti-rabbit IgG (1:200, Vector biolabs, Peterborough, GK) for 90 min at room temperature, followed by incubation in an avidin-biotin complex (Elite vectastain ABC kit, Vector laboratories). The bound antibody-peroxidase complex was visualised using a nickel-enhanced diaminobenzidine (DAB) reaction, to form a black cytoplasmic precipitate.

Brain sections were imaged using an Olympus BX51 light microscope and Olympus UPlanSApo 10/20x lenses. Quantification of kisspeptin neurons in the RP3V, was undertaken by manually counting all labelled neurons present in all three subdivisions, the anteroventral periventricular nucleus (AVPV), rostral preoptic periventricular nucleus (rPVpo), and caudal preoptic periventricular nucleus (cPVpo, bregma 0.02) (2 sections per brain region per mouse) and then averaging this for each animal. As kisspeptin cell bodies in the arcuate nucleus were not easily observed, as previously reported ⁴⁰, kisspeptin fibre immunoreactivity was imaged using a Gryphax NAOS colour camera (Jenoptik) and evaluated using FIJI software and the voxel counter function (National Institutes of Health). Kisspeptin fibre density was measured in the arcuate nucleus across the three subdivisions; rostral arcuate (rARC), middle arcuate (mARC), and caudal arcuate (cARC) with two sections of each area per animal counted, and then averaged across each animal to get total number and reported as total amount of voxels per ROI (voxel fraction).

Characterization of arcuate kisspeptin neuronal activity using GCaMP fibre photometry

Stereotaxic surgery and AAV injections

Adult Kiss1^Cre or Prlr^lox/lox/Kiss1^Cre mice (2-3 months old) were anaesthetised with 2% Isoflurane, given local Lidocaine (4mg/kg, s.c.) and Carprofen (5mg/kg, s.c.) and placed in a stereotaxic apparatus. A custom-made unilateral Hamilton syringe apparatus holding one Hamilton syringe was used to perform unilateral injections into the arcuate nucleus. The needles were lowered into place (−0.14mm A/P, +0.04mm M/L, −0.56mm DV) over 2 minutes and left in situ for 3 minutes before injection was made. 1μl AAV9-CAG-FLEX-GCaMP6s-WPRE-SV40 (1.3×10^-13 GC/ml, University of Pennsylvania Vector Core, Philadelphia, PA, USA) was injected into the arcuate nucleus at a rate of ∼100nl/min with the needles left in situ for 3 minutes prior to being withdrawn over a period of 6 minutes. This was followed by implantation of a unilateral indwelling optical fibre (400 μm diameter, 6.5 mm long, 0.48 numerical aperture (NA), Doric Lenses, Canada, product code: MFC_400/430-0.48_6.5mm_SM3*_FLT) at the same coordinates. Carprofen (5mg/kg body weight, s.c.) was administered for post-operative pain relief. After surgery, mice received daily handling and habituation to the photometry recording procedure over 4-6 weeks before experimentation began.

GCaMP6 fibre photometry

Photometry was performed as reported previously ⁹. Fluorescence signals were acquired using a custom-built fibre photometry system made primarily from Doric components. Violet (405nm) and blue (490nm) fibre-coupled LEDs were sinusoidally modulated at 531 and 211 Hz, respectively, and focused into a 400μm, 0.48 numerical aperture fibre optic patch cord connected to the mouse. Emitted fluorescence was collected by the same fibre and focused onto a femtowatt photoreceiver (2151, Newport). The two GCaMP6s emission signals were collected at 10 Hz in a scheduled 5s on/15s off mode by demodulating the 405nm (non-calcium dependent) and 490nm (calcium dependent) signals. The power output at the tip of the fibre was set at 50μW. Fluorescent signals were acquired using a custom software acquisition system (Tussock Innovation, Dunedin, New Zealand) and analysed using custom templates created by Dr Joon Kim (University of Otago, Dunedin, New Zealand) based on mathematics and calculations similar to those previously described ^{41, 42}. Briefly, the fluorescent signal obtained after stimulation with 405nm light was used to correct for movement artefacts as follows: first, the 405nm signal was filtered using a savitzky-golay filter and fitted to the 490nm signal using least linear square regression. The fitted 405nm signal was then subtracted and divided from the 490nm signal to obtain the movement and bleaching corrected signal. The output of these templates is 490-adjusted405/adjusted405, which was multiplied to get the final ΔF/F as a percentage increase (all photometry data reported as ΔF/F(%)).

All recordings were obtained from freely behaving mice for up to 24 hours and occurred between the hours of 0800 hours and 1200 hours (apart from 24 hours post weaning recording (0900 hours to 1700 hours), and day 18/19 pregnancy recording (1800 hours to 0800 hours the following day). Synchronized events (SE) were defined as when ΔF/F exceeds 3 standard deviations (SD) above the trace mean. Manual event shape analysis was performed in addition to standard deviation method for certain datasets where necessary. Events were counted manually to determine frequency of events per 60 minutes. The between animal variability in total signal means that changes in SE amplitude can only be reported as relative changes within an animal. Relative SE amplitude was calculated by using normalised ΔF/F data and then subtracting the peak of an SE from the nearest nadir to the rise of the SE and averaging that for the number of SEs in a recording. To obtain normalised ΔF/F, three pre-pregnancy datasets from each mouse were used to find the average maximum ΔF/F for that mouse. All datasets were then divided by that normalisation value to get normalised ΔF/F for each trace for each individual mouse.

Monitoring the activity of arcuate kisspeptin neurons across different reproductive stages in the same mice

Adult Kiss1^Cre mice were 8-10 weeks of age at the beginning of experiments, and up to 12 months in age by time of final recording (n = 8 during pregnancy, n = 6 during lactation; 2 mice were euthanised due to dystocia therefore those mice were only followed through pregnancy). Monitoring of vaginal cytology and weights was continuous from 1 week pre-surgery till day 19 of pregnancy and resumed on day 3 of lactation (with all handling stopped on day 19 to avoid potential compromise of parturition and onset of maternal behaviour). To investigate the activity of the arcuate kisspeptin population across different reproductive states in the same animal, the following recording protocol was followed for all Kiss1^Cre mice, unless otherwise stated; virgin (diestrus), day 4 of pregnancy, day 14 of pregnancy, day 18/19 of pregnancy (overnight), day 7 of lactation, day 14 of lactation, day 18 of lactation, 24 hours after weaning, first diestrus after estrous cycles begin following weaning, and 10 days after OVX. In addition, blood sample collection for paired LH measurement was done in virgin (diestrus) state, on day 14 of pregnancy (in 4 mice, maximum of 6 samples were collected around small “peaks” in baseline), day 7 of lactation, and day 14 of lactation. Blood sampling was not carried out at additional time points as blood sampling was undertaken at least a week apart, and stress from repeated sampling was attempted to be kept at a minimum e.g., not blood sampling around the time of birth. Fibre photometry recordings were usually between 2-4 hours in length. The only longer recordings were undertaken on day 18/19 of pregnancy (14 hours) and 24 hours after weaning (8 hours). These particular recording sessions were extended to determine whether there were any longer-term changes occurring in the activity of the arcuate kisspeptin population in the lead up to parturition or following weaning of pups (states closely followed by postpartum estrus and resumption of normal estrous cycles, respectively).

Effect of arcuate kisspeptin neuron-specific deletion of the prolactin receptor gene on the maintenance of lactational infertility and the activity of kisspeptin neurons during lactation

Kisspeptin-specific prolactin-receptor knockout mice (Prlr^lox/lox/Kiss1^{Cre 19}) and their respective Cre-negative controls (Prlr^lox/lox) were generated. RNAscope in-situ hybridization was done to confirm knockdown, with Prlr^lox/lox/Kiss1^Cre mice showing a significant decrease in the percentage of Kiss1-expressing cells co-expressing Prlr compared to controls in the arcuate (p = <0.0001, unpaired two-tailed t test, Extended data Figure 3E, F). Similar to experiments described above, Prlr^lox/lox/Kiss1^Cre (n = 27) and Prlr^lox/lox control (n = 30) dams underwent estrous cycle monitoring from day 3 of lactation onwards to determine whether mice showed an early resumption of estrus cycles.

To determine whether the deletion of the prolactin receptor from arcuate kisspeptin neurons led to early reactivation of these neurons during lactation, adult Prlr^lox/lox/Kiss1^Cre mice (n = 5) and an additional Kiss1^Cre control mouse (n = 1) (8 weeks old at the start of the experiment, and up to 14 months at end of final recording timepoint) were set up for fibre photometry, as described above. Monitoring of vaginal cytology and weights was continuous from 1 week pre-surgery till day 18 of pregnancy and resumed on day 3 of lactation. Recordings were undertaken in a similar timeline as described above, however no pregnancy recordings were done, and in early lactation recordings were performed every 2 days from day 3 to day 9 of lactation, before following the same protocol as described. No blood samples were taken over this lactation period in this genotype. As described previously, recordings were kept between 2-4 hours, apart from the 24 hours after weaning recording (8 hours).

Statistical analysis

Data are presented as mean ± SEM and all statistical analysis was performed with PRISM software 10 (GraphPad Software, San Diego, CA, USA) with a p value of < 0.05 considered as statistically significant. Individual symbols in graphs represent individual mice. Differences in kisspeptin cells number or fibre density was assessed using two-way ANOVAs with Tukey’s multiple comparisons tests or t tests, with both analyses using combined averages of each animal (averaged number of cells or fibre density across the three subdivisions of each nucleus to get total number reported). Resumption of estrous cycles was analysed using Log-rank (Mantel-Cox) test chi square test. LH pulse frequency data and mean LH data was analysed using two-way ANOVAs with Tukey’s multiple comparisons tests. SE frequency and amplitude throughout reproductive cycles was analysed using mixed effect analysis (fixed type III) with Tukey’s multiple comparisons tests where appropriate and day 18/19 of pregnancy data was analysed using t tests. Correlation between SE occurrence and LH pulses was assessed using chi-square test. All fibre photometry data used for quantitative analysis and comparison were from the first pregnancy and lactation. A full list of probability values, inferential statistics, and degrees of freedom for all data can be found in Supplementary Table 1.

Results

Prolactin action on forebrain neurons is necessary to maintain lactational infertility

Lactation has previously been associated with a marked decrease in Kiss1 mRNA levels in both rostral periventricular region of the third ventricle (RP3V) and arcuate nucleus populations during lactation ¹³. To determine whether prolactin is involved in the maintenance of lactational anestrus, the Prlr gene was knocked out of Camk2a expressing neurons (most forebrain neurons, as described in ³⁴) of female mice. Control Prlr^lox/lox mice showed a marked reduction in kisspeptin cell body immunoreactivity in the RP3V of lactating compared to virgin mice (p = 0.0100, Post hoc Tukey’s multiple comparisons test, Figure 1A, C). In contrast, lactation-induced suppression of kisspeptin cell bodies in the RP3V and fibre labelling in the arcuate nucleus was absent in Prlr^lox/lox/Camk2a^Cre mice (p = 0.6409, Post hoc Tukey’s multiple comparisons test, Figure 1A, C; interaction between reproductive state and genotype p = 0.0034, two-way ANOVA, Figure 1A, C; p = 0.0020, unpaired two tailed t test, measured as percentage voxels within the region of interest, Figure 1B, D).

Estrous cycles during lactation were significantly altered by deletion of Prlr in the forebrain, with all Prlr^lox/lox/Camk2a^Cre mice showing a return to estrus between day 6 and day 10 of lactation (Figure 1E), while, as normal, estrus did not occur until day 20 in control Prlr^lox/lox mice (p = <0.0001, log-rank (Mantel-Cox) test, Figure 1E). No differences in litter weight gain from day 3 to day 8 of lactation were observed in either group (p = 0.3282, mixed analysis test, Supplementary Figure 3), indicating that the suckling stimulus that mice received was maintained in the absence of Prlr expression in Camk2a expressing neurons. Collectively, these data show that prolactin action in the brain is absolutely required for the lactation-induced suppression of kisspeptin expression and to maintain lactational infertility in mice.

In a separate cohort of mice, pulsatile LH secretion in Prlr^lox/lox/Camk2a^Cre mice was monitored in early lactation, prior to the return of estrous cycles. To rule out a potential role for progesterone in suppressing fertility during lactation ^{9, 43, 44}, the progesterone receptor antagonist mifepristone (RU486), was administered to mice in early lactation. Vehicle-treated Prlr^lox/lox control mice showed the expected near complete absence of pulsatile LH secretion during lactation (Figure 2A, B). In contrast, nearly all Prlr^lox/lox/Camk2a^Cre mice showed a lack of the normal lactation-induced suppression of pulsatile LH secretion demonstrated by a significant increase in frequency of LH pulses compared to controls (effect of genotype p = 0.0024, two-way ANOVA with Tukey’s multiple comparisons test, Figure 2A, B). There was no effect of mifepristone on this pattern of LH secretion, suggesting that progesterone action is not required for the suppression of LH secretion (LH pulse frequency (interaction genotype x treatment p = 0.2807; treatment p = 0.8588; Figure 2B), mean LH levels (interaction genotype x treatment p = 0.8697; treatment p = 0.8586; Figure 2C), two-way ANOVA with Tukey’s multiple comparisons test). These data indicate that prolactin is the primary signal responsible for the suppression of LH during lactation in mice. Individual LH profiles from all animals are shown in Supplementary Figure 4.

Figure 2.

Episodic activity of arcuate kisspeptin neurons is suppressed during pregnancy and most of lactation

To directly assess the role of prolactin in regulating kisspeptin neuron activity during lactation, GCaMP6 fibre photometry was undertaken to monitor real-time activity of arcuate kisspeptin neurons in freely behaving mice. We first undertook a longitudinal assessment of changes in kisspeptin neuronal activity by tracking individual animals throughout pregnancy and lactation and following weaning (Figure 3).

Figure 3.

Initially, GCaMP fibre photometry recordings were collected in the virgin diestrous state, both with and without serial blood sampling to measure LH concentrations. As can be seen in Figure 4A, photometry recordings were characterized by discrete synchronised events (SEs) of elevated intracellular calcium (indicative of synchronous activity of the kisspeptin population), with each SE correlating perfectly to a single pulse of LH in the minutes following (p = <0.0001, chi-squared test). These were observed to be at a similar rate to that described previously in diestrous mice using GCaMP photometry in a different Kiss1 mouse line ⁹, with periodic SEs occurring about once per hour (1.250±0.250; Figure 3 & 4B).

Figure 4.

The activity of the arcuate kisspeptin population in Kiss1^Cre mice dynamically changed depending on the reproductive state of the mouse (p = 0.0012, mixed-effect analysis, Figure 3 & 4B). On day 4 of pregnancy, SE frequency had markedly decreased (0.297±0.136/hr; Figure 3 & 4B), indicating an early reduction in activity of arcuate kisspeptin neurons during pregnancy. By day 14 of pregnancy, no SEs were seen (0±0/hr; Figure 3 & 4B) and this was confirmed by a lack of pulsatile LH secretion (Supplementary Figure 5). In late pregnancy (day 18), neuronal activity was monitored for 14 hours, during which time low amplitude, SEs were unexpectedly observed at similar frequencies to virgin levels (2.043±0.940/hr; Figure 3 & 4B-C). This unusual pattern of activity is illustrated in more detail in Figure 5 where alongside the resurgence of low amplitude SEs, there was a marked increase in baseline activity observed, relative to other stages. This activity was reminiscent of the miniature SEs observed to be caused by activation of subgroups of cells in a brain slice technique ¹¹, but we are unable to resolve such events using the present methods. Since our aim was to continue longitudinal assessment of the arcuate kisspeptin population into lactation and weaning, mice were not disrupted by blood sampling immediately prior to parturition as we were concerned that this additional stressor might interfere with establishment of maternal behaviour. Hence, we are unable to report whether these low amplitude SEs and elevated baseline activity were associated with LH secretion.

Figure 5.

Evaluation of activity of arcuate kisspeptin neurons during lactation showed complete suppression of activity on day 7 of lactation with a corresponding absence of pulsatile LH secretion (0±0/hr; Figure 3 & 4B). This lactation-induced suppression of activity was partially relieved by day 14 of lactation (0.583±0.083/hr; Figure 3 & 4B), with SEs again corresponding to low frequency pulses of LH secretion. Further increases in SE frequency were seen on day 18 of lactation (0.850±0.100/hr; Figure 3 & 4B), including an increase in baseline activity, similar to that seen in late pregnancy, and by 24 hours after weaning (day 22 postpartum) the frequency of SEs returned to close to non-pregnant levels (1.375±0.114/hr; Figure 3 & 4B). Frequency remained unchanged on the day of first diestrus following a return to estrous cycles after weaning (1.533±0.226/hr; Figure 3 & 4B). As a final manipulation, mice were ovariectomized (OVX), and arcuate kisspeptin population activity observed to increase significantly with clusters of high amplitude activity (4.778±0.222/hr; Figure 3 & 4B), consistent with previous reports following OVX in nulliparous mice ⁴⁵. Collectively, these observations show extensive, dynamic variation in activity of the arcuate kisspeptin neuronal population associated with pregnancy and lactation.

Mice with an arcuate kisspeptin neuron-specific deletion have premature reactivation of estrous cycles and neuronal activity in lactation

To determine whether the prolactin-induced suppression of estrous cycles and LH pulsatile secretion was specifically mediated by kisspeptin neurons, mice were generated with an arcuate-specific deletion of the Prlr from kisspeptin neurons ¹⁹. Similar to the data from forebrain neuron-specific deletion of Prlr, there was early resumption of estrous cycles in Prlr^lox/lox/Kiss1^Cre mice during lactation (63% showing estrus by day 10 of lactation and 83% by day 19 lactation) compared to Prlr^lox/lox controls (4% by day 19 lactation) (p = <0.0001, log-rank (Mantel-Cox) test, Figure 6A). No difference in litter weight gain during lactation (day 3-20 of lactation weight gain) was observed in either group (p = 0.6404, two-way ANOVA, Supplementary Figure 1H) indicating that suckling and/or lactation itself was not impaired. In vivo GCaMP6 fibre photometry in Prlr^lox/lox/Kiss1^Cre mice showed early reactivation of the arcuate kisspeptin population between day 3 and 5 of lactation (Figure 6C). This was accompanied by a clear return to estrous within this early lactation window in 4/5 mice. These data demonstrate that prolactin action specifically on arcuate kisspeptin neurons is responsible for maintaining suppression of those neurons, and thereby fertility, during lactation in mice.

Figure 6.

Discussion

We demonstrate here that prolactin action in arcuate kisspeptin neurons is necessary for the maintained suppression of fertility during lactation in mice. Neuron-specific Prlr deletion (Prlr^lox/lox/Camk2a^Cre) resulted in premature return to estrus in early lactation, even in the presence of ongoing suckling stimulus and the full metabolic consequences of milk production. Accompanying the resumption of estrus was an absence of the normal lactation-induced reduction in kisspeptin immunoreactivity ^{12, 16, 20}, and pulsatile LH secretion was also observed on day 5 of lactation prior to the premature estrus when it would normally have been completely absent ^46–48. To evaluate the specific role of kisspeptin neurons in mediating the prolactin-induced suppression of fertility, we have comprehensively mapped the activity of arcuate kisspeptin neurons throughout a full reproductive cycle: pregnancy, lactation, and after weaning in individual animals. The data show an immediate suppression of activity of arcuate kisspeptin neuronal activity during pregnancy, and this is maintained throughout most of lactation, apart from a brief window of reactivation immediately prior to parturition. Deleting Prlr specifically from arcuate kisspeptin neurons prevented the suppression of activity in early lactation, resulting in premature induction of episodic activation of kisspeptin neurons, and early onset of estrus. Combined, these data provide direct evidence that prolactin action on kisspeptin neurons is necessary for lactation-induced infertility in mice.

It is now well established that the arcuate kisspeptin neurons form the GnRH “pulse generator”, and hence drive pulsatile release of GnRH from the hypothalamus and consequent pulses of LH form the pituitary that is required for fertility ^49–51. This is the first study to monitor activity of the GnRH “pulse generator” across different reproductive states in the same animal, and the data largely match previously described patterns of LH secretion ^{47, 52}. The frequency and dynamics of the synchronised events changed dramatically, initially due to the pregnancy-induced changes in ovarian hormones. The abrupt decrease in “pulse generator” activity in early gestation is likely caused by rising levels of progesterone, known to profoundly suppress activity of arcuate kisspeptin neurons and LH secretion ^{9, 43, 44}. Progesterone is elevated throughout pregnancy, gradually increasing until luteolysis and progesterone withdrawal occurs in the lead up to parturition ^53–56. Interestingly, we observed a transient reactivation of the arcuate kisspeptin neurons in the night between days 18 and 19 of pregnancy. This was characterised by frequent, low amplitude episodes of activity, and increased baseline activity that may represent the intermittent synchronized activity of small subsets of arcuate kisspeptin neurons that have not yet transitioned to full synchronization of the whole population ¹¹. It seems likely that this pattern of activity is associated with progesterone withdrawal in late pregnancy and may be important in stimulating follicular growth leading up to a postpartum ovulation ^{57, 58}.

In early lactation, episodic activity of arcuate kisspeptin neurons was absent, with sporadic low-amplitude activity returning around day 14 of lactation. There was another period of increased baseline activity in late lactation, similar to that seen in late pregnancy, potentially representing a signature of reactivation of synchronized activity of the arcuate kisspeptin neurons. Overall patterns of activity rapidly returned to normal diestrous levels soon after weaning. This increase in “pulse generator” activity during late lactation mirrors the increase in LH levels that has been reported as lactation progresses ⁵⁹. In the absence of Prlr in arcuate kisspeptin neurons, however, synchronized episodic activity re-appeared as early as 3 days after birth, even in the presence of ongoing suckling. These data clearly show that prolactin action in the arcuate kisspeptin neurons is necessary to sustain lactational infertility in mice. The observed disruption of lactational infertility in Prlr^lox/lox/Kiss1^Cre mice is particularly remarkable given that Prlr deletion is restricted to arcuate kisspeptin neurons in this model ¹⁹, and prolactin action on RP3V kisspeptin neurons ^{21, 60} and on gonadotrophs in the pituitary gland ^61–63 are unaffected.

The indispensable role for prolactin in mediating lactation-induced infertility in the mouse is surprising, given the consensus of much work in other species concluding that other factors may be more important (see ^{4, 6, 26, 27, 64}). This may reflect a level of redundancy amongst contributing factors across all species, including ovarian hormones, metabolic cues and neural inputs of suckling. Notably, the conditional deletion approach described here distinguishes prolactin action from the neurogenic effects of suckling without altering the process of lactation itself. Moreover, this approach avoids the potential confounding effects of using dopamine agonists to suppress prolactin ^{28, 29}, given that dopamine can directly inhibit GnRH neuronal activity ⁶⁵.

While the effects of widespread neuronal deletion (Prlr^lox/lox/Camk2a^Cre) on fertility were largely recapitulated by the arcuate kisspeptin-specific model, it was apparent that the global deletion was more effective at inducing the return to estrus during lactation (in 100% of animals by day 10), compared to the arcuate kisspeptin-specific model (63% by day 10, and 83% by day 19). This may be due to the absence of lactation-induced suppression of Kiss1 expression in RP3V kisspeptin neurons of Prlr^lox/lox/Camk2a^Cre mice. Similarly, we cannot rule out the possibility that other populations of prolactin-sensitive neurons, such as GABA or dopamine neurons ⁶⁰, may contribute to suppressing estrous cycles during lactation. Nevertheless, our data collectively provide strong evidence that prolactin action on arcuate kisspeptin neurons is the primary factor mediating lactation-induced infertility in mice. Given that hyperprolactinemia induces infertility in humans and many other species ^{21, 22, 66–75}, it is likely that a conserved mechanism will be contributing to lactational infertility in all mammalian species.

Acknowledgements

We would like to acknowledge the research assistance of Zin Khant-Aung, genotyping by Pene Knowles, and Dr Joon Kim for his assistance with analysis of fibre photometry data.

Financial support

This work was supported by Health Research Council of New Zealand (grant number: 21-560) and the Lions Club of Dunedin South - administered by Perpetual Guardian (Otago Medical Research Foundation).

Author Contributions

ECRH: Designed research, completed research, analysed data, wrote the first draft of the manuscript, edited the manuscript.

SRL: Designed research, completed research, analysed data, edited the manuscript. JC: Completed research, analysed data, edited the manuscript.

HJM: Completed research, edited the manuscript. UB: Provided key resources, edited the manuscript.

AEH: Designed research, analysed data, edited the manuscript. RSEB: Designed research, analysed data, edited the manuscript.

DRG: Designed research, provided key resources, analysed data, edited the manuscript.

Competing Interest Statement

The authors have no competing interests to declare.

Supplementary data

Supplementary Figure 1.

Supplementary Figure 2.

Supplementary Figure 3.

Supplementary Figure 4:

Supplementary Figure 5.

Supplementary Table 1.