Differential retinoic acid responses across testicular development in vitro

Testis, Retinoid, Reproduction, Development, Spermatogonia, in vitro, cell culture. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 2

Spermatogonial differentiation is controlled by distinct spatial and temporal activities of metabolizing enzymes and signaling molecule formation. In vitro models are reductionist ways to examine the interactions between Sertoli cells and spermatogonial stem cells, which may be therapeutic targets for infertility that cause non-obstructive azoospermia, due to either Sertoli-cell only syndrome, hypospermatogenesis or maturation arrest. Here, we find that nanomolar doses of isotretinoin are sufficient to drive Stra8 expression in vitro, an interaction that is both dose-dependent and inhibitable. We compare complex in vitro models (CIVMs) seeded from cells isolated post-natal day 5 (PND 5) and post-natal day 10 (PND 10) Sprague Dawley rat testis. The CIVMs maintain metabolic capacity to produce bioactive retinoids form retinol. We also investigate the impact of common media supplementations on spermatogonial phenotype and find that they can impact the expression of Stra8 and Plzf as markers of early differentiating and undifferentiated spermatogonia, respectively. These results highlight the power of in vitro models to investigate the dynamics of the spermatogonial niche. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 3

1 The testis remains an elusive target for new complex in vitro models (CIVMs), despite 2 9% of males experiencing infertility in their lifetime globally and difficulty in identifying 3 testicular impacts during toxicology screening and drug development [1–4]. A 4 considerable proportion (~40%) of infertility cases are idiopathic [5]. These cases 5 require increased research in the field and particular emphasis on critical developmental 6 windows during testicular development. Neonatal development of testis (post-natal days 7 3-7 in the rat) is critical for the establishment of the testis niche [6], where dysfunction of 8 neonatal testicular development is common with lasting adverse effects on reproductive 9 health and fertility [7–10]. 10 In vitro models of the testis need to consider the complexity of the organ where there 11 are two distinct functional areas: the seminiferous tubule, the site of spermatogenesis, 12 and the interstitial space, the site of testosterone biosynthesis and interaction with 13 vasculature. Sertoli cells are the supporting somatic cells of the seminiferous tubule, 14 responsible for maintaining spermatogenesis through close contact with spermatogonia 15 and establishment of the immune privilege area for post-meiotic cells [11]. Sertoli cells 16 continue to proliferate early in life, ceasing around post-natal day (PND) 15 in rats 17 [12,13]. Spermatogonia reside at the basal lamina of the seminiferous tubule in close 18 contact with the Sertoli cells, where they undergo signaling towards either proliferation 19 (renewal) or differentiation and commitment to meiosis. Retinoic acid (RA) is the well-20 established initiator of spermatogonial differentiation and most bioactive RA in testes 21 (all-trans RA; at-RA) is produced within the testis, where ALDH1A enzymes are present 22 within the Sertoli cells of the neonatal testis to form bioactive RA and CYP26 enzymes 23 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 4 are present within the peritubular myoid cells that encircle the tubules to form a barrier 24 to exogenous retinoic acid and the pre-mature initiation of spermatogenesis [14]. 25 Mechanistic studies on the interactions between Sertoli cells and spermatogonial stem 26 cells are critically needed to study idiopathic infertility conditions that cause non-27 obstructive azoospermia, due to either Sertoli-cell only syndrome (SCOS), 28 hypospermatogenesis, or maturation arrest. Both SCOS and maturation arrest are 29 significant sources of clinical infertility from non-obstructive azoospermia where there is 30 either an arrest of the spermatogonial cells in their pre-meiotic phenotype or an absence 31 of spermatogonial cells altogether (SCOS). In both SCOS and maturation arrest, 32 genetic and environmental factors are implicated, supporting the use of in vitro models 33 where gene x environment interactions are controllable [15–17]. 34 In this study, we compare CIVMs derived from post-natal day 5 (PND 5) and post-natal 35 day 10 (PND 10) Sprague Dawley rat testis. At PND 5, these are neonatal rat testis, 36 where Sertoli cells are actively proliferating and gonocytes are transitioning into type A 37 undifferentiated (type A) spermatogonia [6]. By PND 10, these are now early infantile rat 38 testis where both Sertoli cells and spermatogonia are actively proliferating, alongside 39 the emergence of early differentiating (type B spermatogonia) [6]. While both PND 5 40 and PND 10 Sprague Dawley rats are reproductively immature animals, the likely 41 emergence of type B in the PND 10 testis and the impact of this difference in 42 spermatogonial phenotype, is of particular interest in this study. 43 44

45 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 5 Tissue isolation and culture establishment 46 Testes were isolated from neonatal Sprague Dawley rats (Strain 001, Charles River) on 47 post-natal day 5 (PND 5) or post-natal day 10 (PND 10) using sterile forceps and placed 48 in Minimum Essential Media (MEM, Gibco) containing 1% penicillin-streptomycin (PS, 49 Gibco). Testes in MEM with 1% PS were kept on ice (< 30 minutes) and moved to a 50 sterile Biological Safety Cabinet (BSC) for the remaining tissue isolation. When moved 51 into the BSC, the testes were washed with fresh MEM twice by inversion in a conical 52 tube. The washed testes were then placed in a 6-well cell culture plate with 2-3 mLs of 53 cold MEM in each well. Using sterile forceps, the testes were placed in a well, the plate 54 swirled, and the testis moved to the next well, as a wash. After two washes, the plate 55 was moved under a dissecting microscope in the BSC, where the epididymides were 56 removed and each testis detunicated. Each detunicated testis was fragmented into 3-4 57 pieces and moved to a well with fresh MEM. Then, 2-3 fragments were moved to 58 cryovials with 1mL freezing medium (70 mmol/L sucrose diluted in phenol-red free 59 DMEM/F12 with 10% v/v DMSO; Gibco; freezing media recipe from [18]). The vials 60 were moved to a controlled freezing device (Mr. Frosty) and kept in a -80C freezing for 61 24 hours, after which the cryovials were moved to liquid nitrogen for long term storage. 62 For cell plating, the vials were removed from liquid nitrogen storage and rapidly thawed 63 in a water bath (37C). The fragments were removed from the cryovials into a conical 64 tube with fresh MEM using a transfer pipette to minimize freezing media crossover. The 65 fragments were allowed to settle on ice for 10 minutes before sequential enzyme 66 digestion using hyaluronidase and collagenase; additional enzymatic digestion protocol 67 details in published protocol [19]. The digested tissue was further dissociated with a 68 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 6 brief digestion in 0.05% Trypsin-EDTA solution (Gibco), followed by centrifugation to 69 resuspend in complete testis cell culture medium (containing phenol-red-free MEM 70 [Gibco], 1% sodium pyruvate [Gibco], 1% non-essential amino acids [Gibco], 1% ITS+ 71 culture supplement [Corning], 1% PS [Gibco], epidermal growth factor [Sigma Aldrich], 72 and sodium lactate [Sigma Aldrich]); complete media formulation available in published 73 protocol [19]. The cells were resuspended at a concentration of 1.6 million cells per mL 74 and 100uL plated in each well of a 48-well plate that had previously been coated with a 75 1% v/v Matrigel in MEM solution for > 1 hour at room temperature. Immediately after 76 adding the cell suspension to each well, an additional 100uL of Matrigel overlay solution 77 (complete testis medium with 400ug/mL Matrigel protein), was added to each well, for a 78 final plating density of 800,000 cells per mL and 200 ug/mL Matrigel overlay in a 200 uL 79 total culture volume. Culture media was changed 24-hours after plating, and every 48-80 hours thereafter. 81 Retinoid dosing and inhibitor treatment 82 Isotretinoin (13-cis retinoic acid) and Retinol (both from Sigma Aldrich) stocks were 83 prepared at 1 mg/mL in analytical grade ethanol and mainlined at -80C protected from 84 light. Retinoids were diluted to working stocks in analytical grade ethanol. The working 85 stocks were diluted in complete cell culture media to achieve the experimental doses 86 and consistent 0.1% ethanol v/v. WIN-18446 was used as a specific inhibitor of the 87 aldehyde dehydrogenase (ALDH) enzymes and BMS-189453 was used as a pan-88 retinoic acid receptor (RAR(α,β,γ)) antagonist. WIN-18446 (Sigma Aldrich) was diluted in 89 DMSO to a stock concentration of 1mM. BMS-189453 (MedChem Express) was also 90 diluted in DMSO to a stock concentration of 1mM. All control experiments contained an 91 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 7 equal volume of the organic component as the treatment groups (Ethanol or DMSO), in 92 the stimulus and inhibitory studies, the organic fractions were 0.1% v/v and 0.2% v/v, 93 respectively. 94 Culture media supplementation 95 Culture media supplementation experiments used three media supplement approaches 96 common for in vitro testis models: spermatogonial stem cell (SSC) growth factors [20], 97 lipid-rich albumin (AlbuMAX, Thermo Fisher Scientific) [21], and a defined bovine serum 98 (Knockout Serum Replacement) [22,23]. SSC growth factor media was prepared as 99 published [20]: using recombinant rat growth factors, 40 ng/mL GDNF (R&D 512-GF-100 010/CF), 1 ng/mL FGF2 (R&D 3339-FB-025/CF), and 300n/mL GFRA1 (R&D 560-GR-101 100/CF) in complete testis media. Lipid rich albumin media was prepared by adding 102 20mg/mL of AlbuMAX I Lipid-Rich BSA (Thermo Fisher Scientific) to complete testis 103 media and sterile filtering (0.2um filter), and defined bovine serum media was prepared 104 by adding 10% v/v Knockout Serum Replacement (Gibco) to complete testis media. 105 Control CIVMs were maintained in complete testis media. All cells were plated in control 106 media and changed to treatment media 24 hours after plating, then changed every 48 107 hours thereafter through 15 days in vitro. 108 RNA relative expression analyses and sequencing 109 Cells were lysed from the 48-well plates using 250uL RLT Lysis buffer (Qiagen) per well. 110 The lysis buffer was added to the well, mixed thoroughly, and moved to an RNAse free 111 microcentrifuge tube which was stored at -80C until extraction. 112 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 8 To perform reverse-transcription quantitative PCR (RT-qPCR), total RNA was extracted 113 using a RNeasy Mini Kit (Qiagen), following manufacturer protocols. The quality and 114 concentration of the extracted RNA was determined using a Nanodrop 115 spectrophotometer (ND-1000; Thermo Fisher Scientific). Reverse transcription of the 116 extracted RNA to cDNA was performed using an Applied Biosystems High-Capacity 117 cDNA Reverse Transcription Kit, following manufacturer protocols and using a C1000 118 Touch Thermal Cycler (Bio-Rad). The reverse transcribed cDNA samples were diluted 119 with nuclease-free water to 2.5 ng/μ L and used for quantitative polymerase chain 120 reaction using TaqMan assays and TaqMan Fast Advanced Master Mix in a QuantStudio 121 3 system (Applied Biosystems). Each RT-qPCR mixture consisted of 10 μ L of TaqMan 122 Fast Advanced Master Mix, 5 μ L of nuclease-free water, 1 μ L of TaqMan Gene 123 Expression Assay, and 4 μ L of cDNA (10ng of cDNA total per reaction). The RT-qPCR 124 thermocycler settings followed TaqMan protocols. RT-qPCR was used to target the rat 125 genes Stra8 (Rn01747849_m1) and Plzf (Rn01418644_m1). Each sample’s gene 126 expression was normalized using the difference in cycle threshold (Δ Ct) calculation for 127 relative quantification of gene expression to the housekeeping gene beta-actin (Actb; 128 TaqMan Rn00667869_m1). 129 To perform bulk RNA-sequencing, the cell lysate samples were sent to Novogene 130 America, where the RNA was isolated and messenger RNA was purified from total RNA 131 using poly-T oligo-attached magnetic beads. The purified mRNA was fragmented and 132 the first strand cDNA was synthesized using random hexamer primers, followed by 133 second strand cDNA synthesis using dTTP for nondirectional library, and sequenced 134 using the NovaSeq X Plus sequencing platform (Illumina). RNA quality was determine 135 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 9 using a bioanalyzer, with all RNA Integrity Numbers (RINs) above 9, indicating high 136 quality RNA for sequencing. 137 Transcriptomics 138 Sequencing reads were aligned to the rat genome, GRCr8 (ensembl), using STAR 139 aligner and gene counts tabulated using the STAR gene counts function (genecounts) 140 [24]. Aligned read counts were processed using edgeR, which included removal of lowly 141 expressed genes, normalization using the trimmed-mean of M-values method [25], and 142 comparisons made between PND 5 and PND 10 CIVMs using a Quasi-likelihood F-test 143 [26]. Statistically significant gene expression was determined based on a false 144 discovery rate (FDR) less than 0.05 and a log2 fold change of at least 0.5 in either 145 direction. 146 Pathway enrichment analyses were calculated using the TopGO package (v. 2.60.1; 147 [27]). This method calculated the overrepresentation of differentially genes, as 148 determined by edgeR analyses, annotated in pathways within the Biological Processes 149 family of the Gene Ontology database [28,29]. This analysis considers if a pathway 150 contains more genes that are differentially expressed than would be expected based on 151 the rate of differentially expressed genes in the dataset, the statistical significance is 152 calculated with a fishers exact test to identify how likely the observed differential gene 153 proportion would be if the genes were selected at random. 154 Transcription factor enrichment assessment was completed using the decoupleR 155 package (v. 2.9.7, [30]). The run_ulm function was used for univariate linear modeling 156 on the matrix of differentially expressed genes where a synthetic variable (t), was 157 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 10 created by multiplying the log2 fold change by the -log10(p-value) to incorporate 158 magnitude, directionally, and statistical significance into a single metric for modeling. 159 This follows the standard protocol with the run_ulm function, and a minimum of ten 160 genes were required to be annotated for each transcription factor. The transcription 161 factor and gene associations were downloaded from the omnipath database (rat specific 162 database #10116; [31,32]). The omnipath database data was not specific to any 163 developmental stage. The transcription factor activities with a p-value of less than 0.05 164 based on the fitted univariate linear model were considered. 165 166 Deconvolution 167 Deconvolution estimates of the cell type proportions were completed using the run-168 DWLS and run_OLS functions within the DWLS package (v. 0.1.0) to implement the 169 dampened weighted least squares and ordinary least squares regressions, respectively 170 [33]. These regression-based deconvolution methods require two inputs: a library-size 171 normalized matrix of gene counts from bulk RNA-seq samples and a signature matrix 172 where each cell type expected to be within the bulk RNA-seq sample is represented by 173 a gene expression values, these values are the product of a regression on a larger set 174 of cells annotated to that cell type from publicly available single cells dataset (see 175 supplemental table 2 for details on single-cell references used here). Four single cell 176

were used in these analyses: two from neonatal human samples and two 177 from neonatal mouse samples (supplemental table 2). The cell types were annotated 178 from the publicly available single-cell references to the following groupings: Sertoli cells, 179 Leydig/stroma cells, germ cells, and peritubular myoid cells. The germ cell fraction was 180 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 11 defined by general germ cell marker expression and was a combination of subtypes: 181 gonocytes, pro-spermatogonia, spermatogonial stem cells, pro-spermatogonia, as well 182 as possible early differentiating spermatogonia. To more accurately represent the gene 183 distributions present in a bulk RNA-seq dataset, single-cell gene expression imputation 184 implemented through the Adaptively-thresholded Low Rank Approximation (ALRA) 185 algorithm was applied to the single-cell references in two of the three methods used 186 here [34]. The three methods used in this report are: dampened weighted least squares 187 regression with an ALRA imputed reference dataset, dampened weighted least squares 188 regression with a non-imputed reference dataset, and ordinary least squares regression 189 with an ALRA imputed dataset. See [35] for discussion and benchmarking of 190 deconvolution approaches for neonatal tests bulk RNA-seq. The deconvolution 191 estimates are presented as proportions of the total cell type composition. To measure 192 significant differences in cell type composition between the two developmental stages, 193 PND 5 and PND 10, was assessed with an unpaired Wilcoxon ranked sum test of the 194 estimated proportions across methods and references. 195 196 Fluorescent microscopy 197 All fluorescent microscopy images presented as indicative of representative examples 198 using indirect immunocytochemistry (ICC) with commercially available antibodies. For 199 the ICC of the 48-well culture plates, the plates were fixed with 4% paraformaldehyde 200 solution (PFA) for 15 minutes before washing with D-PBS+/+ (Gibco). Plates were kept 201 at +4C until processing using indirect antibody binding. All antibody buffers were made 202 in D-PBS +/+. The plates were removed from the +4C fridge and warmed to RT. The 203 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 12 cells were permeabilized for 3 minutes using 0.1% Triton X-100 (Sigma Aldrich), 204 followed by non-specific binding blocking at RT in a solution of 1% goat serum (Vector 205 Labs) and 0.1% Tween-20 (Sigma Aldrich). After blocking, the cells were washed with 206 0.1% Tween-20, followed by addition of primary antibodies at a 1:100 v/v dilution in 1% 207 goat serum and 0.1% Tween-20. The cells were moved to +4C to incubate overnight. 208 The cells were washed three times with 0.1% tween, then moved to secondary antibody 209 incubation at a 1:1000 v/v dilution in 1% goat serum and 0.1% Tween-20 for one hour at 210 RT. After incubation, the cells were washed three times more with 0.1% tween, followed 211 by incubation with a nuclear counterstain (Hoechst 33342 , Thermo Fisher Scientific) for 212 10 minutes. The counterstained cells were washed with fresh D-PBS +/+ and stored in 213 the dark at +4C until imaging. 214 Whole mount tubule imaging was performed similarly to the ICC of the cell culture 215 plates with major modifications (tubule protocol adapted from [36]). The whole tubules 216 isolated from detunicated neonatal rat testis were fixed in 4% PFA for two hours at +4C. 217 After fixation, the tubules underwent heat activated antigen retrieval in a 10 mM sodium 218 citrate buffer with 0.05% v/v Tween-20 (adjusted to pH of 6), for 15 minutes at +95C in 219 microcentrifuge tubes using a block heater. The tubules were permeabilized twice, each 220 for 15 minutes. The primary antibody incubations were extended to three days at +4C 221 and the secondary antibody incubations extended to two days at +4C. The serum 222 concentration in antibody buffers was increased to 5% v/v and washes were extended 223 to one hour of submersion in 0.1% Tween-20 buffer three times. Nuclear counterstain 224 incubation was extended to 30 minutes. Tubules were mounted whole on standard 225 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 13 microscopy slides and covered with a standard #1.5 coverslip in prolong gold mounting 226 media (Thermo Fisher scientific). 227 Fluorescent microscopy of 48-well plates was completed on an inverted Nikon TI-228 Eclipse microscope, widefield scans were completed using a Nikon TI- High-Resolution 229 Widefield microscope, and whole mount tubule imaging was completed using an 230 Olympus Fluoview-1000 confocal microscope. 231

232 Transcriptomics analyses 233 Comparison of the transcriptomes of CIVMs derived from PND 5 and PND 10 tissue 234 reveal significant differences between the developmental stages. CIVMs seeded with 235 cells from three PND 5 tissue isolations were compared against CIVMs from three PND 236 10 tissue isolations. Each sample was a mixture of four CIVMs, representing a single 237 biological replicate. The expression of canonical germ cell marker genes was more 238 variable in the PND 5 derived CIVMs compared to the PND 10 CIVMs, somatic cells 239 markers were stable across both developmental stages (Figure 1.A). Based on a 240 significance threshold of log2 fold change greater than 0.5 or less than -0.5 and false 241 discovery rate (FDR) threshold of less than 0.05, 850 genes were differentially 242 expressed between the developmental stages (268 genes upregulated in PND 5 CIVMs 243 and 582 genes upregulated in PND 10 CIVMs; Figure 1.B). Multidimensional scaling 244 shows a higher intra-group distance between the PND 5 samples compared to the PND 245 10 samples, though the developmental stages remain clustered by age (Figure 1.C). 246 Gene Ontology (GO) over-representation analysis identified that 1,337 annotated 247 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 14 pathways (of 6,077 total annotated pathways) were over-represented, based on a fisher 248 statistic less than 0.05, among the differentially expressed genes. A keyword search of 249 the enriched GO terms using, ‘retino,’ ’male,’ ‘testis’ or ‘gonad’ identified nine pathways 250 with GO terms containing these keywords that are over-represented among the 251 differentially expressed genes between CIVMs from PND 5 and PND 10 tissue (Table 252 1). Critically, these overrepresented pathways include regulation of gonad development, 253 metabolism, and responses to retinoids, as well as responses to gonadotropins. An 254 additional 15 GO gene sets containing the keywords, ‘sertoli’, ‘leydig’, ‘testos’, ‘steroid’, 255 or ‘hormone’, were found to be overrepresented in the differentially expressed genes 256 between CIVMS from PND 5 and PND10 (Supplemental Table 2). The enrichment of 257 transcription factor-target genes between the two developmental stages reveals distinct 258 transcriptional regulatory activities, with target gene upregulation in PND 5 CIVMs (in 259 red and with positive activity score) or upregulation in PND 10 (in blue and with positive 260 activity score) (Figure 1.D). The top 20 transcription factors in each direction, based on 261 activity score from univariate linear modeling of the synthetic t statistic, are presented 262 (Figure 1.D), the full list of transcription factors with a statically significant association to 263 genes differentially regulated at either timepoint is included (supplemental figure 5). 264 265 Deconvolution analyses 266 The composition of CIVMs derived from each developmental stage was estimated using 267 bulk RNA-seq deconvolution methods. Three deconvolution approaches were used, 268 each a combination of an imputation algorithm and a deconvolution regression: DWLS 269 deconvolution with ALRA imputed reference data, OLS deconvolution with ALRA 270 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 15 imputed reference data, and DWLS deconvolution with non-imputed reference data. 271 These methods were identified in previous analysis as the best for neonatal testis 272 models (method benchmarked in [35]). Deconvolution estimates for each cell type were 273 relatively consistent between developmental stages and no significant differences were 274 identified through an unpaired Wilcoxon rank-sum test comparing the cell types at each 275 developmental stage (Figure 2.A). The average estimates for cell proportion across 276

and single-cell references were: 0.091 and 0.083 for germ cells, 0.116 and 277 0.097 for Leydig/Stromal cells, 0.309 and 0.330 for peritubular myoid cells, and 0.484 278 and 0.490 for Sertoli cells, in the PND 5 and PND 10 CIVMs respectively (Table 2). 279 280 Isotretinoin stimulus 281 The expression of stimulated by retinoic acid-8 (Stra8) and promyelocytic leukemia zinc 282 finger (Plzf, also referred to as Zbtb16) in neonatal testis CIVMs derived from PND 5 283 tissue (Figure 3.A for Stra8 and 3.B for Plzf) and PND 10 tissue (Figure 3.C for Stra8 284 and 3.D for Plzf) was quantified using RT-qPCR after exposure to isotretinoin for 24 285 hours. In the PND 5 derived cultures, Stra8 was not detectable in all samples from the 286 CIVMs in the control, 0.3, 3, and 30nM dose groups. Samples with Stra8 expression 287 below the threshold of quantification, considered any sample with a cycle threshold of 288 37 or greater in the RT-qPCR analyses, are displayed as “X” in the figure. All CIVMs 289 measured had quantifiable expression of Stra8 after exposure to the 300nM isotretinoin 290 dose. The expression of Stra8 was significantly higher in 300nM dose group relative to 291 the control using a linear mixed effects model of delta-CT as a function of dose with a 292 random intercept set for each biological sample to estimate marginal means, which 293 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 16 were adjusted post-hoc using Tukey’s correction. Expression of Plzf was quantifiable in 294 all PND 5 derived CIVMs across isotretinoin doses and the expression of Plzf was 295 significantly lower in the 300nM dose group using the same analyses as reported for 296 Stra8. 297 In the PND 10 testis derived CIVMs, the expression of Stra8 was detectable in all but 298 one sample in control group. The addition of isotretinoin significantly increased the 299 expression of Stra8 in the 3, 30, and 300 nM PND 10 CIVM dose groups (p-values 300 <0.001), with a trend in the 0.3 nM dose group (p-value of 0.10). The expression of Plzf 301 was quantifiable in all PND 10 CIVMs, with a significant decrease in the 30 and 300 nM 302 isotretinoin dose groups. Statistical significance in the PND 10 CIVMs was determined 303 following the same approach as for the PND 5 CIVMs. In all cases, when the cycle 304 threshold from RT-qPCR was below the threshold of quantification (CT of 37), the 305 sample was marked as an X on the plot and CT value used for delta-CT calculations. If 306 cycle threshold was below detection the sample CT was set to 40 for statistical 307 calculations and plotting purposes. Data displayed as expression relative to beta-actin 308 (housekeeping gene), by transforming the data to be 2 raised to the difference of the 309 gene of interest and housekeeping gene (2^ [gene – HK]). This representation 310 preserves the non-quantifiable samples, which are important for contextualizing the 311 Stra8 induction, and does not inflate fold-changes due to very lowly expressed genes. 312 Due to the low number of non-detects in the PND 10 CIVM data, the data is also 313 presented as log2 fold change (Supplemental Figure 1). 314 Expression of STRA8 in the PND 10 CIVM was confirmed through 315 immunocytochemistry (supplemental figure 1). STRA8 protein expression was nuclear 316 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 17 located and qualitatively at higher frequency in the PND 10 CIVM after 48 hours of 30 317 nM isotretinoin treatment compared to control. 318 319 Retinol metabolism 320 The expression of Stra8 in PND 10 testis CIVMs after addition of 3, 30, and 300 nM of 321 retinol was determined through RT-qPCR (Figure 4.A). The addition of retinol to the 322 CIVM increased Stra8 expression across all doses (p-values <0.0001). The co-323 treatment of the CIVMs with retinol and 1 µM BMS-189453 resulted in no meaningful 324 change in Stra8 expression at 3 nM retinol, but significant decreases in Stra8 325 expression at the 30 and 300 nM doses (p-values <0.0001). A two-hour pretreatment 326 and continued co-exposure of retinol with 1 µM WIN-18446 resulted in a significant 327 decrease in Stra8 expression across all retinol doses (at 3 nM retinol a p-value <0.0001, 328 at 30nM retinol a p-value of 0.0009, and at 300 nM retinol a p-value of 0.0142). The 329 magnitude of Stra8 reduction was relatively consistent in WIN-18446 pretreated CIVMs, 330 where the differences in delta CT estimates were 1.67, 1.19, and 0.90 for 3, 30, and 331 300nM retinol, respectively. Changes in Stra8 expression with BMS-189453 co-332 exposure were not consistent across retinol doses, where there was no significant 333 difference in delta CT estimates at the 3 nM retinol dose, but estimated delta CT 334 differences of 1.58 and 2.47 at 30 nM and 300 nM retinol respectively (calculated from 335 estimated marginal means). Data displayed as log2 fold change relative to median of 336 controls. The physiological relevance of retinoid metabolism within the PND 10 neonatal 337 testis CIVM is additionally supported by detection of retinoid metabolism related genes 338 within the CIVMs after three days in vitro (Figure 4.B). 339 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 18 340 Media formulation impact on spermatogonial phenotype 341 The impact of media supplementations on spermatogonial phenotype was measured 342 through RT-qPCR to determine the expression of two marker genes, Plzf and Star8, 343 over 15 days in vitro. Similar to the isotretinoin stimulation experiments (Figure 3), the 344 variation in response was higher in the PND 5 derived testis CIVMs than in the PND 10 345 testis CIVMs. (A) The expression of Plzf in PND 5 and PND 10 CIVMs was quantified 346 after 4 and 15 days in vitro maintained in base media (serum-free) conditions and after 347 supplementation with SSC growth factors (GDNF, GFRA1, FGF2), lipid rich albumin 348 (AlbuMAX), and defined bovine serum (Knockout Serum Replacement). In PND 5 testis 349 CIVMs, the expression of Plzf was not statistically different with supplementation of SSC 350 growth factors but was significantly decreased after four days of supplementation with 351 lipid rich albumin (p-value of 0.0001) and defined bovine serum (p-value <0.0001) 352 (Figure 5.A). Similarly, the expression of Stra8 was not statistically different with 353 supplementation of SSC growth factors but was significantly increased after four days of 354 supplementation with lipid rich albumin and defined bovine serum (p-values <0.0001) 355 (Figure 5.B). The expression of Stra8 in PND 5 control CIVMs after four days in vitro 356 was close to the limit of quantification, with many samples below quantification (a cycle 357 threshold of 37). After 15 days in vitro, a qualitatively higher proportion of samples were 358 above quantification, though no statistical assessment is reported due to high proportion 359 of sample below quantification. The expression of Stra8 increased between days in vitro 360 4 and 15 with lipid-rich albumin supplementation (p-value 0.004) and decreased with 361 defined serum supplementation (p-value <0.0001). The expression of Plzf significantly 362 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 19 decreased between days in vitro 4 and 15 in the control media CIVMs (p-value 363 <0.0001), while all supplemented CIVMs had no notable change in Plzf expression 364 (Figure 5.A). 365 366 In PND 10 testis tissue derived CIVMs, the expression of Plzf trended lower, but was 367 not statistically different with supplementation of SSC growth factors (p-value of 0.07). 368 Plzf was significantly decreased after four days of supplementation with lipid rich 369 albumin and defined bovine serum (p-values <0.0001) (Figure 5.A). Stra8 expression 370 was not statistically different with supplementation of SSC growth factors but was 371 significantly increased after four days of supplementation with lipid rich albumin (p value 372 of 0.011) and defined bovine serum (p-values <0.0001) (Figure 5.B). After 15 days in 373 vitro, a higher proportion of CIVMs maintained in control media or SSC growth factors 374 had Stra8 expression below quantification relative to day in vitro four. The expression of 375 Stra8 increased between days in vitro 4 and 15 with lipid-rich albumin supplementation 376 (p-value 0.03) but was unchanged with defined serum supplementation (p-value 0.96). 377 378 Overall, the culture viability was not qualitatively different with media supplementation 379 and was qualitatively similar to our previous investigations using similar culture methods 380 [19]. Notably, media supplementations that contained serum products (Knockout Serum 381 Replacement and AlbuMAX; Figures 6.A-D) led to distinct morphological differences 382 relative to the serum-free cultures after four days (Control and SSC growth factors 383 supplementation; Figures 6.E-H) in CIVMs derived from both developmental stages 384 (PND 10 [Figure 6.C-D,G-H] and PND 5 [Figure 6.A-B,E-F]) after four days in vitro. 385 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 20 Sertoli cells and spermatogonia were retained through 15 days in vitro and confirmed 386 through immunocytochemistry for SOX9 (Sertoli cells) and DDX4 (spermatogonia) 387 (Supplemental Figures 2.A-B). 388 389

390 391 Neonatal rodents remain the preeminent model organism for studying testicular 392 development and reproductive dynamics. The extension of in vivo rodent research to in 393 vitro models provides a reductionist approach to expand the breadth of molecular 394 analyses and biological manipulation possible in reproductive and developmental 395 studies. Further, the progression of in vitro models based on non-human tissue remains 396 a critical way to reduce the need for animals in toxicology and preclinical research for 397 organ systems where human tissue is not available for research, like the developing 398 testis [37]. However, in vitro models require careful assessment to identify biological 399 significance from in vitro artifacts and to benchmark model phenotype within a dynamic 400 developing system. The impact of in vitro culture on spermatogonial phenotype is a 401 critical consideration for a range of reproductive and developmental research: from 402 novel models of in vitro spermatogenesis to molecular analyses of toxicants on specific 403 spermatogonial stages. We find that while complex in vitro models (CIVMs) seeded with 404 primary testis cells from post-natal day (PND) 5 and PND 10 Sprague Dawley rats 405 retain similar cell type compositions, with transcriptomic signatures of Sertoli cells, 406 Leydig cells, Germ cells, and peritubular myoid cells, there are distinct differences in the 407 overall transcriptome, transcription factor activity, and retinoid responsiveness. These 408 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 21 manifest in disparate responses of spermatogonia within the CIVMs at PND 10 to those 409 at PND 5 after exposure to physiologically relevant levels of retinoid and distinct 410 responses to common cell culture media supplements. These findings have critical 411 implications for in vitro studies of spermatogonial dynamics, where various neonatal 412 ages or media supplements have been used. Our results support that in neonatal rat 413 models, PND 10 testis can be a consistent and reproducible model for retinoic acid 414 metabolism and spermatogonial response. 415 416 Transcriptomics analyses 417 Overall, the transcriptome of CIVMs from PND 5 rat testis was distinct from that of 418 CIVMs seeded from PND 10 rat testis. We found that more than 700 genes were 419 differentially expressed between CIVMs at the two developmental stages following the 420 standard edgeR workflow. To apply biological meaning to these differentially expressed 421 genes, we considered if certain annotated pathways were over-represented within our 422 differentially expressed gene set, suggesting those pathways may be specifically 423 different between the developmental stages. This over-representation analysis revealed 424 significant enrichment in a number of pathways (supplemental table 1), but review of 425 this list identified specific developmentally relevant pathways pertaining to retinoic acid 426 response and metabolism (Table 1) that were differentially expressed between the PND 427 5 and PND 10 CIVMs. Assessment of the major genes considered part of retinoic acid 428 metabolism in the testis revealed that while certain retinoic acid metabolism related 429 genes were differentially expressed (i.e. Stra6, Cyp26b1, Rbp1, Crabp2; adj. p-value 430 <0.05), each was upregulated in PND 5 CIVMs relative to the PND 10 CIVMs 431 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 22 (supplemental figure 6). Aside from Cyp26b1, which removes the bioactive retinoic acid 432 and is critical for normal testis development [38], the impact of the differential 433 expression of Stra6, Rbp1, or Crabp2 is uncertain as these are responsible for 434 intracellular transport or binding of retinoids. This follows the trend of the broader 435 dataset where the majority of genes differentially expressed within the retinoic acid 436 related ontologies (GO:0042573, GO:0071300, and GO:0001523) are upregulated in 437 PND 5 relative to PND 10 (supplemental figure 7). 438 To further address the differences in transcriptome between the PND 5 and PND 10 439 testis CIVMs, the differentially expressed genes were modeled through their canonical 440 association with transcription factors (Figure 1.D). Target genes of the transcription 441 factor cleraxis (Scx), associated with Sertoli cell growth and maturation [39], were more 442 active in the PND 5 CIVMs than in the PND 10 CIVMs. Limited information is available 443 on the role of Scx in the developing testis. Interestingly, the previous report on Scx in 444 the testis [39], using freshly isolated rat Sertoli cells, identified a transient increase in 445 Scx expression at PND 10, whereas our report did not find a significance difference in 446 expression of Scx itself between PND 5 and PND 10 CIVMs, rather an increase in 447 inferred Scx activity in PND 5 CIVMs based on differential expression of target genes. 448 The canonical Sertoli cell marker, Sox9, was not differentially expressed between the 449 CIVM developmental stages, though the transcription factor activity inference identified 450 that multiple Sry‐ type HMG box proteins (SOX Family), Sox9, Sox5, and Sox6, were 451 more active in the PND 5 CIVMs. Based on previous reports (referenced in [40]), it is 452 unlikely that all of these SOX transcription factors were expressed in the same cell type, 453 however, they are related to a coordinated response with Col2a1 upregulation, which is 454 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 23 significantly upregulated in the PND 5 CIVMs reported here. This supports a 455 coordinated matrix remodeling in the PND 5 CIVMs after three days in vitro that is not 456 present in the PND 10 CIVMs after three days in vitro. 457 458 Deconvolution analyses 459 While there were significant differences in the transcriptomes of CIVMs derived from 460 PND 5 and PND 10 testis, our deconvolution analyses support that there are not broad 461 differences in cell type proportions between these developmental stages (Figure 2). 462 Qualitatively, the whole-mount imaging of the primary seminiferous tubules confirms that 463 DDX4+ germ cells are present at both developmental stages in the rat, though variation 464 across tubule length make quantitative assessments unreliable (supplemental figure 2). 465 This is also supported by a lack of significant diffntial expression of canonical somatic 466 cell marker genes: Cyp11a1 for Leydig cells, Sox9 and Wt1 for Sertoli cells, and Acta2 467 for peritubular myoid cells (Figure 1.A). Our analyses find that the bulk transcriptome of 468 both the PND 5 and PND 10 CIVMs are composed of ~ 9-10% germ cells, ~10-11% 469 Leydig cells, ~30-33% peritubular myoid cells, and ~48-49% Sertoli cells (Table 2.). The 470 PND 10 testis are larger than the PND 5 testis, though our data support that in vitro 471 each maintains a similar cell type proportion over time in vitro, suggesting the cells are 472 proliferating at similar rates. The convergence on this cell type proportion may be a 473 function of differential viability in culture, where germ cells may be more fragile 474 compared to peritubular cells which may be more robust, as well as the lack of pituitary 475 signaling that can drive cell type specific proliferation. In vivo, it is expected that the 476 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 24 Sertoli cells are highly proliferative at PND 5, whereas at PND 10 the spermatogonia 477 and other somatic cells are highly proliferative as well [6]. 478 479 Isotretinoin stimulus 480 While the cell type proportions may be similar between the PND 5 and PND 10 CIVMs, 481 the response to retinoic acid is distinctly different between the two CIVM developmental 482 stages based on induction of stimulated by retinoic acid-8 (Stra8, a marker of 483 spermatogonial differentiation) and promyelocytic leukemia zinc finger (Plzf; a marker of 484 undifferentiated spermatogonia) expression (Figure 3). Using doses of isotretinoin that 485 are reflective of physiologically relevant levels of bioactive retinoids based on published 486 adult human tissue data [41], where 0.3 nM is reflective of isotretinoin in testicular 487 homogenate, 3 nM is reflective of isotretinoin in serum, while 30nM and 300nM are 488 dose escalations. Isotretinoin (13-cis retinoic acid), is an endogenous isoform of the 489 bioactive all-trans-retinoic acid (ATRA) and is used as a pharmaceutical as it is a poor 490 substrate for the CYP26B- enzymes that degrade bioactive retinoids [42]. In the PND 5 491 testis derived CIVMs, Stra8 was largely undetectable at baseline or at the 0.3 nM dose 492 of isotretinoin. In the 3 nM and 30 nM dose groups, there were more samples with 493 quantifiable expression of Stra8 (i.e., above the 37-cycle threshold in RT-qPCR), though 494 only at the 300 nM dose group did Stra8 expression increase significantly in the PND 5 495 CIVMs (Figure 3.A). In the same PND 5 CIVMs, Plzf was expressed above quantitation 496 in all samples, but showed no significant dose-relationship (Figure 3.B). In the PND 10 497 CIVMs, there was a significant increase in Stra8 across isotretinoin doses parallelling a 498 decrease in Plzf expression (Figure 3.C-D). The response of spermatogonia in PND 10 499 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 25 CIVMs to physiological levels of bioactive retinoic acid (isotretinoin) is a promising 500 advance in in vitro spermatogonial differentiation studies, where supraphysiological 501 levels may exceed relevant enzyme capacity, decreasing translational power. Our 502 finding that nanomolar doses of isotretinoin elicit a spermatogonial response suggest 503 that a possible “target” for tissue concentrations of RA in patients with NOA in clinical 504 trials of retinoid-based treatments of men being treated for infertility from reduced sperm 505 production may be at least 30 nM or higher to initiate or stimulate spermatogenesis. 506 While follow-up clinical studies to refine this hypothesis are needed, clinical dose ranges 507 can be refined using this in vitro method. This is very advantageous for dosing high 508 affinity interactions, like the retinoids and retinoid receptors, where the nanomolar range 509 is sufficient for activity. Particularly in cases where a therapeutic may target increased 510 spermatogonial differentiation as a treatment for a non-obstructive oligozoospermia, 511 possible therapeutics could be trialed in vitro, where impacts on spermatogonial 512 differentiation can be measured in a reductionist system at the cellular level. 513 While the response to exogenous isotretinoin is important for mechanistic studies of 514 spermatogonial differentiation and for screening therapeutics to increase sperm count, 515 the metabolism of bioactive ATRA itself is a critical step to recapitulate in vitro to 516 progress in vitro investigation of infertility causes and screening potential 517 contraceptives. Estimates show that a very small percentage (< 1%) of bioactive retinoic 518 acid within the testis originated elsewhere and was brough via circulation, rather the 519 extensive metabolic network of retinoid metabolizing enzymes are responsible for 520 maintaining the intratesticular retinoid environment [43]. Inhibition of the aldehyde 521 dehydrogenase enzymes (ALDH family) within the testis has historically been a focus 522 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 26 for development of male contraceptives [44,45], and inhibiting retinoic acid activity 523 continues to be a target for novel contraceptives [46–48]. Our results here show that the 524 PND 10 testis derived CIVM retains the capacity to metabolize retinol to ATRA at levels 525 sufficient to drive Stra8 expression (Figure 4.A), our RNA-seq analyses show that genes 526 for key the critical metabolizing enzymes, transporters, and binding proteins are 527 expressed within the CIVMs (Figure 4.B). Through the addition of specific inhibitors of 528 retinoic acid formation by inhibiting ALDH retinyl metabolism (WIN-18446) and retinoic 529 acid receptor activation by allosterically binding the RAR complex (BMS-189453), our 530 PND 10 CIVM data supports that these critical processes are retained in vitro and are 531 inducible and inhibitable (Figure 4.A). 532 Given our data supporting that these developing testis CIVMs are responsive to 533 retinoids, we investigated if residual retinoids in commonly used cell culture 534 supplements were sufficient to drive spermatogonial phenotype. Using three media 535 supplement approaches common for in vitro testis models: spermatogonial stem cell 536 (SSC) growth factors [20], lipid-rich albumin (AlbuMAX, Thermo Fisher Scientific) [21], 537 and a defined bovine serum (Knockout Serum Replacement) [22,23], we found that 538 addition of serum derived components impacted the phenotype of the spermatogonia in 539 vitro (Figure 5). Using Plzf and Stra8 gene expression as metrics for undifferentiated 540 and differentiating spermatogonia respectively, our data support that addition of serum 541 derived components, lipid rich albumin and defined bovine serum, increase the 542 expression of Stra8 after four days in vitro in both the PND 5 and PND 10 derived testis 543 CIVMs. The upregulated of Stra8 in the PND 10 CIVMs was accompanied by a 544 significant decrease in Plzf, while there was not a significant decrease in Plzf 545 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 27 expression in the PND 5 CIVMs (Figure 5). Notably, the additions of lipid rich albumin 546 and defined bovine serum were sufficient to drive Stra8 expression in the PND 5 547 CIVMs, while it was generally unquantifiable under serum-free (control and SSC growth 548 factors) conditions. This is hypothesized to be related to residual retinoids within the 549 serum products, though not directly measured in this report. Previous reports show that 550 residual retinoids in cell culture products can be biologically active and our data 551 presented in this report support that the activity of residual retinoids is critical to 552 consider in in vitro testis models [49]. 553 When these CIVMs were maintained for 15 days in vitro, there was a significant 554 decrease in Plzf expression in the control media CIVMs derived from both ages, PND 5 555 and PND 10, there was not a significant decrease in Plzf gene expression in the SSC 556 growth factor CIVMS between in vitro days 4 and 15. While 15 days was likely not a 557 long enough culture period to see significant proliferation of spermatogonia, where 558 previous studies are carried out for up 10 weeks and see significant expansion of SSCs 559 [20], our data suggest the addition of SSC growth factors may support the 560 undifferentiated spermatogonial phenotype (characterized here by Plzf expression). As 561 expected based on the isotretinoin exposure experiments (Figure 3), there is little to no 562 baseline expression of Stra8 in PND 5 testis CIVMs, whereas it is consistently 563 expressed above quantification in PND 10 derived CIVMs. Additionally, there is a lower 564 variation of spermatogonial response within the PND 10 derived testis compared to a 565 much larger variation in the PND 5 derived CIVMs. This is an interesting finding 566 considering a previous work which found that retinoic acid responsiveness was a factor 567 of spermatogonial phenotype, as KIT+ differentiating spermatogonia were less 568 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 28 responsive. However, this could be due to comparing relative changes, where the 569 expression of Stra8 in already differentiating spermatogonia is less inducible than 570 undifferentiated spermatogonia [50]. 571 Overall, we propose that in vitro models of the developing testis are powerful tools to 572 study the dynamics of retinoids and spermatogonial phenotype. We find that 573 physiological levels of retinoids are sufficient to drive Stra8 expression in testis CIVMs 574 derived from PND 10 tissue. We report similarities in cell type proportions, but distinct 575 differences in transcriptomes and inferred transcription factor activity of CIVMs derived 576 from PND 5 testis and PND 10 testis. Additional research is warranted to further 577 investigate the dynamics of spermatogonia in vitro. A model that recapitulates the 578 spermatogonial niche dynamics that support proliferation of undifferentiated 579 spermatogonia through retinoic acid driven differentiation will advance reproductive 580 biology and toxicology and reduce the need for lengthy in vivo studies. 581 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 29 Data Availability 582 All code is available on GitHub: github.com/bchansen3/Hansen_2025_Spermatogonia 583 RNA-seq data is available as FASTQ files and gene counts on the Gene Expression 584 Omnibus: GSE314458. 585 Any other additional data available upon request. 586 Author Contributions 587 Conceptualization: BCH, EMF, JKA, EJK; Methodology: BCH, SMH, LAH; Formal 588 analysis: BCH; Resources: Z.J; Writing—original draft: BCH; Writing—review & editing: 589 BH, SMH, LAH, EMF, JKA, EJK; Supervision: JKA, EMF, EJK; Funding acquisition: 590 BCH, EMF, EJK. 591

592 The authors would like to thank members of the Isoherranen laboratory at the University 593 of Washington for their expertise on using retinoids in vitro, specifically: Aprajita Yadav, 594 Aurora Authement, and Jiayao Chen. 595 The authors would like to thank Dashiel Cockrill for insightful questions and support on 596 the experiments presented here. 597 The authors would also like to thank Dr. Dale Hailey at the UW ISCRM Garvey Imaging 598 core for microscopy support and the University of Washington Department of 599 Environmental and Occupational Health Sciences IT team for maintenance and support 600 of computing servers. 601 602 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 30

603 [1] Chandra A, Copen CE, Stephen EH. Infertility and impaired fecundity in the United 604 States, 1982-2010: data from the National Survey of Family Growth. Natl Health 605 Stat Report 2013:1–18, 1 p following 19. 606 [2] US FDA. Testicular Toxicity: Evaluation During Drug Development. Guidance for 607 Industry. Center for Drug Evaluation and Research (CDER); 2018. 608 [3] Saldutti LP , Beyer BK, Breslin W, Brown TR, Chapin RE, Enright B, Faustman E, 609 Foster PMD, Hartung T, Kelce W, Kim JH, Loboa EG, et al. In vitro testicular toxicity 610 models: Opportunities for advancement via biomedical engineering techniques. 611 ALTEX 2013; 30:353–377. 612 [4] Nam CS, Campbell KJ, Acquati C, Bole R, Adler A, Collins DJ, Collins E, Samplaski 613 M, Anderson-Bialis J, Andino JJ, Asafu-Adjei D, Gaskins AJ, et al. Deafening 614 Silence of Male Infertility. Urology 2023; 182:111–124. 615 [5] Krausz C, Escamilla AR, Chianese C. Genetics of male infertility: from research to 616 clinic. Reproduction 2015; 150:R159–R174. 617 [6] Picut CA, Remick AK, de Rijk EPCT, Simons ML, Stump DG, Parker GA. Postnatal 618 Development of the Testis in the Rat: Morphologic Study and Correlation of 619 Morphology to Neuroendocrine Parameters. Toxicol Pathol 2015; 43:326–342. 620 [7] Orth JM, Gunsalus GL, Lamperti AA. Evidence from Sertoli cell-depleted rats 621 indicates that spermatid number in adults depends on numbers of Sertoli cells 622 produced during perinatal development. Endocrinology 1988; 122:787–794. 623 [8] Rebourcet D, Darbey A, Monteiro A, Soffientini U, Tsai YT, Handel I, Pitetti J-L, Nef 624 S, Smith LB, O’Shaughnessy PJ. Sertoli Cell Number Defines and Predicts Germ 625 and Leydig Cell Population Sizes in the Adult Mouse Testis. Endocrinology 2017; 626 158:2955–2969. 627 [9] Matilionyte G, Tharmalingam MD, Sanou I, Lopes F, Lane S, Stukenborg J-B, 628 Spears N, Anderson RA, Mitchell RT. Maintenance of Sertoli Cell Number and 629 Function in Immature Human Testicular Tissues Exposed to Platinum-Based 630 Chemotherapy—Implications for Fertility Restoration. Frontiers in Toxicology 2022; 631 4. 632 [10] Berkowitz GS, Lapinski RH, Gazella JG, Dolgin SE, Bodian CA, Holzman IR. 633 Prevalence and Natural History of Cryptorchidism. Pediatrics 1993; 92:44–49. 634 [11] Fijak M, Meinhardt A. The testis in immune privilege. Immunological Reviews 2006; 635 213:66–81. 636 [12] Sharpe RM, McKinnell C, Kivlin C, Fisher JS. Proliferation and functional 637 maturation of Sertoli cells, and their relevance to disorders of testis function in 638 adulthood. Reproduction 2003; 125:769–784. 639 [13] Hofmann M-C, McBeath E. Sertoli Cell-Germ Cell Interactions Within the Niche: 640 Paracrine and Juxtacrine Molecular Communications. Frontiers in Endocrinology 641 2022; 13. 642 [14] Wu J-W, Wang R-Y, Guo Q-S, Xu C. Expression of the retinoic acid-metabolizing 643 enzymes RALDH2 and CYP26b1 during mouse postnatal testis development. 644 Asian J Androl 2008; 10:569–576. 645 [15] Tharakan T, Luo R, Jayasena CN, Minhas S. Non-obstructive azoospermia: current 646 and future perspectives. Fac Rev 2021; 10:7. 647 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 31 [16] Kanbar M, Vermeulen M, Wyns C. Organoids as tools to investigate the molecular 648 mechanisms of male infertility and its treatments. Reproduction 2021; 161:R103–649 R112. 650 [17] Ghanami Gashti N, Sadighi Gilani MA, Abbasi M. Sertoli cell-only syndrome: 651 etiology and clinical management. J Assist Reprod Genet 2021; 38:559–572. 652 [18] Baert Y, Goossens E, van Saen D, Ning L, in’t Veld P, Tournaye H. Orthotopic 653 grafting of cryopreserved prepubertal testicular tissue: in search of a simple yet 654 effective cryopreservation protocol. Fertility and Sterility 2012; 97:1152-1157.e2. 655 [19] Wegner S, Hong S, Yu X, Faustman EM. Preparation of Rodent Testis Co-Cultures. 656 Curr Protoc Toxicol 2013; 0 16:Unit-16.10. 657 [20] Kubota H, Avarbock MR, Brinster RL. Growth factors essential for self-renewal and 658 expansion of mouse spermatogonial stem cells. Proc Natl Acad Sci U S A 2004; 659 101:16489–16494. 660 [21] Matsumura T, Sato T, Abe T, Sanjo H, Katagiri K, Kimura H, Fujii T, Tanaka H, 661 Hirabayashi M, Ogawa T. Rat in vitro spermatogenesis promoted by chemical 662 supplementations and oxygen-tension control. Sci Rep 2021; 11:3458. 663 [22] Richer G, Baert Y , Goossens E. In-vitro spermatogenesis through testis modelling: 664 Toward the generation of testicular organoids. Andrology 2020; 8:879–891. 665 [23] Sato T, Katagiri K, Yokonishi T, Kubota Y , Inoue K, Ogonuki N, Matoba S, Ogura A, 666 Ogawa T. In vitro production of fertile sperm from murine spermatogonial stem cell 667 lines. Nat Commun 2011; 2:472. 668 [24] Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P , Chaisson 669 M, Gingeras TR. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 2013; 670 29:15–21. 671 [25] Robinson MD, Oshlack A. A scaling normalization method for differential expression 672 analysis of RNA-seq data. Genome Biology 2010; 11:R25. 673 [26] Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for 674 differential expression analysis of digital gene expression data. Bioinformatics 675 2010; 26:139–140. 676 [27] Alexa A, Rahnenfuhrer J. topGO: Enrichment Analysis for Gene Ontology. 2024. 677 [28] The Gene Ontology Consortium, Aleksander SA, Balhoff J, Carbon S, Cherry JM, 678 Drabkin HJ, Ebert D, Feuermann M, Gaudet P , Harris NL, Hill DP, Lee R, et al. The 679 Gene Ontology knowledgebase in 2023. Genetics 2023; 224:iyad031. 680 [29] Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP , 681 Dolinski K, Dwight SS, Eppig JT, Harris MA, Hill DP, et al. Gene Ontology: tool for 682 the unification of biology. Nat Genet 2000; 25:25–29. 683 [30] Badia-i-Mompel P, Vélez Santiago J, Braunger J, Geiss C, Dimitrov D, Müller-Dott 684 S, Taus P, Dugourd A, Holland CH, Ramirez Flores RO, Saez-Rodriguez J. 685 decoupleR: ensemble of computational methods to infer biological activities from 686 omics data. Bioinformatics Advances 2022; 2:vbac016. 687 [31] Türei D, Valdeolivas A, Gul L, Palacio‐ Escat N, Klein M, Ivanova O, Ölbei M, Gábor 688 A, Theis F, Módos D, Korcsmáros T, Saez‐ Rodriguez J. Integrated intra‐ and 689 intercellular signaling knowledge for multicellular omics analysis. Molecular 690 Systems Biology 2021; 17:e9923. 691 [32] Türei D, Korcsmáros T, Saez-Rodriguez J. OmniPath: guidelines and gateway for 692 literature-curated signaling pathway resources. Nat Methods 2016; 13:966–967. 693 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 32 [33] Tsoucas D, Dong R, Chen H, Zhu Q, Guo G, Yuan G-C. Accurate estimation of cell-694 type composition from gene expression data. Nat Commun 2019; 10:2975. 695 [34] Linderman GC, Zhao J, Roulis M, Bielecki P, Flavell RA, Nadler B, Kluger Y. Zero-696 preserving imputation of single-cell RNA-seq data. Nat Commun 2022; 13:192. 697 [35] Hansen BC, Arian CM, Zeng Y, Takezawa MG, Theberge AB, Faustman EM, 698 Thummel KE, Kelly EJ. Leveraging RNA-seq deconvolution to improve complex in 699 vitro model characterization. Journal of Biological Chemistry 2025; 301:110510. 700 [36] Alves-Lopes JP , Söder O, Stukenborg J-B. Use of a three-layer gradient system of 701 cells for rat testicular organoid generation. Nat Protoc 2018; 13:248–259. 702 [37] Devine P, Guha M, Ekert J, Kopec A, Gosset J, Freag M, Wagoner M, Hewitt P , 703 Harris K, Lemmens M, Sadrieh N, Mendrick D, et al. Considerations from the 704 pharmaceutical industry (IQ MPS affiliate) workshop on animal microphysiological 705 systems and 3Rs in drug development. ALTEX - Alternatives to Animal 706 Experimentation 2025. 707 [38] Bowles J, Feng C-W, Ineson J, Miles K, Spiller CM, Harley VR, Sinclair AH, 708 Koopman P . Retinoic Acid Antagonizes Testis Development in Mice. Cell Reports 709 2018; 24:1330–1341. 710 [39] Muir T, Sadler-Riggleman I, Skinner MK. Role of the Basic Helix-Loop-Helix 711 Transcription Factor, Scleraxis, in the Regulation of Sertoli Cell Function and 712 Differentiation. Molecular Endocrinology 2005; 19:2164–2174. 713 [40] Lefebvre V, Li P , de Crombrugghe B. A new long form of Sox5 (L‐ Sox5), Sox6 and 714 Sox9 are coexpressed in chondrogenesis and cooperatively activate the type II 715 collagen gene. The EMBO Journal 1998; 17:5718–5733. 716 [41] Nya-Ngatchou JJ, Arnold SLM, Walsh TJ, Muller CH, Page ST, Isoherranen N, 717 Amory JK. Intratesticular 13-cis retinoic acid is lower in men with abnormal semen 718 analyses: a pilot study. Andrology 2013; 1:325–331. 719 [42] Amory JK, Ostrowski KA, Gannon JR, Berkseth K, Stevison F, Isoherranen N, 720 Muller CH, Walsh T. Isotretinoin administration improves sperm production in men 721 with infertility from oligoasthenozoospermia: a pilot study. Andrology 2017; 5:1115–722 1123. 723 [43] Vernet N, Dennefeld C, Rochette-Egly C, Oulad-Abdelghani M, Chambon P, 724 Ghyselinck NB, Mark M. Retinoic Acid Metabolism and Signaling Pathways in the 725 Adult and Developing Mouse Testis. Endocrinology 2006; 147:96–110. 726 [44] Paik J, Haenisch M, Muller CH, Goldstein AS, Arnold S, Isoherranen N, Brabb T, 727 Treuting PM, Amory JK. Inhibition of Retinoic Acid Biosynthesis by the 728 Bisdichloroacetyldiamine WIN 18,446 Markedly Suppresses Spermatogenesis and 729 Alters Retinoid Metabolism in Mice. J Biol Chem 2014; 289:15104–15117. 730 [45] Heller CG, Moore DJ, Paulsen CA. Suppression of spermatogenesis and chronic 731 toxicity in men by a new series of bis(dichloroacetyl) diamines. Toxicology and 732 Applied Pharmacology 1961; 3:1–11. 733 [46] Hogarth CA, Amory JK, Griswold MD. Inhibiting Vitamin A Metabolism As An 734 Approach To Male Contraception. Trends Endocrinol Metab 2011; 22:136–144. 735 [47] Noman MAA, Kyzer JL, Chung SSW, Wolgemuth DJ, Georg GI. Retinoic acid 736 receptor antagonists for male contraception: current status†. Biol Reprod 2020; 737 103:390–399. 738 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 33 [48] Mannowetz N, Chung SSW, Maitra S, Noman MAA, Wong HL, Cheryala N, Bakshi 739 A, Wolgemuth DJ, Georg GI. Targeting the retinoid signaling pathway with YCT-529 740 for effective and reversible oral contraception in mice and primates. Commun Med 741 2025; 5:68. 742 [49] Czuba LC, Zhong G, Yabut K, Isoherranen N. Analysis of Vitamin A and Retinoids 743 in Biological Matrices. Methods Enzymol 2020; 637:309–340. 744 [50] Zhou Q, Li Y, Nie R, Friel P, Mitchell D, Evanoff RM, Pouchnik D, Banasik B, 745 McCarrey JR, Small C, Griswold MD. Expression of Stimulated by Retinoic Acid 746 Gene 8 (Stra8) and Maturation of Murine Gonocytes and Spermatogonia Induced 747 by Retinoic Acid In Vitro1. Biology of Reproduction 2008; 78:537–545. 748 [51] Guo J, Grow EJ, Mlcochova H, Maher GJ, Lindskog C, Nie X, Guo Y , Takei Y , Yun 749 J, Cai L, Kim R, Carrell DT, et al. The adult human testis transcriptional cell atlas. 750 Cell Research 2018; 28:1141–1157. 751 [52] Sohni A, Tan K, Song H-W, Burow D, de Rooij DG, Laurent L, Hsieh T-C, Rabah R, 752 Hammoud SS, Vicini E, Wilkinson MF. The Neonatal and Adult Human Testis 753 Defined at the Single-Cell Level. Cell Reports 2019; 26:1501-1517.e4. 754 [53] Tan K, Song H-W, Wilkinson MF. Single-cell RNAseq analysis of testicular germ 755 and somatic cell development during the perinatal period. Development 2020; 756 147:dev183251. 757 758 759 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 34 Figures: 760 Figure 1. Transcriptomic Comparison of CIVMs derived from PND 5 and PND 10 761 after three days in vitro. 762 Bulk-RNA transcriptomes were collected from PND 5 and PND 10 testis CIVMs after 763 three days in culture. (A) The expression of canonical marker genes for germ cells in the 764 neonatal testis alongside markers of somatic cell types. (B) Differentially expressed 765 genes based on log2 fold change greater than 0.5 or less than -0.5 and false discovery 766 rate (FDR) threshold of less than 0.05 shown in volcano plot with genes upregulated in 767 PND 10 shown as a negative log2 fold change and genes upregulated in PND5 shown 768 as a positive log2 fold change. (C) Multidimension scaling supports that the intra-769 developmental stage samples cluster together and suggests that the intra-sample 770 variance is higher in PND 5 CIVMs. Percentage for each dimension represents the 771 variance captured by that component. (D) Inferred transcription factor (TF) activity from 772 differentially expressed genes, top 20 TFs for each developmental stage included (PND 773 5 in red with positive activity score, PND 10 in blue with negative activity score). 774 775 Figure 2. RNA-seq Deconvolution based- estimates of cell type proportion within 776 CIVMs of neonatal testis from PND 5 and PND 10 tissue. 777 Bulk-transcriptomes from the PND 5 and PND 10 derived CIVMs were deconvolved 778 using publicly available methods. Shape indicated the combination of deconvolution 779

and color indicates the single-cell reference used. Each point is the mean of the 780 three biological replicate CIVM RNA-seq samples for each developmental stage per 781 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 35 each combination of method and single cell reference. Boxplots represent the full 782 distribution of estimates (including biological replicate CIVMs). There were no 783 statistically significant differences identified between the cell type estimates at the two 784 stages, as determined by an un-paired Wilcoxon ranked sum test p-value greater than 785 0.05 for all cell types (denoted by ‘ns’). 786 787 Figure 3. Expression of markers of spermatogonial phenotype after exposure to 788 isotretinoin in vitro. 789 CIVMs derived from neonatal testis from PND 5 and PND 10 testis were exposure to 790 isotretinoin for 24 hours. (A) Expression of Stra8 in PND 5 derived CIVMs determined 791 by RT-qPCR. In the control, 0.3, 3, and 30 nM isotretinoin dose groups the expression 792 of Stra8 contained samples below the threshold of quantification. The expression of 793 Stra8 was significantly increased in the 300 nM isotretinoin dose group (p-value 794 <0.0001). (B) Expression of Plzf in PND 5 derived CIVMs determined by RT-qPCR. All 795 samples were above the threshold of quantification. The expression of Plzf in the 300 796 nM isotretinoin dose group was significantly lower than control (p-value of 0.026). (C) 797 Expression of Stra8 in PND 10 derived CIVMs determined by RT-qPCR. Aside from one 798 control sample, all samples were above the threshold of quantification. The expression 799 of Stra8 was significantly increased in the 3,30, and 300 nM isotretinoin dose groups (p-800 values all <0.0001). (D) Expression of Plzf in PND 10 derived CIVMs determined by RT-801 qPCR. All samples were above the threshold of quantification. The expression of Plzf in 802 the 30 and 300nM isotretinoin dose groups were significantly lower than control (p-803 values <0.0001). 804 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 36 Figure 4. Retinol metabolism and inhibition within a PND 10 testis CIVM. 805 Quantification of Stra8 activation after addition of retinol to PND 10 testis CIVM using 806 RT-qPCR. (A) Addition of retinol into PND 10 testis CIVM at 3, 30, and 300 nM 807 increases Stra8 expression (blue). Co-exposure with BMS-189453 (green) and pre-808 treatment with WIN-18446 (red) show reduction in Stra8 expression across doses. (B) 809 Expression of key retinoid metabolism genes in RNA-seq analysis of PND 10 derived 810 testis CIVMs. Includes expression of retinol transporter Stra6; metabolic enzymes 811 Rdh10, Aldh1a1, and Cyp26b1; retinoid nuclear receptors and co-receptors Rxr (alpha 812 and beta isoforms) and Rar (alpha and gamma isoforms); and transport proteins Rbp4, 813 Crabp1, and Crabp2. Dashed line marks a log2 CPM of 1 as a gene expression 814 threshold. 815 816 Figure 5. Spermatogonial phenotype after media supplementation media for 15 817 days in vitro. 818 The expression of Plzf and Stra8 quantified by RT-qPCR at day in vitro 4 and day in 819 vitro 15 in PND 5 tissue CIVMs and PND 10 tissue CIVMs in control and media 820 supplementation conditions. Color indicates day in vitro, dashed line divides plot by 821 PND 5 CIVMs and PND 10 CIVMs. Points that are “X” indicate samples that were below 822 the limit of quantification (cycle threshold of 37). Data displayed as expression relative 823 to beta-actin (housekeeping gene (HK), (2^ [gene – HK])). 824 825 Figure 6. Media supplementation cell morphology 826 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 37 Brightfield images of PND 5 testis tissue derived CIVMs (A-B, E-F) and PND 10 testis 827 tissue derived CIVMs (C-D, G-H) after 4 days in vitro maintained in either: control 828 media, SSC growth factors (GDNF, GFRA1, FGF2), lipid rich albumin (AlbuMAX), and 829 defined bovine serum (Knockout Serum Replacement). 830 831 832 Tables: 833 Table 1. Gene Ontology overrepresentation analyses of differentially expressed genes 834 between CIVMs derived from PND 5 and PND 10 testis. 835 GO Pathway GO Pathway Term Annotated Genes Significant DE Genes Expected DE Genes Fisher Statistic GO:0042573 retinoic acid metabolic process 11 6 1.02 0.00020 GO:0060135 maternal process involved in female pregnancy 76 17 7.07 0.00050 GO:0071371 cellular response to gonadotropin stimulus 37 10 3.44 0.00157 GO:0034698 response to gonadotropin 62 13 5.77 0.00403 GO:0071300 cellular response to retinoic acid 58 12 5.39 0.00629 GO:1905939 regulation of gonad development 14 5 1.30 0.00676 GO:0001523 retinoid metabolic process 39 9 3.63 0.00825 GO:1905941 positive regulation 10 4 0.93 0.00989 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 38 of gonad development GO:0044752 response to human chorionic gonadotropin 12 4 1.12 0.02005 836 837 Table 2. Means and ranges of RNA-seq deconvolution based- estimates of cell type 838 proportion within CIVMs of neonatal testis from PND 5 and PND 10 tissue. 839 CIVM Developmental Stage Cell Type Mean Proportion Estimate Range of Proportion Estimates PND 5 Germ cells 0.091 0.011 – 0.271 Leydig/Stroma cells 0.116 0 – 0.238 Peritubular cells 0.309 0.157 – 0.493 Sertoli cells 0.484 0.324 – 0.818 PND 10 Germ cells 0.083 0.007 – 0.249 Leydig/Stroma cells 0.097 0 – 0.200 Peritubular cells 0.330 0.145 – 0.568 Sertoli cells 0.490 0.334 – 0.776 840 841 842 843 844 845 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 39 Supplemental Information: 846 847 Supplemental figures: 848 849 Supplemental Figure 1. Immunocytochemistry of STRA8 expression in PND 10 850 CIVMs after 48 hours of exposure to 30nM isotretinoin. 851 Widefield scan of STRA8 expression across an example CIVM derived from PND 10 852 testis. (A) Expression of STRA8 (yellow) and nuclei (blue) in a CIVM of neonatal testis 853 after 48 hours of exposure to 30nM isotretinoin. (A-inset) Zoomed-in view of (A) to 854 highlight the nuclear localization of STRA8. (B) Expression of STRA8 (yellow) and 855 nuclei (blue) in a CIVM of neonatal testis in control group. 856 857 Supplemental Figure 2. Whole-mount immunohistochemistry of PND 5 and PND 858 10 Sprague Dawley rat seminiferous tubules. 859 Confocal images reduced to single image (using z-stack average, FIJI). Sertoli cell 860 nuclei are marked with a SOX9 primary antibody (red), spermatogonia are marked with 861 a cytoplasmic DDX4 primary antibody (green). (A) Single PND 5 seminiferous tubule 862 whole mount image. (B) Single PND 10 seminiferous tubule whole mount image. 863 864 Supplemental Figure 3. Widefield fluorescent immunocytochemistry of PND 5 and 865 PND 10 CIVMs. 866 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 40 Widefield scans showing spermatogonia marked with a cytoplasmic DDX4 primary 867 antibody (green). Nuclei are marked with Hoechst 33342 (blue). (A) Scan of example 868 PND 5 control culture. (B) Inset of (A). (C) Scan of example PND 10 control culture. (D) 869 Inset of (C). 870 871 Supplemental Figure 4. Log2 fold change of markers of spermatogonial 872 phenotype after exposure to isotretinoin in vitro. 873 CIVMs derived from neonatal testis from PND 10 testis were exposure to isotretinoin for 874 24 hours. (A) Expression of Stra8 in PND 10 derived CIVMs determined by RT-qPCR. 875 Aside from one control sample, all samples were above the threshold of quantification. 876 The expression of Stra8 was significantly increased in the 3,30, and 300 nM isotretinoin 877 dose groups (p-values all <0.0001). (B) Expression of Plzf in PND 10 derived CIVMs 878 determined by RT-qPCR. All samples were above the threshold of quantification. The 879 expression of Plzf in the 30 and 300nM isotretinoin dose groups were significantly lower 880 than control (p-values <0.0001). Same data as presented in Figure 3, but shown as 881 Log2 Fold Change for comparability to other log2 fold change data presented. 882 883 Supplemental Figure 5. Fluorescent immunocytochemistry of example PND 5 and 884 PND 10 CIVMs after 15 days in vitro. 885 Fluorescent microscopy images showing spermatogonia marked with a cytoplasmic 886 DDX4 primary antibody (green) and Sertoli cell nuclei marked with SOX9 (red). Both are 887 examples maintained in AlbuMAX for 15 days. Not representative of quantitative 888 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 41 changes with treatment, rather included to support maintenance of spermatogonia 889 through 15 days in vitro. (A) Example PND 5 AlbuMAX CIVM after 15 days. (B) 890 Example PND 10 AlbuMAX CIVM after 15 days. 891 892 Supplemental Figure 6. Complete results of transcription factor activity 893 inferenced from bulk RNA-seq data using decoupleR. 894 Extension of Figure 1.D, showing all transcription factors with significant activity inferred 895 based on differentially expressed gene from RNA-seq edgeR analysis. Significance 896 determined based on p-value <0.05 during modeling permutation. Analyses done in R 897 software using the decoupleR package. 898 899 Supplemental Figure 7. Retinol metabolism and inhibition within PND 10 and PND 900 5 testis CIVMs. 901 Extension of Figure 4.8 to include PND 5 gene expression as well. Expression of key 902 retinoid metabolism genes in RNA-seq analysis of PND 10 derived testis CIVMs. 903 Includes expression of retinol transporter Stra6; metabolic enzymes Rdh10, Aldh1a1, 904 and Cyp26b1; retinoid nuclear receptors and co-receptors Rxr (alpha and beta 905 isoforms) and Rar (alpha and gamma isoforms); and transport proteins Rbp4, Crabp1, 906 and Crabp2. Dashed line marks a log2 CPM of 1 as a gene expression threshold. 907 908 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 42 Supplemental Figure 8. Differentially expressed genes within retinoic acid 909 metabolism related pathways. 910 Extension of Table 1 to show the expression of differentially expressed genes within 911 the three significant gene ontologies related to retinoic acid metabolism or response 912 (GO:0042573, GO:0034698, and GO:0001523). 913 914 Supplemental Tables: 915 Supplemental Table 1 916 CIVM Developmental Stage

Cell Type Mean Proportion Estimate Minimum proportion estimate Maximum proportion estimate PND 5 OLS deconvolution using ALRA imputed

Germ cells 0.121 0.031 0.271 Leydig/Stroma cells 0.106 0.000 0.206 Peritubular cells (PTM) 0.341 0.157 0.493 Sertoli cells 0.432 0.373 0.480 DWLS deconvolution using ALRA imputed

Germ cells 0.088 0.060 0.118 Leydig/Stroma cells 0.171 0.048 0.238 Peritubular cells (PTM) 0.281 0.210 0.317 Sertoli cells 0.460 0.377 0.594 DWLS deconvolution using non- imputed

Germ cells 0.065 0.011 0.194 Leydig/Stroma cells 0.070 0.000 0.147 Peritubular cells (PTM) 0.306 0.169 0.436 Sertoli cells 0.559 0.324 0.818 PND-10 OLS deconvolution using ALRA Germ cells 0.108 0.007 0.249 Leydig/Stroma 0.089 0.000 0.186 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint 43 imputed

cells Peritubular cells (PTM) 0.393 0.145 0.568 Sertoli cells 0.410 0.334 0.541 DWLS deconvolution using ALRA imputed

Germ cells 0.079 0.038 0.104 Leydig/Stroma cells 0.142 0.020 0.200 Peritubular cells (PTM) 0.315 0.248 0.393 Sertoli cells 0.464 0.338 0.623 DWLS deconvolution using non- imputed

Germ cells 0.063 0.016 0.185 Leydig/Stroma cells 0.059 0.032 0.083 Peritubular cells (PTM) 0.282 0.175 0.347 Sertoli cells 0.595 0.427 0.776 917 918 Supplemental Table 2 919 Name GSE Species Age Reference Infant-120 GSE120506 Human 2 donors (13 months old) [51] Infant-124 GSE124263 Human 2 neonatal donors (2 days old, 7 days old) [52] Mouse - PND2 GSE130593 Mouse Post natal day 2 C57BL/6 mice (two technical replicates, each replicate is a combination of testis tissue from two mice) [53] Mouse - PND7 GSE130593 Mouse Post natal day 7 C57BL/6 mice (two technical replicates, each replicate is a combination of testis tissue from two mice) [53] 920 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint Figure 1. Transcriptomic Comparison of CIVMs derived from PND 5 and PND 10 after three days in vitro. A. C. B. D. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint Figure 2. RNA-seq Deconvolution based- estimates of cell type proportion within CIVMs of neonatal testis from PND 5 and PND 10 tissue. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint Figure 3. Expression of markers of spermatogonial phenotype after exposure to isotretinoin in vitro. C. Expression of Stra8 in PND 10 derived CIVMs D. Expression of Plzf in PND 10 derived CIVMs A. Expression of Stra8 in PND 5 derived CIVMs B. Expression of Plzf in PND 5 derived CIVMs .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint Figure 4. A. Retinol metabolism and inhibition with neonatal testis CIVM B. .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint Figure 5. A. Plzf B. Stra8 .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint Figure 6. A. PND 5 KSR B. PND 5 AlbuMAX C. PND 10 KSR D. PND 10 AlbuMAX E. PND 5 SSC GFs F. PND 5 control G. PND 10 SSC GFs H. PND 10 control .CC-BY-NC 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted December 31, 2025. ; https://doi.org/10.64898/2025.12.30.696417doi: bioRxiv preprint