Pituitary hormones are specifically expressed in trigeminal sensory neurons and contribute to pain responses in the trigeminal system

Trigeminal (TG), dorsal root (DRG), and nodose/jugular (NG/JG) ganglia each possess specialized and distinct functions. We used RNA sequencing of two-cycle sorted Pirt-positive neurons to identify genes exclusively expressing in L3–L5 DRG, T10-L1 DRG, NG/JG, and TG mouse ganglion neurons. Transcription factor Phox2b and Efcab6 are specifically expressed in NG/JG while Hoxa7 is exclusively present in both T10-L1 and L3–L5 DRG neurons. Cyp2f2, Krt18, and Ptgds, along with pituitary hormone prolactin (Prl), growth hormone (Gh), and proopiomelanocortin (Pomc) encoding genes are almost exclusively in TG neurons. Immunohistochemistry confirmed selective expression of these hormones in TG neurons and dural nerves; and showed GH expression in subsets of TRPV1+ and CGRP+ TG neurons. We next examined GH roles in hypersensitivity in the spinal versus trigeminal systems. Exogenous GH produced mechanical hypersensitivity when injected intrathecally, but not intraplantarly. GH-induced thermal hypersensitivity was not detected in the spinal system. GH dose-dependently generated orofacial and headache-like periorbital mechanical hypersensitivity after administration into masseter muscle and dura, respectively. Periorbital mechanical hypersensitivity was reversed by a GH receptor antagonist, pegvisomant. Overall, pituitary hormone genes are selective for TG versus other ganglia somatotypes; and GH has distinctive functional significance in the trigeminal versus spinal systems.


Results
To isolate sensory neurons, we used Pirt/TdTomato reporter mice 17 and captured small-to-large-sized (10-80 μm) sensory neurons using a 100 μm nozzle. We first gated singlets from doublets (Fig. 1A). Live cells separated from all singlets were used to gate Pirt/TdTomato + cells (Fig. 1A). Medium-to-strongly expressing Pirt/TdTomato + neurons, were gated against TdTomatowild-type neurons (Fig. 1B). Omitting low expressing Pirt/TdTomato + neurons from gating is a critical step since it increases enrichment levels for Pirt + neurons. Additionally, two cycles of fluorescence-activated cell sorting (FACS) insured maximal purity (> 90%) of sensory neuronal fractions in samples 13 (Fig. 1C). To validate size distributions of sorted Pirt + neurons, two-cycle FACS-sorted cells were plated on coverslip and their size was assessed using NIS-elements (Nikon Instruments, Melville, NY) (Fig. 1D). TG Pirt + cell size distribution shows that described two-cycle sorting procedure captures small as well as large sensory neurons (Fig. 1E). Proportions of Pirt/TdTomato + cells to live cells were found to vary from ganglion-toganglion and between samples (Fig. 1F). TG samples had the greatest cell number with NG/JG samples having the least. However, we used approximately similar numbers of cells for RNA-seq experiments (Fig. 1G).
Differentially expressing genes (DEG)s for L3-L5 DRG versus T10-L1 DRG sensory neurons. Single-cell sequencing showed substantial differences in the subset of sensory neuronal groups innervating leg and paws (L3-L5 DRG) compared to colon and intestine (T10-L1 DRG) in male mice 11,12 . Accordingly, we compared transcriptomics of sorted sensory neurons from these ganglia. Using strengthened selection criteria outlined in the Materials and Methods section, no DEGs were revealed. However, lowering strength of the selection criteria using fold change (FC) > 5 and P value < 0.05 showed that compared to T10-L1 DRG versus L3-L5 DRG had 59 DEGs at RPKM > 5 and 28 DEGs at RPKM > 10. DEG numbers of T10-L1 relative to L3-L5 DRG with the same selection criteria were 58 at RPKM > 5 and 36 at RPKM > 10. Notable DEGs are highlighted in Table 1. Thus, nervous system related genes such as Ntsr2, Th, Trpv1, Accn1, Kcnh6, Cacna1b, and Gabrb3 were enriched in L3-L5 DRG, while immune system related genes Il6, Ccr1, Cxcl10, Nfkbie, Icam1, and Cd248 are mainly expressed in T10-L1 (Table 1). Consequently, gene clustering according to statistical overrepresentation test for biological processes using the PANTHER software assigned 9 predominant DEGs from L3-L5 DRG to the regulation of membrane potential. Biological processes assigned for T10-L1 DRG DEGs were involved in the regulation of MAPK cascade (10 DEGs) and cellular response to organic substance (15 DEGs). Importantly, many of these DEGs, including Ccr1, Cxcl10, Pdgfc, Nfil3, Irgm1, Il6, and Icam1 are linked to immune processes.

DEGs for NG/JG versus L3-L5 DRG and T10-L1 DRG.
Nodose-jugular ganglion complex has a specialized role in regulation of several vital visceral organs such as heart, lung, trachea, esophagus, and intestine 16 . Comparison of NG/JG sensory neuronal transcriptomic profiles to DRG revealed only several NG/JG-selective DEGs using outlined selection criteria (see "Materials and Methods") ( Table 2). DRG sensory neurons contain much more predominant DEGs compared to NG/JG. Thus, T10-L1 DRG sensory neurons have 113 DEGs at RPKM > 5 compared to NG/JG; L3-L5 DRG sensory neurons contain 99 such DEGs with 50 overlapping (Fig. 2A). These DRG-selective DEGs relative to NG/JG sensory neurons cannot be broken onto biological processes using statistical overrepresentation test. Nevertheless, DRG has several notable selective DEGs compared to NG/JG complex sensory neurons, including Mrgprd and Hoxa7 (Table 3).

DEGs with strong specificity for DRG, NG/JG and TG sensory neurons. Using pair comparison
and Venn diagram analysis we found Phox2b (paired-like homeobox 2b) and Efcab6 (EF-hand calcium binding domain 6) as DEGs that were strongly specific to male mouse NG/JG sensory neurons with little-to-no expression in DRG and TG ( Fig. 2B; Table 2). Phox2b is a transcription factor specifically expressed in neurons of the peripheral and central nervous system 18 (Table 3), an established marker of IB4 + DRG neurons 11,21 . A marker of proprioceptors, Pvalb, is also absent in NG/JG neurons, but is highly expressed in TG neurons, despite the fact that TG do not have proprioreceptors 1,11,22 . Another example is calcitonin-related polypeptide beta, Calcb, which is at   (Table 3). Hox genes play critical roles in development of many cell types, especially a subset of neurons, during embryogenesis 23 . We found Hoxa7, Hoxa9, and Hoxa10 were selectively expressed in L3-L5 DRG compared to NG/JG or TG ( Fig. 2C; Table 3). Comparison of DRG sensory neuron selective expression relatively to only TG sensory neurons found Hoxb2, Hoxb5, and Hoxb7 as L3-L5 DRG sensory neuron-specific DEGs. Evaluation of T10-L1 DRG sensory neuronal transcriptomic profiles relatively to NG/JG or TG outlined 4 DEGs, including Hoxa7 (Fig. 2D). Hoxb2 and Hoxb7, but not Hoxb5 are also specific for T10-L1 DRG sensory neurons compared to TG neurons. Interestingly most Hox genes, except for Hoxa7, did not differentially express in T10-L1 DRG when compared to NG/JG sensory neurons. Overall, our data show that Hoxa7 was the only DEG distinctively expressed in both T10-L1 and L3-L5 DRG, but not NG/JG or TG. We did not find Hox genes that lack transcription in DRG neurons.
Venn analysis of DEGs showed that 6 genes had significantly higher presence in TG compared to L3-L5 DRG, T10-L1 DRG as well as NG/JG sensory neurons (Fig. 3A,B). Cyp2f2 gene product is critical in the metabolism and toxicity of numerous xenobiotic compounds 24 . Ker18 plays a role in intestinal pathology 25 and is linked to  www.nature.com/scientificreports/ peripherin, a well-known marker for small-diameter sensory neurons 26 , located in chromosome 12 27 . Ptgds is a key enzyme in prostaglandin synthesis and specifically translated in female lumbar DRG neurons 28 . Accordingly, the PTGDS inhibitor, AT-56 produces hypersensitivity in male but is only effective at high doses in female mice 28 . Prl, Gh, and Pomc genes encode classical master-hormones, which are highly expressed in the pituitary 29,30 . PRL contribution in sex-dependent pain has been proposed in several studies [31][32][33][34] . These studies have mainly focused on expression and function of PRL receptor (Prlr) in DRG or TG neurons [35][36][37][38] . Exogenous GH plays an anti-nociceptive role in the spinal system 39 . POMC, which undergoes post-translational processing into multiple peptides including alpha, beta and gamma melanocyte-stimulating hormones (MSH), and adrenocorticotropin (ACTH), is also involved in anti-nociception in the spinal system due to opioids processed from POMC 40 . Further analysis of RNA-seq data showed that several DEGs are in DRG or NG/JG, but not TG sensory neurons. These genes are Map3k12, Slc35c1, Slc35b4, Ranbp6, Rab9b, Rapgef5, Tspan12, Ggta1, and Coro1c. Involvement of these genes in nociceptive pathway is unknown.  . According to these results, approximately 74% of sensory neurons were GH + / trpV1and 45% GH + TG neurons did not express CGRP. Overall, these data indicate that GH is expressed in both nociceptive and non-nociceptive TG sensory neurons.

Expression of Prl, Gh and
To further investigate GH and POMC expression, we performed whole mount IHC on male mouse dura biopsies as it was previously done for PRL 38 . GH is present in a set of neurofilament heavy chain positive (NFH + ) dural fibers ( . We note that NFH + fibers (i.e. A-fibers) usually travel with C-fiber inside of perineural sheath in dura 42 . Hence, Fig. 6 cannot definitively tell whether GH and POMC are expressed in C-and/or A-fiber containing TG neurons. Altogether, these data suggest a surprising expression of classical pituitary hormones POMC, GH, and PRL in TG, but not DRG or NG/JG sensory neurons.
Exogenous GH-induced hypersensitivity in the spinal system. Sex-dependent actions of exogenous and endogenous PRL in the spinal and trigeminal system are reported 31,34,38 . Growth hormone receptor 43 is expressed on DRG or TG neurons 11 (see also RNA-seq Supplementary data). IHC data (Figs. 4,5,6) indicate that along with their endocrine effects, endogenous GH could exert autocrine or paracrine actions upon release from non-neuronal extra-pituitary cells at periphery (i.e. hindpaws, dura, masseter muscle, etc.), spinal cord, brain stem or TG neurons. Here, we evaluated whether exogenously delivered GH can produce hypersensitivity in the spinal and trigeminal systems.
Exogenous GH-induced hypersensitivity in the trigeminal system. Single administration of exogenous GH into the masseter muscle produced up to 4-days-lasting orofacial mechanical hypersensitivity in a

Discussion
The various sensory ganglia have distinct anatomical locations and unique functions as well as discrete pain pathological conditions associated with their sensory neurons [1][2][3][4] . Therefore, characterization of sensory neuron transcription and protein profiles in different ganglia are critically important for understanding underlying mechanisms of pain conditions associated with these sensory ganglia. There are multiple published studies delineating the differences in transcriptional and translational profiles between lumbar DRG and TG in rodents and humans 14,15,41,[46][47][48] . In our and others estimation, sensory ganglia are composed of 90-95% non-neuronal cells (Fig. 1C) 13,14 . Hence, to truly delineate transcriptomic profiles for sensory neurons, there is a need for vigorous and meticulous purification of sensory neuronal fractions from sensory ganglion preparations. Meticulous sensory neuronal purifications by FACS or sensory neuronal ribosome isolation have been performed in some [13][14][15]41 , but not all studies [46][47][48] . Sensory neuronal fraction purification by FACS requires the use of reporter mice 14 or back-labeling of sensory neurons with fluorescent tracers 13 . Reporter mice must possess high specificity in all (or almost all) sensory neurons. Specificity of the advillin promotor (Avil-GFP) 14 has been disputed 49 . Accordingly, we selected Pirt-cre/TdTomato reporter, specificity of which has been confirmed by two independent differential   21,50 and detailed anatomical studies 17,51 . Another challenge in purification of sensory neurons is created by their large size (15-70 μm). To avoid missing large neurons and damaging mediumto-large neurons, 100-130 μm nozzle is required during FACS 13 (Fig. 1D,E). However, having a large nozzle, in turn, undermines FACS efficiency. Thus, a minimum of two cycles of sorting (used here) is required (Fig. 1C). Moreover, to achieve high enrichment, weakly Pirt-expressing neurons should be omitted by gating (Fig. 1B,C).
Next, it appears that ganglion non-neuronal cells have tendency to create doublets with sensory neurons 41 . Separation of doublets from singlets during FACS is paramount for accurate purification of sensory neurons (Fig. 1A). Finally, preparation from wild-type mouse sensory ganglia is necessary for correct setting of gates (Fig. 1B). All these precautions are important for accurate and highly enriched sensory neuron-specific purifications.
We have detected differences in gene expression between sensory neurons of different ganglia by using RNAsequencing on population of Pirt + neurons. Moreover, DEGs selection for further validation, and consideration for functional studies were quite strong: > 5 RPKM, fold-change (FC) > 5 and statistical significance for DEGs as Padj < 0.05. Populational based RNA-sequencing has strong and weak sites. Thus, this approach has strong signal to noise ratio, as well as reproducibility and reliability of results. Drawback of this approach is that for heterogeneous population as Pirt + neurons from different ganglia, it is impossible to tell whether a variation in transcript reads between distinct ganglia could be due to a differential proportion of gene-expressing neurons, or a differential level of transcript production in the gene-expressing neurons. Such questions could be answered by performing single-cell sequencing and analysis of sensory neurons from different ganglia. However, this approach has its own limitations such as difficulty in performing high quality single-cell sequencing on sensory neurons and especially single cell sequencing analysis for differences between two distinct groups of sensory neurons. In any case, transcriptomic data on differences between populations of sensory neurons generated by these approaches require further lengthy functional studies to prove meaningfulness of findings.
Besides traditional targets-L3-L5 DRG and TG, T10-L1 DRG and NG/JG were used as described experiments, as this allows for a broader picture on diversity for sensory neuronal transcriptomic profiles. Additionally, T10-L1 DRG and NG/JG are functionally distinct from L3-L5 DRG or TG neurons [1][2][3][4] . There is a set of DEGs, which were predominantly expressed in L3-L5, T10-L1 DRG, NG/JG or TG sensory neurons (Tables 1,2,3, Figs. 2, 3). These DEGs do not cover any particular biological processes according to the statistical      Here, we have focused our efforts to investigate whether GH contributes to regulation of nociception in the spinal versus trigeminal system. GH did not exert acute heat and mechanical hypersensitivity in male mouse hindpaw after local administration (Figs. 7A,B). However, intra-spinal (i.e. intrathecal) injection of exogenous GH (1 or 5 μg) produced up to 4-day lasting mechanical, but not heat hypersensitivity (Figs. 7C,D). Interestingly, no dose-dependency of GH action was recorded. This could indicate that lower GH dosage is able to produce hypersensitivity via acting on central terminals of sensory neurons. Another possibility is that GH activates spinal cord cells, which in turn release factors activating or sensitizing the central terminals in the spinal system. Unlike the spinal system, stimulation of peripheral terminal in the trigeminal system by GH produced profound acute hypersensitivity. Thus, orofacial mechanical hypersensitivity was induced by after single administration of GH into masseter muscle and periorbital mechanical hypersensitivity was detected after injection of GH into dura mater of male mice (Figs. 8A,B). The effect of GH in the trigeminal system was dose-dependent (Fig. 8A). This GH-induced hypersensitivity was meditated by GHR and reversed by pegvisomant (Fig. 8C). These data indicate www.nature.com/scientificreports/ that ability of GH in induction of hypersensitivity in male mice depends on several factors such as modality (thermal vs. mechanical), sensory system (spinal vs. trigeminal), and application site (peripheral vs. central terminals). Previous study shows that systemic application of GH (0.5 mg/kg) increased baseline mechanical nociception in P7, but not P14 male mice 39 . In contrast, heat baseline nociception was reduced in P7, but not P14 male mice 39 . These results agree with our data (Fig. 7A,B). Interestingly, multiple systemic GH treatments attenuated carrageenan acute inflammatory hypersensitivity 39 . GH release hormone receptor (GHRHR) ablation induced behavioral and afferent hypersensitivity during early developmental stages but resolved at P21 age male mice 52,53 . Reported effects of GH in young male mice are attributed to insulin-like growth factor 1 receptor 39 . It is not clear whether the action of GH is local, in the spinal cord or brain. However, systemically delivered GH is not readily cross the blood brain barrier 54 . Moreover, effect of GH in the spinal system has yet to be assessed in females, especially female mice at the reproduction age. Our data and previous literature suggest that GH could have differential effects on modulation of nociception and hypersensitivity in the trigeminal versus spinal systems. This is also supported by clinical data. Thus, acromegaly patients having excess GH often reported severe and prolonged headaches, but no pain in limbs (see review 55 ).
An increased ACTH brings to Cushing and Addison disease. Patients with Cushing's syndrome and Addison disease seldom report abdominal pain 56,57 . However, pain during these diseases was not associated with elevation of ACTH. Despite PRL, GH, and POMC-derived peptides/proteins are predominantly expressed in the pituitary at high levels, there is evidence for extra-pituitary presence of these hormones, especially in immune cells [58][59][60][61] . POMC-derived peptide could be involved in cell-cell communication via autocrine and paracrine mechanisms. Moreover, reduction of already low levels of POMC expression in DRG neurons of female and male mice with diabetic neuropathy contributes to hypersensitivity 40 . This effect was attributed to endorphins that could be processed from POMC 40 . Overexpression of POMC-derived endo-opioids in L3-L4 DRG does not change baseline nociception in female mice but suppresses diabetic neuropathy-induced hypersensitivity 40 . Interestingly, diabetic female mice develop heat hypersensitivity, as opposed to hyposensitivity in males, while POMC-MOR expression is downregulated in both female and male mice, and POMC effects in diabetic mice was not sex-dependent 40 . Like GH, POMC-derived peptides could have distinct signaling pathways and outcomes on modulation of nociception and hypersensitivity in the trigeminal versus spinal systems.
Overall, our approaches identify several DEGs that are specifically expressed in sensory neurons of DRG, NG/JG or TG. Interestingly, these genes include very prominent players in the endocrine system-PRL, GH, POMC as well as prolactin and oxytocin receptors. Moreover, Prl, Prlr, Pomc and Ptgds, which are involved in differential occurrences of pain disorders in men and women 28,31,34,40,62 , have differential expression for TG versus DRG and NG/JG neurons. Based on our findings, we favor the hypothesis that certain critical proteins for the endocrine system have different signaling pathways as well as pathophysiological outcomes for the spinal versus trigeminal system. Our results advance our understanding of unique properties of sensory ganglion neurons and provide a building step for further studies on regulation of nociceptive pathways by endogenous GH and POMC for pathological pain conditions affecting head and neck area (i.e. the trigeminal system).

Materials and methods
Mouse lines and reagents. All animal experiments conformed to APS's Guiding Principles in the Care and Use of Vertebrate Animals in Research and Training, and to protocols approved by the University Texas Health Science Center at San Antonio (UTHSCSA) Animal Care and Use Committee (IACUC). We followed guidelines issued by the National Institutes of Health (NIH) and the Society for Neuroscience (SfN) to minimize the number of animals used and their suffering. The reporting in the manuscript follows the recommendations in the ARRIVE guidelines.

Isolation of ganglion sensory neurons.
Left and right whole L3-L5 DRG, T10-L1 DRG, NG/JG complex and TG tissue biopsies were dissected after perfusion of Pirt/TdTomato mice with phosphate buffer pH 7.3 (PB). Ganglion tissues were used for single-cell suspension generation as previously described 13  RNA isolation, cDNA synthesis and RNA-sequencing. RNA was isolated from single-cell sensory neuron suspension using Qiagen RNeasy (Universal Mini Kit) as was previously described 74 . RNA (< 10 ng) quality was accessed after cDNA preparation using Fragment Analyzer Agilent 2100 Bioanalyer RNA 6000 Nano chip (Agilent Technologies, Santa Clara, CA). RNA-seq cDNA libraries from sensory neuronal fraction (3000-35,000 neurons depending on types of ganglia) were prepared using oligo dT according to SMART-seq-2 protocol 75,76 with previously described modifications 13 . cDNA libraries were subjected to quantification and subsequent 50 bp single read sequencing run with Illumina HiSeq 3000 platform (Illumina, San Diego, CA). Each group have n = 4 samples. Depth of reads was 30-50 × 10 6 bp for each sample.
Transcriptomic data analyses and statistics. Sequencing data from all samples were processed in the same way as previously described 13 . Briefly, RNA-seq readings were de-multiplexed with CASAVA and the FastQ files were generated. Raw reads were aligned to mouse genome build mm9/UCSC hg19 using TopHat2 default settings 77,78 . The BAM files obtained after alignment were processed using HTSeq-count to obtain the counts per gene, and then converted to RPKM (Read Per Kilobase of gene length per Million reads of the library) 79 . Differentially expressing genes (DEGs) were identify using DESeq software after performing median normalization 80 .
Quality control statistical analysis of outliers, intergroup variability, distribution levels, PCA and hierarchical clustering analysis was performed to statistically validate the experimental data. Multiple test controlling was performed with Benjamini-Hochberg procedure and adjusted p value (Padj) was generated. Criteria for selection of DEGs for the further analysis are following: > 5 RPKM, fold-change (FC) > 5 and statistical significance for DEGs as Padj < 0.05. This allows to select DEGs with high levels expression and significant difference in expression levels. DEGs were clustered according to biological processes using the PANTHER software (http:// www. panth erdb. org/).
Immunohistochemistry. Immunohistochemistry (IHC) was performed on L3-L5 DRG, NG/JG complex and TG sections, and dura mater biopsies dissected from naïve male 4% paraformaldehyde-perfused mice. Cryo-section (25 μm) generation and IHC process were performed as described 22,36 . Whole-mount IHC on dura biopsies was carried out. Intact dura was fixed again with 4% paraformaldehyde and cryoprotected with 30% sucrose in phosphate buffer. Labeling with primary and secondary antibodies were done on submerged dura samples in wells of 12-well plates. IHC was simultaneously performed on 4-8 sections generated from 3 animals. The following primary antibodies were used: anti-CGRP rabbit polyclonal (Sigma; C8198; 1:300) 64 Control IHC was performed on tissue sections processed as described but either lacking primary antibodies or lacking primary and secondary antibodies. Images were acquired using a Keyence BZ-X810 All-in-One Fluorescent Microscope (Keyence, Itasca, IL), a Nikon Eclipse 90i microscope (Nikon Instruments, Melville, NY) equipped with a C1si laser scanning confocal imaging system or Zeiss (Carl Zeiss, Jena) LSM single photon confocal microscope. Images were processed with NIS-elements (Nikon Instruments, Melville, NY), ZEN (Carl Zeiss, Jena) or Adobe Photoshop CS2 software. Gain setting was constant during acquisition, and it was established on no primary control slides. Cell counts from IHC images acquired as Z-stack were performed using Image J software. Total cells/section and positive cells were counted. Cell counting were performed independently by two investigators. We used 3 independent mice to generate sections and counted 3-5 sections per mouse. Thus, each group has n = 3, and data for each sample are represented by mean values from 3 to 5 sections generated per animal.

Dural, masseter muscle, hindpaw (intraplantar) and spinal cord (intrathecal) injections.
Injection into hindpaw (i.e. intraplantar) was done as previously described 69 . Briefly, mice were anesthetized with 5% isoflurane (v/v) for ≈20-30 s. The plantar surface of the footpad was cleaned with betadine and 70% ethanol. Solutions with hormones (10 μl) were injected using 1 ml-insuline syringe with 30-gauge needles into the metatarsal region of the hindpaw. Pressure on hindpaw was maintained several seconds after withdrawal of the needle.
Injection into spinal cord (i.e. intrathecal) was done as described 34 . Briefly, tissue above spinal L3-L5 levels was cleaned with betadine and 70% ethanol. The mice were anesthetized with 5% isoflurane (v/v) for ≈1-1.5 min. Injection were performed with 30-gauge 1/2-inch needle mated to a 10-pl luer tip syringe (Hamilton, Reno, NV). The needle is inserted into the tissue above L4 or L5 spinal levels so that it slips into the groove between the spinous and transverse processes. The needle is then moved carefully forward to the intervertebral space as the angle of the syringe is decreased to about 10°. The tip of the needle is inserted so that approximately 0.5 cm is within the vertebral column. Solutions with hormones (10 μl) were injected and the needle rotated on withdrawal.
Injection into masseter muscle were also performed on mice anesthetized with 5% isoflurane (v/v) for 1-1.5 min. The skin over masseter muscle was cleaned with betadine and 70% ethanol. Solutions with hormones (10 μl) were injected using 1 ml-insuline syringe with 30-gauge needles into region closer to tendinous aponeurosis of the superficial head of the masseter muscle.  70 . Briefly, mice were anesthetized under isoflurane for < 2 min and injected with 5 μl hormone solution using a modified internal cannula (Invivo1, part #8IC313ISPCXC, Internal Cannula, standard, 28 gauge, fit to 0.5 mm). The inner portion of the cannula was adjusted with calipers to extend from 0.5 to 0.65 mm in length. The cannulas were inserted through the soft tissue at the intersection of the lambdoidal and sagittal sutures.
Hypersensitivity testing of the hindpaws, periorbital skin and V2 facial skin area over masseter muscle. All experimenters performing testing on mice and data analysis were done blinded for all behavior experiments. Allocation of animals to treatment groups was randomized by a "blinder" that drew animal numbers from a bag of paper slips. Mice were habituated to the testing environment for at least 1 h prior to nociception measurements on the hindpaw. Heat-induced nociception was measured using an automated Hargreaves' apparatus as previously described 71 . Mechanical stimulus-induced nociception following the intraplantar injection was assessed by determining paw withdrawal threshold using up-down von Frey method 72 .
Mechanical hypersensitivity in head and neck area was performed on unrestrained animals 70,73 . To reduce any effects of restraint by hands or grasping the tail, mice were habituated 73 . For measurements of mechanical hypersensitivity following masseter muscle injection, sequence of following procedures was carried out. First, naïve mice were placed in a black wire mesh box (4 × 4 × 4″) and allowed to freely move for ≈1 h. This procedure was repeated 3 days. Next, test von Frey filament probing of V2 facial skin area over masseter muscle was applied to each mouse for 3-4 consecutive days. Each grade of von Frey filament was applied 3 times at intervals of a few seconds. The stimulation always began with the filament producing the lowest force and stopped when mice are responded to 3 consecutive stimulations with a graded von Frey filament. A brisk or active withdrawal of the head from the probing filament was defined as a response. Mice with mechanical threshold > 0.6 g at V2 facial skin area, which is considered baseline mechanical nociception, were selected for drug/hormone injection. After injection into masseter muscle, mechanical hypersensitivity was regularly assessed for 1 week.
Headache-like behavior following dura injection were assessed as described 70 . To habituate mice, they were placed in 4 oz paper cups (Choice) for 2 h a day for 3 consecutive days. von Frey testing of the periorbital skin (the midline of the forehead at the level of the eyes), which is used to assess headache-like behavior, was carried out on each mouse located in a paper cup for the 3-4 consecutive post-habituation days. Baselined animals were defined as animals that exhibited a withdrawal threshold > 0.6 g. Mice with a baseline threshold < 0.6 g at the end of 3 habituation days and 4 test days were excluded from experiments. After application of drug/hormone to dura, mechanical thresholds were regularly determined for 1 week by applying von Frey filaments to the periorbital skin in an ascending/descending manner starting from the 0.02 g filament. If the animal responded to this filament, decreasing forces were applied until the 0.008 g filament was reached.
Statistical analysis. GraphPad Prism 9.0 (GraphPad, La Jolla, CA) was used for statistical analyses. Data in the figures are mean ± standard error of the mean (SEM), with "n" referring to the number of analyzed mice for IHC or behavioral experiments. Statistical changes between 2 or more groups with two variables were analyzed by regular 2-way ANOVA with Sidak's post-hoc tests. A difference is accepted as statistically significant when p < 0.05. Number of animals in a group, interaction F ratios, and the associated p values are reported.