Identification of Genes Enriched in GnRH Neurons by Translating Ribosome Affinity Purification and RNAseq in Mice

Abstract

Gonadotropin-releasing hormone (GnRH) neurons are a nexus of fertility regulation. We used translating ribosome affinity purification coupled with RNA sequencing to examine messenger RNAs of GnRH neurons in adult intact and gonadectomized (GDX) male and female mice. GnRH neuron ribosomes were tagged with green fluorescent protein (GFP) and GFP-labeled polysomes isolated by immunoprecipitation, producing one RNA fraction enhanced for GnRH neuron transcripts and one RNA fraction depleted. Complementary DNA libraries were created from each fraction and 50-base, paired-end sequencing done and differential expression (enhanced fraction/depleted fraction) determined with a threshold of >1.5- or <0.66-fold (false discovery rate P ≤ 0.05). A core of ∼840 genes was differentially expressed in GnRH neurons in all treatments, including enrichment for Gnrh1 (∼40-fold), and genes critical for GnRH neuron and/or gonadotrope development. In contrast, non-neuronal transcripts were not enriched or were de-enriched. Several epithelial markers were also enriched, consistent with the olfactory epithelial origins of GnRH neurons. Interestingly, many synaptic transmission pathways were de-enriched, in accordance with relatively low innervation of GnRH neurons. The most striking difference between intact and GDX mice of both sexes was a marked downregulation of genes associated with oxidative phosphorylation and upregulation of glucose transporters in GnRH neurons from GDX mice. This may suggest that GnRH neurons switch to an alternate fuel to increase adenosine triphosphate production in the absence of negative feedback when GnRH release is elevated. Knowledge of the GnRH neuron translatome and its regulation can guide functional studies and can be extended to disease states, such as polycystic ovary syndrome.

Characterization of the translatome of preoptic GnRH neurons reveals a core of ∼800 differentially expressed genes; these genes are affected by both sex and gonadal status in adults.

Gonadotropin-releasing hormone (GnRH) neurons are required for fertility. Failure of these neurons to mature and/or function properly can result in reproductive dysfunction, such as idiopathic hypogonadotropic hypogonadism or polycystic ovary syndrome, the most common form of infertility in women (1–7). Pulsatile release of GnRH into the hypophyseal portal system acts on the pituitary to stimulate the synthesis and secretion of the gonadotropins luteinizing hormone (LH) and follicle-stimulating hormone, which in turn, regulate gonadal steroidogenesis and gametogenesis. Both GnRH secretion and GnRH neuronal activity are regulated by steroid feedback (8–15).

In adult mice, there are ∼800 GnRH neurons scattered throughout the preoptic area (POA) and anterior hypothalamus (16, 17). The advent of transgenic mice synthesizing green fluorescent protein (GFP) under the control of the GnRH promoter (18–20) advanced the study of GnRH neurons and allowed for the examination of their electrophysiological properties, pharmacology, and to a limited extent, gene transcription profiles (21–25). Although powerful, single-cell polymerase chain reaction (PCR) requires laborious cell harvest, is low throughput and qualitative, and typically requires a priori selection of transcripts of interest to allow amplification. There is heterogeneity in both gene expression and electrophysiologic parameters of GnRH neurons (18, 26, 27). It is thus challenging to collect sufficient individual cells to develop a representative transcriptional profile of the GnRH neuronal population. Fluorescence-activated cell sorting (FACS) of GnRH neurons overcomes many of the limitations of single-cell PCR in that hundreds of cells can be collected and transcripts simultaneously profiled (28). FACS, however, requires proteolytic digestion and likely results in the loss of extracellular protein domains (such as receptors) and neuronal processes; these changes may alter signaling into the cell and thus, alter gene expression before the sample is captured.

To expand our knowledge of the GnRH neuron and help overcome some of the obstacles of single-cell PCR and FACS, we set out to examine the GnRH neuron translatome, defined as RNAs that are actively associated with ribosomes. We used translating ribosome affinity purification (TRAP) combined with RNA sequencing (RNAseq; TRAPseq). TRAPseq has the advantages of examining cell-specific transcripts and unbiased profiling of transcripts in the cells of interest (29–33). We harnessed the power of mouse genetics and cre-lox technology to express a GFP-tagged ribosome protein in a GnRH–neuron-specific manner (34, 35). TRAPseq was done on POA tissue from both intact and gonadectomized (GDX) female and male mice. We had three goals: first, to identify gene transcripts that are enriched or de-enriched in GnRH neurons; second, to identify populations of transcripts that are sensitive to sex and/or gonadal factors; and third, to relate changes in GnRH neuron transcripts to normal reproductive physiology.

Materials and Methods

Definitions

Throughout this work, RNA fractions are referred to as enhanced or depleted for GnRH–neuron transcripts, whereas genes differentially expressed between these two RNA fractions are referred to as enriched or de-enriched in GnRH neurons (Fig. 1A).

Figure 1.

Experimental design and quantitative PCR confirmation of GnRH neuron messenger RNA enrichment by TRAP on samples used for RNAseq. (A) Diagram of experimental flow defining GnRH neuron–depleted vs –enhanced RNA fractions and enriched vs de-enriched transcripts. (B) Representative raw PCR amplification graphs for (upper) Gnrh1 and (lower) glial fibrillary acidic protein (Gfap) with β-actin (Actb) (both graphs) as a housekeeping gene. (C) Means ± standard error of the mean (n = 13 biological replicates) threshold cycles for Actb, Gnrh1, and Gfap from GnRH neuron–enhanced and –depleted samples. (D) Expression of Gnrh1 and Gfap normalized to Actb. *P < 0.001 enhanced vs depleted, two-way analysis of variance/Bonferroni. (E) Fold enrichment (enhanced/depleted) for Actb, Gnrh1, and Gfap.

Animals

Mice were provided with water and Harlan 2916 chow ad libitum and were held on a 14-hour light:10-hour dark light cycle with lights on at 0400 Eastern Standard Time. Homozygous GnRH-cre mice (C57BL/6 background) (36) were crossed with homozygous R26-loxSTOPlox-L10-GFP mice (mixed C57BL/6/129 background (34, 35) to generate mice expressing GFP-labeled L10a ribosome protein targeted to GnRH neurons (GnRH-cre/L10a-GFP mice). Adult male and female mice (aged 55 to 165 days) were used. Some mice were GDX under isoflurane anesthesia (VetOne Fluoriso, Boise, ID), followed by bupivacaine (Sensorcaine MPF; APP Fresenius Kabi, Lake Zurich, IL) as a local analgesic. Vaginal cytology was monitored ≥7 days before tissue collection in intact females to determine estrous cycle stage. The Institutional Animal Care and Use Committee of the University of Michigan approved all procedures.

Confirmation of L10a-GFP expression in GnRH neurons

GnRH-cre/L10a-GFP mice (n = 3/treatment) were perfused transcardially with 4% paraformaldehyde and brains removed. Brains were cryopreserved overnight in 20% sucrose, then frozen in optimum cutting temperature (Tissue-Tek; Sakura Finetek, Torrance, CA), and sectioned at 30 µm into four series on a cryostat (Leica, Buffalo Grove, IL). One series from just caudal to the olfactory bulb through the optic chiasm was processed for free-floating dual immunofluorescence using standard procedures (37). The primary antibodies (Table 1) used were the following: chicken anti-GFP (ab13970, 1:1000; Abcam, Cambridge, UK) and rabbit anti-GnRH (EL-14, 1:4000; generous gift from Dr. Oline Ronnekleiv, Oregon Health & Science University, Portland, OR) (38). Primary antibodies were visualized with Alexa Fluor-488 conjugated goat anti-chicken and Alexa Fluor 546-conjugated goat anti-rabbit (A-11039 and A-11035, respectively; Thermo Fisher Scientific, Waltham, MA), respectively. Sections were mounted on Superfrost Plus glass slides (Thermo Fisher Scientific) and coverslipped with ProLong Gold antifade reagent containing 4′,6-diamidino-2-phenylindole (Thermo Fisher Scientific). Immunofluorescence was detected using a fluorescent Axio Imager microscope (Zeiss, Jena, Germany). The number of cells expressing GnRH, GFP, or both was counted in the three sections surrounding the organum vasculosum of the lamina terminalis containing the highest number of GnRH neurons. The region contained within the tissue punch used for TRAPseq was focused on, defined as ventral to the anterior commissure, and extended 0.6 mm lateral from midline on each side.

Table 1.

Antibody Table

Peptide Target	Antigen Sequence	Name of Antibody	Source, Catalog Number, RRID Information (if Available)	Species/Type
GFP, immunohistochemistry	Recombinant full-length protein corresponding to GFP	Anti-GFP antibody	Abcam, ab13970, RRID: AB_300798	Chicken/polyclonal
GnRH	GnRH conjugated to bovine serum albumin	EL-14	Dr. Oline Ronnekleiv, Oregon Health & Science University, RRID: AB_2715535	Rabbit/polyclonal
GFP, TRAP	GFP	Heintz laboratory TRAP anti-GFP 19F7 antibody (31)	Antibody and Bioresource Core Facility, Memorial Sloan Kettering Cancer Center, Htz-GFP-19F7, RRID: AB_2716736	Mouse/monoclonal
GFP, TRAP	GFP	Heintz laboratory TRAP anti-GFP 19C8 antibody (31)	Antibody and Bioresource Core Facility, Memorial Sloan Kettering Cancer Center, Htz-GFP-19C8, RRID: AB_2716737	Mouse/monoclonal

Abcam, Cambridge, UK; Oregon Health & Science University, Portland, OR; Antibody and Bioresource Core Facility, Memorial Sloan Kettering Cancer Center, New York, NY.

Abbreviation: RRID, research resource identifier.

Experimental design

TRAPseq was performed on four groups of adult mice: intact males, intact females (diestrus), orchidectomized (ORC) males, and ovariectomized (OVX) females; experiments on GDX mice were done 7 days postsurgery. Preliminary experiments determined it was necessary to pool POA punches from five to seven mice per treatment for RNAseq. Each pooled set of POA punches is considered a single biological replicate. A minimum of three biological replicates per treatment was used (n = 3 each intact diestrous females, OVX females, ORC males; n = 4 intact males). Steroid bioavailability, or lack thereof, was confirmed by measuring uterine mass in females and serum LH from trunk blood collected at the time of tissue collection from both sexes. Seminal vesicle mass in males was observed, but was sufficiently reduced by eye in castrates that exact mass was not obtained.

TRAP and RNAseq

We followed the published TRAP procedure (29) with modifications. In brief, at the time of tissue collection, mice were anesthetized with isoflurane and decapitated. Brains were quickly removed and placed in an ice-cold, 1 mm coronal brain matrix (Zivic Instruments, Pittsburgh, PA). A coronal slice, extending 2 mm rostral from the caudal extent of the optic chiasm, was placed in a pool of ice-cold dissection buffer [Hanks’ balanced salt solution (1×; Invitrogen/Thermo Fisher Scientific); 2.5 mM HEPES-KOH (Affymetrix/Thermo Fisher Scientific); 4 mM NaHCO₃ (Gibco/Thermo Fisher Scientific); and 35 mM glucose, 0.5 mM dithiothreitol, and 0.1 mg/ml cycloheximide (all from Sigma-Aldrich, St. Louis, MO)]. A Palkovits punch (1.2 mm diameter) was centered on the midventral region to dissect the GnRH neurons. Tissue punches were immediately homogenized in ice-cold lysis buffer [20 mM HEPES-KOH, 150 mM KCl, and 10 mM MgCl₂ (Affymetrix/Thermo Fisher Scientific); 1× EDTA Free Protease Inhibitor and 1.25% volume-to-volume ratio (v/v) Protector RNase Inhibitor (Roche, Indianapolis, IN); 0.625% v/v RNasin (Promega, Madison, WI); 0.625% v/v Superasin (Invitrogen/Thermo Fisher Scientific); and 0.5 mM dithiothreitol and 0.1 mg/ml cycloheximide]. Lysis buffer volume was adjusted for input amounts of POA punches (100 µL lysis buffer per POA). Anti-GFP (HtzGFP-19F7 and HtzGFP-19C8; Antibody and Bioresource Core Facility, Memorial Sloan Kettering Cancer Center, New York, NY)-coated streptavidin magnetic beads (Streptavidin T1 Dynabeads; Invitrogen/Thermo Fisher Scientific) and wash buffers were prepared as described (29–31). As GnRH neurons are few in number, preliminary experiments indicated smaller volumes than recommended of anti-GFP-coated beads reduced contamination with known non-GnRH neuronal RNAs, such as glia-specific RNAs. Bead volume was thus reduced to 25 µL per POA punch. Immunoprecipitation occurred overnight (14 to 16 hours) at 4°C. Polysome-RNA complexes bound to the anti–GFP-coated streptavidin magnetic beads (GnRH neuron specific) were separated from the supernatant by a magnet; RNA was isolated using the RNeasy Micro Kit with on-column DNasing (Qiagen, Valencia, CA). A portion (approximately one-third) of the depleted RNA (supernatant) was also isolated. There are thus two RNA pools for each biological replicate: one that is enhanced (immunoprecipitant) for GnRH RNAs and one that is depleted (supernatant).

RNA quantity and quality were determined by the RNA 6000 Pico Chip (Agilent Technologies, Santa Clara, CA). Despite the pooling of POAs from five to seven mice for each biological replicate, GnRH-enhanced samples were all below the 50-pg/µL threshold for Pico Chip analysis. Because of this, before constructing complementary DNA (cDNA) libraries, 5 µl of GnRH-enhanced and -depleted RNAs were reverse transcribed, blind to concentration (39), and amplified for Gnrh1 (known GnRH-specific transcript), glial fibrillary acidic protein (Gfap; known non-GnRH transcript), and β-actin (Actb; housekeeping transcript) to determine enrichment for Gnrh1 and potential glial contamination. Gnrh1 and Gfap were normalized to Actb [normalized relative expression, comparative cycle threshold method (40)] and enrichment calculated as normalized relative expression in the GnRH-enhanced RNA sample, divided by the normalized relative expression in the GnRH-depleted sample. Representative raw PCR amplification graphs are shown in Fig. 1B; summary data are in Fig. 1C–1E. Gnrh1 was enriched 271 ± 52-fold (n = 13), demonstrated by the right-shift in the Actb amplification curve in the enhanced vs depleted fraction. In contrast, the cycle threshold for Gnrh1 is similar in the enhanced and depleted fractions (Fig. 1B, upper). Gfap was only 2.4 ± 0.3-fold (n = 13) enriched in GnRH-enhanced RNA samples (Fig. 1B, lower). The entirety of the remaining GnRH-enhanced RNA and 5 to 20 ng of the depleted RNA were used to create cDNA libraries with the SMARTer v4 Ultra Low Kit and Low Input DNA Library Prep Kits (Clontech, Mountain View, CA). Libraries were checked for quality (D1000 ScreenTape; Agilent Technologies). Eight samples per lane were 50-base, paired-end sequenced using the HiSeq 4000 platform (Illumina, San Diego, CA) by the University of Michigan DNA Sequencing Core.

RNAseq analysis was done by the University of Michigan Bioinformatics Core. The Tuxedo Suite software package was used for alignment, differential expression analysis, and postanalysis diagnostics (41). Cufflinks/CuffDiff (http://cole-trapnell-lab.github.io/cufflinks/) (42) was used for expression quantitation, normalization, and determination of differential expression using University of California Santa Cruz (Santa Cruz, CA) mm10.fa as the reference genome sequence (http://genome.ucsc.edu/). Enrichment for RNAseq is defined as read counts [fragments per kilobase of transcript per million mapped reads (FPKM)] in the GnRH-enhanced fraction, divided by the FPKM in the GnRH-depleted fraction. Genes and transcripts were identified as being enriched (or de-enriched) based on three criteria: test status = “OK,” false discovery rate P ≤ 0.05 [P value adjusted for false discovery rate (q) ≤ 0.05], and fold change (FC) > 1.5 or < 0.66. Before calculation of the q value, a pseudocount of “1,” was added to all FPKM values. This reduces the problem of false discovery, as a result of a large FC, calculated using values near zero for low-abundance genes (43). All FPKM values shown in the tables and throughout the manuscript are the raw values without the added pseudocount. Genes and isoforms were annotated with National Center for Biotechnology Information (NCBI) Entrez GeneIDs and differentially expressed genes with Gene Ontology (http://www.geneontology.org/) terms using NCBI annotation. Database for Annotation, Visualization and Integrated Discovery (DAVID; version 6.7; http://david.abcc.ncifcrf.gov/) was used to identify functional categories. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways that were different between treatments were analyzed using Advaita’s iPathwayGuide (http://www.advaitabio.com/ipathwayguide). The RNAseq data (GSE110042) were submitted to the NCBI Gene Expression Omnibus database.

Quantitative PCR validation of RNAseq data

Mice were prepared for quantitative PCR (qPCR) validation as for RNAseq mentioned previously. Additionally, four GnRH-cre mice, not expressing floxed L10a-GFP, were included to determine background expression from nonspecific polysome binding to anti-GFP-coated magnetic beads. For these studies, a biological replicate is an individual mouse. TRAP was conducted, as described previously, with two modifications: the lysis buffer volume was 350 µL per POA punch, and the anti-GFP streptavidin magnetic bead volume was 50 µL. RNA was isolated as described previously, and 10 µL was reverse transcribed (39), blind to concentration. cDNA pools (from GnRH-enhanced and -depleted RNA fractions), controls, and a standard curve were preamplified using TaqMan PreAmp Master Mix (Invitrogen/Thermo Fisher Scientific) with primer:probes for TaqMan PrimeTime qPCR assays from Integrated DNA Technologies (Coralville, IA; Table 2). Assay primers span an intron, when possible, to minimize amplification of genomic DNA. cDNA was preamplified for 15 cycles, according to the manufacturer’s recommendation. Preamplified DNAs were diluted 1:10 with Tris-EDTA; a second dilution to 1:100 (final) was created for use with more abundant transcripts. Dilutions were stored at −20°C until used for qPCR, which was performed using 2 to 5 µL diluted, preamplified DNA per reaction, in duplicate, for 40 cycles (TaqMan Gene Expression Master Mix; Invitrogen/Thermo Fisher Scientific). PCR reaction efficiencies were calculated from the slope of the standard curve. To confirm the linearity and parallelism of the preamplification step, 2 µL nonpreamplified standard curve cDNA was also amplified in singlicate. Amplicon size was confirmed by agarose gel electrophoresis or sequencing for custom primer probe sets.

Table 2.

qPCR Primers and Probes

Gene		Sequence	Source	IDT Reference No.
Ar	Fwd	5′ CTGCCTTGTTATCTAGCCTCA 3′	IDT	Mm.PT.47.17416675
	Rev	5′ ATACTGAATGACCGCCATCTG 3′
	Probe	5′ ACCACATGCACAAGCTGCCTCT 3′
Arg1	Fwd	5′ GAATGGAAGAGTCAGTGTGGT 3′	IDT	Mm.PT.58.8651372
	Rev	5′ AGTGTTGATGTCAGTGTGAGC 3′
	Probe	5′ ACAGTCTGGCAGTTGGAAGCATCT 3′
Bahcc1	Fwd	5′ GGTAGGAAGATGGATGCCAGA 3′	IDT	Mm.PT.58.12197037
	Rev	5′ CCACATCCTCTCCCTCTGA 3′
	Probe	5′ CACTCCCATGAAGCTCCTGCGTAC 3′
Egfl7	Fwd	5′ CAGCCTTACCTCACCACTTG 3′	IDT	Mm.PT.58.6782252
	Rev	5′ CAAGCATAGCGAGGCCTT 3′
	Probe	5′ CAGCACCTACCGAACCATCTACCG 3′
Epcam	Fwd	5′ GGCATTAAGCTCTCTGTGGAT 3′	IDT	Mm.PT.58.11851150
	Rev	5′ TGGTGTCATTAGCAGTCATCG 3′
	Probe	5′ ACCCATCTCCTTTATCTCCGACTTCTCA 3′
Esr1	Fwd	5′ CCTGTTTGCTCCTAACTTGCT 3′	IDT	Mm.PT.47.16003033
	Rev	5′ GAACCGACTTGACGTAGCC 3′
	Probe	5′ TGCCTTCCACACATTTACCTTGATTCCT 3′
Esr2	Fwd	5′ AAGTAGGAATGGTCAAGTGTGG 3′	IDT	Mm.PT.47.17681375
	Rev	5′ CTGGTTCTCTTGGCTTTGTTC 3′
	Probe	5′ AGTACGAAGACAGAGAAGTGCCAGC 3′
Fhl1	Fwd	5′ GTAGCGATTCTTATAATGCACCTC 3′	IDT	Mm.PT.58.9685318
	Rev	5′ GTCACTGCTGCCTGAAGT 3′
	Probe	5′ TTGACAAGTTCTGCGCCAACACC 3′
Gfap	Fwd	5′ CAACCTCCAGATCCGAGAAAC 3′	IDT	Mm.PT.51.11727348
	Rev	5′ GTCCTTGTGCTCCTGCTTC 3′
	Probe	5′ TCCTTAATGACCTCACCATCCCGC 3′
Gnrh1	Fwd	5′ GGGAAAGAGAAACACTGAACAC 3′	Custom
	Rev	5′ AGTACATTCGAAGTGCTGGG 3′
	Probe	5′ GATGGGCAAGGAGGTGGATCAAAT 3′
Hspa8	Fwd	5′ AAGAGCACAGGAAAGGAGAAC 3′	IDT	Mm.PT.58.33325534.gs
	Rev	5′ CTCAGCCTTGTACTTCTCAGC 3′
	Probe	5′ TTACTCAAGCGGCCCTTGTCATTGG 3′
Kiss1	Fwd	5′ CTGCTTCTCCTCTGTGTCG 3′	IDT	Mm.PT.45.16269514
	Rev	5′ TTCCCAGGCATTAACGAGTTC 3′
	Probe	5′ CGGACTGCTGGCCTGTGGAT 3′
Kiss1r	Fwd	5′ CTCACTGCATGTCCTACAGC 3′	IDT	Mm.PT.49a.16255718.g
	Rev	5′ GCCTGTCTGAAGTGTGAACC 3′
	Probe	5′ TCAATCCGCTGCTCTATGCCTTCC 3′
Npffr1	Fwd	5′ GACAATGCCACATGCAAGATG 3′	IDT	Mm.PT.49a.12363343
	Rev	5′ GTCAGCTTCTCACGGAAAGG 3′
	Probe	5′ TCTGCGTCGGTTTTCACACTGGT 3′
Otx1	Fwd	5′ GAAGTCCTTCAAGCTGTTGC 3′	IDT	Mm.PT.58.42606516
	Rev	5′ CTAAACAACCGAGCAAGACAAG 3′
	Probe	5′ AGAGGTAGATGGTGAAAAGCCGCG 3′
Otx2	Fwd	5′ TTGTTCTGACCTCCATTCTGA 3′	IDT	Mm.PT.58.13383028
	Rev	3′ AAATCAACTTGCCAGAATCCAG 3′
	Probe	5′ ATCGAAGAGCTAAGTGCCGCCAA 3′
Oxt	Fwd	3′ CTTGGCTTACTGGCTCTGAC 5′	IDT	Mm.PT.58.29698679.g
	Rev	3′ CCGAAGCAGCGTCCTTT 5′
	Probe	5′ TGTGCTGGACCTGGATATGCGC 3′
Pgr	Fwd	5′ CGCCATACCTTAACTACCTGAG 3′	IDT	Mm.PT.47.10254276
	Rev	5′ CCATAGTGACAGCCAGATGC 3′
	Probe	5′ AGATTCAGAAGCCAGCCAGAGCC 3′
Ppia	Fwd	5′ CAAACACAAACGGTTCCCAG 3′	IDT	Mm.PT.39a.2.gs
	Rev	5′ TTCACCTTCCCAAAGACCAC 3′
	Probe	5′ TGCTTGCCATCCAGCCATTCAG 3′
Rab25	Fwd	5′ TGTCGTGGCTGAACTCATTG 3′	IDT	Mm.PT.58.42268495
	Rev	5′ CGTCTTCACTCTCCATGCTG 3′
	Probe	5′ CACGCCTGACTCGCCGATCA 3′
Rai1	Fwd	5′ CAAGAGAAAAGGGAGACTGAGAG 3′	IDT	Mm.PT.58.9078821
	Rev	5′ TCTCGAAAAGACTGCATGACT 3′
	Probe	5′ AGGAGCCCGCAGATAACCAGC 3′
Six1	Fwd	5′ GAGAGAGTTGATTCTGCTTGTTG 3′	IDT	Mm.PT.58.13903884
	Rev	5′ GGTCAGCAACTGGTTTAAGAAC 3′
	Probe	5′ CGAGGCCAAGGAAAGGGAGAACA 3′
Six6	Fwd	5′ GGTCCCTCACCCTCCGA 3′	IDT	Mm.PT.58.43583201
	Rev	5′ CAAAAACCGCAGACAAAGAGAC 3′
	Probe	5′ TGCAGCCAAAAACAGACTCCAGC 3′
Tmprss6	Fwd	5′ CCCAAAGGAGTAGACAGAACTAGA 3′	IDT	Mm.PT.58.42367147
	Rev	5′ CTATTGCTTTCCGCAGTGAATC 3′
	Probe	5′ CCAAAGCCCAGAAGATGCTCCAAGA 3′
Tnrc18	Fwd	5′ AAGAACTGCTCTACCTGCAAC 3′	IDT	Mm.PT.58.11343509
	Rev	5′ ATCTTCTCCTCCATCTCCGT 3′
	Probe	5′ CTCTTGTGCTTTCAACCGCTCCAC 3′’
Tstd1	Fwd	5′ GCGATTTCCAATTCAGACACAG 3′	IDT	Mm.PT.58.10697454.g
	Rev	5′ CGTTCACTCGTAGCCTCAG 3′
	Probe	5′ TGTTGAGTGCCCCAGGAATGGTAC 3′

Primer probe sets that may amplify genomic DNA are indicated in bold.

Abbreviations: Fwd, forward; IDT, Integrated DNA Technologies; Rev, reverse.

Assays

For all blood samples, serum was separated by centrifugation and stored at −20°C until assayed. All assays were conducted by the University of Virginia Center for Research in Reproduction Ligand Assay Core (Charlottesville, VA). LH was measured in singlicate by a sensitive two-site sandwich immunoassay (44, 45) using monoclonal antibodies against bovine LH (no. 581B7) and against the human LH–β subunit (no. 5303: Medix, Kauniainen, Finland), as described (45). Intra- and interassay coefficients of variation were 4.5% and 8.3%, respectively. Mouse LH reference prep (AFP5306A; provided by Dr. A. F. Parlow and the National Hormone and Peptide Program, Harbor University of California Los Angeles Medical Center, Los Angeles, CA) was used as standard. The assay has a sensitivity of 0.04 ng/mL LH. Serum estradiol levels, in females, were measured in singlicate using the CalBiotech Estradiol ELISA kit (ES180S-100; Calbiotech, El Cajon, CA). Assay sensitivity was 3 pg/mL estradiol. Intra- and interassay coefficients of variation were 6.1% and 8.9%, respectively.

Statistics

Data for parameters other than TRAPseq are reported as means ± standard error of the mean (SEM). Statistical analyses were performed using Prism 7 (GraphPad Software, La Jolla, CA). Normality of data was analyzed using a Shapiro-Wilk test. Specific statistical tests were dictated by experimental design, and data distribution is indicated in the figure legends. Significance was set at P < 0.05.

Results

Validation of animal models

Serum LH in both sexes and estradiol and uterine mass for females are shown in Fig. 2; each point is the mean of the samples within a biological replicate. Serum LH was low in intact female and male groups and was increased 7 days after either ovariectomy or orchidectomy. Serum estradiol and uterine mass were greater in intact rather than OVX females; uterine mass was below the proestrous range (46).

Figure 2.

Validation of intact vs castrate biological replicates. Individual values and means ± SEM for (A) serum LH, (B) serum estradiol, and (C) uterine mass. Different letters indicate P < 0.0001 within each panel using two-way analysis of variance, (A) Bonferroni, or (B and C) Mann-Whitney U test.

To examine the fidelity of L10a-GFP expression in GnRH neurons, dual immunofluorescence for GnRH and GFP was used. GnRH-immunoreactive cells were identified in the expected regions (47) from the diagonal band of Broca through the anterior hypothalamus (Fig. 3). All cells positive for GnRH were also positive for GFP. Between 81% and 90% of GFP-labeled cells was immunoreactive for GnRH (intact female 87% ± 3%, OVX 81% ± 6%, intact male 86% ± 5%, ORC 90% ± 4%, n = 3/group). These TRAP RNA samples should thus be viewed as enhanced for GnRH neurons, not purified for these cells.

Figure 3.

Colocalization of GnRH and GnRH cre-driven L10a-GFP immunoreactivity in TRAP mice. Representative images from POA showing (A) GFP and (B) GnRH immunostaining and (C) overlay. *GFP-positive cell that is not immunopositive for GnRH. Note all GnRH-positive cells were also GFP positive.

TRAPseq reveals a core of genes differentially expressed in GnRH neurons

TRAPseq enriches RNAs that are in the process of being translated from other RNAs, generating two RNA samples per biological replicate: one that is enhanced for GnRH neuron RNAs and one that is depleted. This depleted pool contains RNAs from other cell types and GnRH neuron RNAs not associated with polysomes at the time of tissue harvest. The mean sequencing read yield for each sample was ∼48 million (±1.5 million, n = 26), of which ∼83% was mapped. All samples were highly enriched for Gnrh1 (see later).

We analyzed the data in three different ways. First, we examined enrichment vs de-enrichment in GnRH neurons across all four groups (Fig. 4A). Second, we examined enrichment vs de-enrichment in intact vs GDX samples (males and females combined; Fig. 4B). Third, we examined enrichment vs de-enrichment in males vs females (intact and GDX combined; Fig. 4C). When all groups were examined independently, there was a core of 840 genes that were differentially expressed in all the treatments, 333 of which were enriched in GnRH neurons and 507 of which were depleted (Fig. 4A). Of note, de-enriched (expressed but at levels lower than surrounding tissue) is different than not detected.

Figure 4.

Enrichment or de-enrichment of transcripts in GnRH neurons identified by RNAseq depends on sex and gonadal status. (A) Venn diagrams of the number of transcripts (left) enriched in GnRH neurons and (right) de-enriched from GnRH neurons as a function of treatment. (B and C) Venn diagram of the number of transcripts (left in each) enriched and (right in each) de-enriched as a function of (B) gonadal status or (C) sex.

Genes highly enriched in GnRH neurons

The most highly enriched genes in the GnRH neuron are shown in Table 3. Most of these genes have not, to our knowledge, been identified as expressed in GnRH neurons, with the exceptions of Gnrh1 (∼39-fold enriched) and the transcription factors Otx1 and Six1 (∼41- and 13-fold enriched, respectively), which are important in GnRH or gonadotrope development and/or physiology (48–50). Besides Gnrh1, the most highly enriched transcripts are a mixture of transcription factors (Otx1, Mn1, Six1, and Rai1), regulators of transcription factors (Bcor and Kdf1), genes associated with neurite growth and/or cytoskeleton (Prph, Gse1, and Ccdc68), cell–cell contact (Epcam and Cld9), enzymes (Rab25, Arg1, Tstd1, and Tmprss6), calcium binding (Cabp4), and DNA/chromatin-binding elements (Bahcc1, Tnrc18, and Prr12). Many of the most highly enriched genes in the GnRH neuron are expressed at very low levels in the depleted RNA (see Supplemental data files), raising the question of whether enrichment was attributable to division of the enhanced RNA values by a low number. A selection of these highly enriched transcripts was thus examined by qPCR in TRAP samples from individual POAs (Fig. 5A). Enrichment was confirmed for eight of the 11 genes examined: Rab25, Otx1, Gnrh1, Tstd1, Epcam, Tmprss6, Arg1, and Six1. We could not confirm enrichment for Bahcc1, its paralogue Tnrc18, or Rai1, as a result of high background expression in the bead-bound RNA fraction in samples from mice not expressing L10a-GFP (Fig. 5A). No further verification was done on Cldn9, Kdf1, Bcor, Gse1, Prr12, Ccdc68, Prph, Mn1, or Cabp4; thus, their enrichment in GnRH neurons remains to be confirmed.

Table 3.

The 20-Most Enriched Transcripts in GnRH Neurons From TRAPseq With a q Value < 0.001

Gene	NCBI No.	Fold ∆
		Intact	GDX
Rab25	53868	45	64
Otx1	18423	34	45
Bahcc1	268515	36	42
Gnrh1	14714	39	39
Tstd1	226654	40	37
Epcam	17075	38	37
Tmprss6	71753	28	38
Tnrc18	231861	28	36
Arg1	11846	32	31
Cldn9	56863	25	34
1810019J16Rik (Kdf1)	69073	30	28
Bcor	71458	25	33
Gse1	382034	25	29
Prr12	233210	22	28
Ccdc68	381175	20	28
Prph	19132	26	19
Mn1	433938	19	24
Six1	20471	21	20
Cabp4	73660	22	17
Rai1	19377	15	22

Abbreviation: Fold ∆, FC.

Figure 5.

qPCR validation of (A) highly enriched and (B) other selected transcripts. Means ± SEM gene expression normalized to the mean of Ppia, Fhl1, and Hspa8 of preamplified cDNA from single POA TRAP. *P < 0.05, two-way analysis of variance/Bonferroni. IP, immunoprecipitate.

To test for potential contamination of the TRAPseq samples with glial RNAs, we examined the top 40 genes that were enriched >1.5-fold in either astrocytes or oligodendrocytes, as identified by Cahoy et al. (51); the top 15 genes for each are shown in Fig. 6. Most astrocyte and oligodendrocyte markers were de-enriched or not differentially expressed in the GnRH neuron-enhanced RNA fraction, with the exception of mild enrichment of the glial marker Gfap in GDX and male mice; this was also observed in the qPCR data from single POA TRAP (Fig. 5B). Additionally, there is enrichment of the actin-binding protein gelsolin (Gsn); interestingly, Gsn has been identified in neurons as important for neurogenesis, and Gsn-deficient mice have retarded migration of olfactory neurons (52).

Figure 6.

Heat map of selected marker genes encoding (upper) astrocyte and (lower) oligodendrocyte markers in GnRH neuron TRAPseq samples. Legend is log₂ of the quotient of the FPKM of the enhanced divided by the depleted sample. Genes that are not differentially expressed (−0.58 > FC < 0.58 log₂ and/or q value > 0.05) are in gray, enriched genes are red, and de-enriched genes in green. INT, intact.

Genes highly expressed in GnRH neurons

The most highly expressed transcripts in GnRH neurons (as determined by FPKM in the enriched RNA fraction), regardless of enrichment/de-enrichment, are shown in Table 4. Gnrh1 is the most highly expressed transcript; the average FPKM for Gnrh1 is approximately four to five times greater than the next most highly expressed gene, the ubiquitous protein and common housekeeper, cyclophilin (Ppia). Gnrh1 is the only gene to be both highly enriched (FC > 4) and highly expressed in GnRH neurons. The rest of the top 20 most highly expressed transcripts are dominated by genes that are important in regulating cellular functions, including the cytoskeleton (Tmsb4x, Tmsb10, and Stmn1), metabolism/energy (Usmg5, Ckb, Cox4i1, Aldoa, Cox8a, Ndufa4, and Atpif1), protein synthesis/degradation (Rpl41, Rpl13, Ubb, and Uchl1), intracellular signaling (Ppia, Calm2, and Gnas), transcription (Bex2), or iron regulation (Fth1).

Table 4.

The 20 Most Highly Expressed Transcripts in GnRH Neurons

Gene	NCBI No.	Intact		GDX
		q≤	Fold ∆	q≤	Fold ∆
Gnrh1	14714	0.001	39	0.001	39
Ppia	268373	0.104	1.6	0.543	1.1
Ubb	22187	0.505	2.0	0.123	1.4
Usmg5	66477	0.001	2.5	0.041	1.3
Fth1	14319	0.652	1.2	0.114	0.8
Tmsb4x	19241	0.021	1.6	0.744	0.9
Rpl41	67945	0.001	3.6	0.001	1.6
Cox4i1	12857	0.002	1.8	0.544	1.1
Aldoa	11674	0.480	1.3	0.938	1.0
Bex2	12069	0.001	2.3	0.001	1.6
Tmsb10	19240	0.001	1.9	0.901	1.0
Cox8a	12868	0.001	2.0	0.262	1.2
Uchl1	22223	0.004	2.1	0.007	1.5
Ndufa4	17992	0.001	1.9	0.273	1.2
Calm2	12314	0.569	1.2	0.873	1.0
Stmn1	16765	0.002	2.0	0.031	1.4
Rpl13	270106	0.001	2.6	0.004	1.4
Ckb	12709	0.988	1.0	0.607	0.9
Atpif1	11983	0.001	2.6	0.003	1.4
Gnas	14683	0.620	1.2	0.831	1.1

Genes that are also enriched in GnRH neurons are in bold. Enrichment, FC ≥ 1.5 with q value < 0.05.

Gonadectomy affects gene enrichment/de-enrichment in GnRH neurons in both sexes

Several of the genes highly expressed in GnRH neurons were altered by GDX (bold in Table 4). Tmsb4x and Tmsb10 are enriched in samples from intact but not GDX mice. Five other highly expressed genes that are enriched in intact but not GDX samples are Cox4i1, Cox8a, Ndufa4, Atpif1, and Usmg5, all of which are associated with the mitochondrial electron transport chain.

Expression of nuclear receptors in GnRH neurons

GnRH neuron functions are sensitive to steroid feedback, but lack of detection of steroid receptors (other than estrogen receptor β) in GnRH neurons using histochemical or in situ approaches has led to the prevailing view that most of these effects are mediated via steroid-sensitive afferents (53). Expression of steroid receptors and other members of the nuclear receptor superfamily from TRAPseq are shown in Table 5 and Fig. 5B. Expression levels of the classic transcription factor steroid receptors Ar, Esr1, Esr2, and Nr3c1 (glucocorticoid receptor); Nr3c2 (mineralocorticoid receptor); and Pgr were low in both enhanced and depleted RNA fractions (see accompanying datasets); Ar, Esr1, and Nr3c2 were nonetheless de-enriched in GnRH neurons across all treatments, and Nr3c1 was de-enriched in samples from intact mice of both sexes. This de-enrichment is sufficiently large in that it can be detected within each sex. Interestingly, expression of Esr2, which is identified in GnRH neurons of rodents, humans, and sheep (54–56), was low and de-enriched in GnRH neurons from intact mice if both sexes were pooled but not if each sex was independently analyzed. Isoforms of the thyroid hormone receptor (Thra and Thrb) were relatively abundant in the POA punch; Thra was de-enriched in GnRH neurons, but Thrb was neither enriched nor de-enriched. The vitamin D receptor (Vdr) receptor was not detected by TRAPseq. Steroids can also act via nongenomic mechanisms to regulate GnRH neurons. Gpr30, a G-coupled orphan receptor with reported roles in the GnRH neuron (57–60), was expressed at low levels but not differentially expressed in GnRH neurons.

Table 5.

Expression of Nuclear Receptor Superfamily Members in GnRH Neurons

Gene	NCBI No.	Intact		GDX
		Fold ∆	q≤	Fold ∆	q≤
Ar	11835	0.560	<0.01	0.605	0.01
Esr1	13982	0.391	<0.01	0.346	<0.01
Esr2	13983	0.263	0.02	1.081	0.90
Gpr30	76854	1.397	0.73	1.143	0.91
Nr3c1	14815	0.560	<0.01	0.810	0.11
Nr3c2	110784	0.484	<0.01	0.472	<0.01
Pgr	18667	0.828	0.18	0.826	0.26
Thra	21833	0.640	<0.01	0.629	<0.01
Thrb	21834	0.990	0.97	1.170	0.40

Genes that are de-enriched are in bold. No genes in this group were enriched in GnRH neurons.

Expression of neuropeptides and their receptors in GnRH neurons

GnRH neurons are regulated by neuropeptides. The expression of select neuropeptides and their receptors identified in GnRH neurons by TRAPseq are shown in Table 6 and Fig. 5B. The kisspeptin receptor Kiss1r was only enriched in GnRH neurons from intact mice. Receptors for neurokinin B and dynorphin (Tacr3 and Opkr1) were detected in all groups but de-enriched in GnRH neurons from intact mice (Tacr2 was extremely low, FPKM < 0.1). Genes encoding kisspeptin and neurokinin B peptides were detected in GnRH neurons but not enriched or de-enriched, whereas the more widely expressed dynorphin was de-enriched. In contrast, the gene encoding galanin, Gal, was enriched in GnRH neurons from intact and female mice, consistent with its immunochemical detection in these cells (61). Gal1r is de-enriched in samples from intact mice. The receptor for the gonadotropin inhibitory hormone, Npffr1 is enriched (FC, approximately 4) in all treatments, highlighting a potential role of this peptide in regulating GnRH neurons (62, 63). Enrichment of Npffr1 by qPCR was not observed, however, in single POA TRAP.

Table 6.

Expression of Neuropeptides and Their Receptors in GnRH Neurons

Gene	NCBI No.	Intact		GDX
		Fold ∆	q≤	Fold ∆	q≤
Kiss1	280287	1.40	0.13	0.97	0.96
Kiss1r	114229	4.18	0.04	3.53	0.19
Tac1	21333	1.20	0.13	0.81	0.11
Tacr1	21336	0.54	<0.01	0.70	0.06
Tac2	21334	1.02	0.93	0.77	0.25
Tacr3	21338	0.36	<0.01	0.67	0.07
Pdyn	18610	0.58	<0.01	0.62	0.01
Oprk1	18387	0.31	<0.01	0.80	0.37
Gal	14419	1.72	<0.01	1.21	0.18
Galr1	14427	0.42	0.02	0.86	0.77
Galr3	14429	0.85	0.94	0.92	0.98
Npvf	60531	8.09	0.18	2.91	0.60
Npffr1	237362	4.36	<0.01	4.17	<0.01
Adcyap1	11516	0.60	<0.01	0.64	<0.01
Adcyap1r1	11517	0.52	<0.01	0.80	0.10
Npy	109648	0.87	0.49	0.68	0.04
Npy1r	18166	0.61	0.01	0.80	0.32
Npy2r	18167	0.77	0.17	0.52	0.00
Npy5r	18168	0.64	0.15	0.92	0.83
Lepr	16847	0.74	0.44	0.61	0.17
Insr	16337	1.18	0.37	1.05	0.86

Genes that are enriched or de-enriched are in bold.

Pathway analysis of differentially enriched genes in GnRH neurons

In addition to individual genes, examination of pathways regulated by sex and gonadal factors can inform us about the GnRH neuron (Table 7). The regulation of the oxidative phosphorylation pathway (KEGG 00190) in GnRH neurons by gonadal factors is particularly striking; 58 genes encompassing components of all five complexes of the electron transport chain are enriched in GnRH neurons from intact, but not GDX, mice (Fig. 7). Whereas the fold enrichment is not high (approximately twofold) in intact mice, the uniformity of this shift is notable. Reduced dependence on oxidative phosphorylation in GDX mice might indicate a shift to glycolysis. We examined the regulation of glucose transporters (Gluts), as glucose entry into the cell is a point of control for glycolysis. The Glut Slc2a1 (Glut1) is enriched in GnRH neurons from GDX relative to intact mice in both sexes (2.4-fold in OVX and 2.7-fold in ORC). In contrast to the enrichment of Glut1, Slc2a3 (Glut3), the main neuronal Glut (64), is de-enriched in samples from intact mice (0.65-fold). Additionally, Slc2a8 is enriched, and Slc2a13 (H⁺-myo-inositol cotransporter) is de-enriched in both intact and GDX mice.

Table 7.

The Most Differentially Regulated Pathways in GnRH Neurons

KEGG Identifier	Pathway Name	q Values
		Intact	GDX
All GnRH neurons
4728	Dopaminergic synapse	1.47E-04	4.21E-04
4024	cAMP signaling pathway	1.47E-04	9.64E-04
4723	Retrograde endocannabinoid signaling	1.49E-04	1.00E-03
5032	Morphine addiction	2.07E-04	5.01E-04
4727	GABAergic synapse	3.01E-04	9.64E-04
Intact only
0190	Oxidative phosphorylation	1.96E-13	1.000
4260	Cardiac muscle contraction	8.77E-05	0.970
4932	Nonalcoholic fatty liver disease	8.77E-05	0.970
5010	Alzheimer’s disease	9.16E-05	0.604
5012	Parkinson disease	1.44E-04	1.000
5016	Huntington disease	3.01E-04	0.533
GDX only
4010	MAPK signaling pathway	0.243	2.63E-03
4360	Axon guidance	0.157	2.63E-03
4310	Wnt signaling pathway	0.053	6.10E-03
4931	Insulin resistance	0.069	6.15E-03
5222	Small cell lung cancer	0.167	7.03E-03
Males only		Males	Females
0190	Oxidative phosphorylation	9.88E-07	0.833
4932	Nonalcoholic fatty liver disease	8.87E-05	0.492
5010	Alzheimer’s disease	1.23E-04	0.919
5012	Parkinson disease	1.50E-04	0.626
5016	Huntington disease	3.80E-04	0.513
Females only
5205	Proteoglycans in cancer	0.131	0.007
4720	Long-term potentiation	0.054	0.007
4261	Adrenergic signaling in cardiomyocytes	0.063	0.010
4918	Thyroid hormone synthesis	0.099	0.020
4015	Rap1 signaling pathway	0.063	0.026

Significant q values are in bold.

Abbreviations: cAMP, cyclic adenosine monophosphate; GABAergic, pertaining to γ-aminobutyric acid; MAPK, mitogen-activated protein kinase; Rap1, Ras-related protein 1.

Figure 7.

Heat map of genes encoding mitochondrial complexes I to V (see bold Roman numerals on left; KEGG 00190) in GnRH neuron TRAPseq samples. Legend is log₂ of the quotient of the FPKM of the enhanced divided by the depleted sample. Genes that are not differentially expressed (−0.58 > FC < 0.58 log₂ and/or q value > 0.05) are in gray, enriched genes are in red, and de-enriched genes are in green.

Pathways for dopamine, γ-aminobutyric acid, glutamate, and retrograde endocannabinoid transmission are de-enriched in GnRH neurons

The pathway with the most substantial regulation in all treatments (i.e., pan-GnRH neuron as opposed to intact vs GDX, as observed for oxidative phosphorylation) is the dopaminergic synapse (KEGG 04728); of 135 annotated genes, between 32 and 47 are differentially regulated in this pathway across treatment groups, and all but four of these are de-enriched in GnRH neurons compared with the surrounding tissue, including the dopamine receptors Drd1, Drd2, and Drd5 (Fig. 8). De-enrichment of many genes in the dopaminergic pathway, such as Slc18a2, the vesicular monoamine transporter 2, is not unexpected, as these genes are associated with the presynaptic terminal, and GnRH neurons have not been reported to be dopaminergic (65, 66) and are not enriched for tyrosine hydroxylase. The four genes that are enriched in GnRH neurons from this pathway are γ subunits of the guanine nucleotide-binding protein: Gng3, Gng5, Gng8, and Gng13. Consistent with receptor de-enrichment, genes associated with intracellular signaling cascades downstream of the dopamine receptor systems are also de-enriched. Likewise, genes associated with both sides of the retrograde endocannabinoid signaling pathway (KEGG 04723) are largely de-enriched in GnRH neurons (Fig. 8).

Figure 8.

Heat map of genes encoding the retrograde endocannabinoid signaling (KEGG 04723; left) and dopaminergic synapse (KEGG 04728; right) pathways in GnRH neuron TRAPseq samples. Legend is log₂ of the quotient of the FPKM of the enhanced divided by the depleted sample. Genes that are not differentially expressed (−0.58 > FC < 0.58 log₂ and/or q value > 0.05) are in gray, enriched genes are in red, and de-enriched genes are in green.

Likewise, transcripts associated with the two most prevalent fast synaptic transmission signaling pathways, γ-aminobutyric acid (GABA; KEGG 04727) and glutamate (KEGG 04724), are both de-enriched in GnRH neurons, including both pre- and postsynaptic components (Fig. 9).

Figure 9.

Heat map of genes encoding the (left) GABAergic synapse (KEGG 04727) or (right) glutamatergic synapse (KEGG 04724) pathways in GnRH neuron TRAPseq samples. Legend is log₂ of the quotient of the FPKM of the enhanced divided by the depleted sample. Genes that are not differentially expressed (−0.58 > FC < 0.58 log₂ and/or q value > 0.05) are in gray, enriched genes are in red, and de-enriched genes are in green.

Discussion

The GnRH neuronal population has been difficult to study using molecular methods because of the paucity of these cells and their wide distribution. Our experimental design for these TRAPseq studies examined the effects of both sex and gonadal factors on the transcriptome of GnRH neurons. Examination of known genes demonstrated appropriate enrichment (e.g., Gnrh1 and Otx1) with low contamination of GnRH-enhanced samples with non-neuronal markers.

The effects of sex (male vs female) were not as dramatic as the effects of gonads (intact vs GDX) on the GnRH neuron transcriptome, suggesting that gonadal factors help maintain the differentiated GnRH neuron phenotype in both sexes. The most striking example of gonadal status altering GnRH neuron translation was upregulation of the oxidative phosphorylation pathway in intact vs GDX mice. Steroid hormones can directly or indirectly regulate mitochondrial respiration (67, 68). The loss of enrichment of electron transport complex genes in GDX mice suggests the hypothesis that removal of gonadal factors, which increases GnRH neuron activity and hormone release, may necessitate a switch to the higher-throughput glycolysis pathway, which produces fewer adenosine triphosphate (ATP) per run but operates at rates >100-fold higher than electron transport. Consistent with this, Slc2a1 (Glut1) was enriched in GDX mice. Although the enrichment of oxidative phosphorylation genes in intact mice was modest (approximately twofold), it was uniform over almost five dozen genes. This difference would be difficult to establish using standard methods, such as qPCR, as twofold enrichment is only about one cycle difference in PCR. Of interest, other pathways that are highly differentially regulated in intact vs GDX mice have substantial mitochondrial electron transport components, including cardiac muscle contraction, nonalcoholic fatty liver disease, and the neurodegenerative diseases Alzheimer’s, Parkinson’s, and Huntington’s.

Tmsb4x and Tmsb10 are examples of other highly expressed genes enriched in samples from intact but not GDX mice. These actin-binding proteins are important in central nervous system development and neuroprotection (69) and are regulated by steroids (70, 71). In addition to their intracellular role regulating actin, secreted thymosins play roles in wound healing, angiogenesis, migration, and recovery from neurologic injury (72). Thymosinß4 stimulates GnRH release from rat hypothalamic explants (73). In endothelial cells, thymosinß4 interacts with cell-surface ATP synthase, increasing extracellular ATP, which can activate P2X4 purinergic receptors (74). Cell-surface ATP synthase is also found in neurons (75, 76), and the P2X4 purinergic receptor is expressed by cultured embryonic monkey GnRH neurons, where its activation alters calcium oscillations (77). P2x4 is enriched 1.5-fold in GnRH neurons from intact mice. Kiss1r, one of the most critical components of the neuroendocrine regulation of puberty and fertility (78), is also enriched only in GnRH neurons from intact mice. This suggests its regulation by gonadal factors but also points to its indirect effect through the regulation of GnRH neuron afferents (79) and its nonreproductive central roles (80, 81).

TRAPseq, as it profiles transcripts associated with ribosomes, reveals the transcripts in which GnRH neurons invest in converting to protein. This includes both highly expressed and highly enriched genes. Gnrh1 is the most highly expressed transcript in these cells and the only transcript to be both highly expressed and highly enriched. Both Bex2—expressed in the olfactory epithelium (82) and other endocrine progenitor cells (83)—and the deubiquitination gene Uchl1—largely limited to neurons and neuroendocrine cells (84)—are highly expressed in GnRH neurons from both intact and GDX mice. In contrast, Tmsb4x and Tmsb10, discussed previously, are enriched in samples from intact but not GDX mice.

Many transcripts identified as most highly enriched in the GnRH neuron translatome followed a pattern of low expression in the GnRH-enhanced RNA fraction and very low expression in the depleted fraction. An example is Rab25, the most highly enriched gene identified in the GnRH neuron (∼55-fold). Rab25 is a member of the Ras superfamily of small GTPases that has roles in cell migration, vesicular transport, and/or receptor recycling (85). It is admittedly an unusual candidate for a GnRH neuron marker gene. Expression of Rab25 in the mouse brain is extremely low, according to the Allen Brain Atlas (http://www.brain-map.org/). Rab25 has most often been reported as an oncogene in epithelial cells, where it is important in cell migration (metastasis) and apical recycling (86). Examination of the Gene Expression Omnibus database (GSE87544), from a recent report profiling gene expression of single dissociated mouse hypothalamic cells by drop-Seq (87), revealed that Rab25 expression is extremely rare, detected in only five neurons (TPM > 2; of ∼14,000 cells). Gnrh1 was also extremely rare, detected in only 29 cells (transcripts per kilobase million > 2), with only eight expressing the transcript at substantial levels (15 to 1500 TPM). Three neurons expressing Gnrh1 at high levels also expressed Rab25.

In examining fold enrichment or de-enrichment of transcripts in TRAPseq studies as described previously, it is critical to examine transcript abundance and to confirm rare transcripts with a secondary approach. We confirmed that eight of the 11 most highly enriched transcripts observed with TRAPseq, including Rab25, are indeed enriched when examined via qPCR of single POA TRAP samples. Some of these rare, but highly enriched transcripts may be useful to track these cells before GnRH expression is detectable in developmental studies.

Pathway analysis revealed that major synaptic signaling pathways for GABA, glutamate, dopamine, and retrograde endocannabinoids were largely de-enriched in GnRH neurons. The lone enriched gene in several of these pathways was Gng8. The enrichment of Gng8, like Bex2 and Gsn, may be related to the origins of the GnRH neuron in the olfactory placode. Gng8 is highly expressed in developing olfactory neurons (88), and mice null for Gng8 have a loss of vomeronasal neurons, although they are fertile (89). The relative de-enrichment of genes in these synaptic transmission pathways does not mean they are unimportant. GABA, endocannabinoids, and dopamine have been reported to alter GnRH neuronal activity (90–92), and GABAergic and glutamatergic postsynaptic currents have been detected in these cells (79, 93). Relative de-enrichment of RNAs encoding proteins for these neurotransmitter systems is consistent with anatomical (94–96) and electrophysiological (97, 98) studies indicating that GnRH neurons receive few synapses compared with other neuronal types in the brain. The observations of de-enrichment do not suggest that these transmitter systems are unimportant for regulation of GnRH neurons or their afferents in the case of endocannabinoids. Rather, they point again to an investment decision by GnRH neurons; the production of large amount of transmitter receptor system-related genes in a cell with fewer anatomical connections may be wasteful.

TRAPseq also confirmed the rarity of sex steroid receptors in GnRH neurons. It is well established that GnRH neurons are sensitive to steroid feedback, but with the exception of estrogen receptor β (Esr2), the classical steroid receptors have not been consistently detected in GnRH neurons using either histochemical or in situ approaches (53). Both the enhanced and depleted RNA fractions had low expression of sex steroid receptors, and many of these were de-enriched in the GnRH-enhanced fraction. This includes Esr2, which was de-enriched in the GnRH neurons from intact mice. Esr2 has been identified in GnRH neurons of rodents, humans, and sheep (54–56). It is important to reiterate that de-enrichment does not indicate a lack of expression but rather, a reduction compared with the nontargeted cell types.

As with all methods, TRAPseq has caveats, some of which have been mentioned previously but will be summarized here. First, as with other uses of TRAP, we enriched, rather than purified, these cells (33). Marker genes from nontargeted cell types, such as Gfap for glia, should ideally be de-enriched in the enhanced RNA fraction, but background is often detected. There may be multiple sources of this background. A technical source may result from association of non–GFP-labeled polysomes with antibody-coated beads; we identified three such cases for Bahcc1, Tnrc18, and Rai1, which were enriched in the bead-bound RNAs from GnRH-cre-only mice that lacked the GFP tag. There may also be biological sources of this background, such as shuttling of ribosomes or messenger RNAs among cell types (33) or low-level off-target expression of the transgene (enhanced GFP-L10a). Approximately 10% of GFP-labeled cells in the preoptic-area dissection did not coexpress GnRH, consistent with previous reports with this GnRH-cre mouse line (36). This likely stems from transient expression of the GnRH promoter during development (99) and resultant lifetime expression of Cre. Second, whereas TRAPseq is robust at detecting large differences, smaller shifts may not rise to the level of significance without a large number of biological replicates. In this regard, we were able to detect differential expression of most genes in the oxidative phosphorylation pathway despite a relatively small magnitude change, suggesting that this experiment was sufficiently powered. Third, it is possible that changes attributed to the studied cell type as enriched may actually be attributable to a downregulation of that transcript in the depleted fraction as a result of treatment. In this regard, for the data highlighted in this discussion, changes in enrichment were driven largely by changes in the GnRH-enhanced RNA fraction. Finally, it is important to remember that de-enrichment does not mean not expressed.

This hypothesis-generating approach revealed several possible lines of investigation that can be pursued in future research. These include the role of different energy metabolism pathways, identification of GnRH-enriched genes that may be useful in tracking these cells before GnRH expression in developmental studies, and the possible role of gonadal factors in maintaining enrichment of many genes that contribute to the mature GnRH neuron phenotype.

Glossary

Abbreviations:

Actb

β-actin

ATP

adenosine triphosphate

cDNA

complementary DNA

FACS

fluorescence-activated cell sorting

FC

fold change

FPKM

fragments per kilobase of transcript per million mapped reads

GABA

γ-aminobutyric acid

GDX

gonadectomized

Gfap

glial fibrillary acidic protein

GFP

green fluorescent protein

Glut

glucose transporter

GnRH

gonadotropin-releasing hormone

Gsn

gelsolin

KEGG

Kyoto Encyclopedia of Genes and Genomes

LH

luteinizing hormone

NCBI

National Center for Biotechnology Information

ORC

orchidectomized

OVX

ovariectomized

PCR

polymerase chain reaction

POA

preoptic area

qPCR

quantitative polymerase chain reaction

RNAseq

RNA sequencing

SEM

standard error of the mean

TRAP

translating ribosome affinity purification

TRAPseq

translating ribosome affinity purification combined with RNA sequencing