Throughout the male reproductive lifespan, spermatogonial stem cells (SSCs) produce committed progenitors that proliferate and then remain physically connected in growing clones via short cylindrical intercellular bridges (ICBs). These ICBs, which enlarge in meiotic spermatocytes, have been demonstrated to provide a conduit for postmeiotic haploid spermatids to share sex chromosome-derived gene products. In addition to ICBs, spermatogonia exhibit multiple thin cytoplasmic projections. Here, we have explored the nature of these projections in mice and find that they are dynamic, span considerable distances from their cell body (≥25 μm), either terminate or physically connect multiple adjacent spermatogonia, and allow for sharing of macromolecules. Our results extend the current model that subsets of spermatogonia exist as isolated cells or clones, and support a model in which spermatogonia of similar developmental fates are functionally connected through a shared dynamic cytoplasm mediated by thin cytoplasmic projections.
The foundation of mammalian spermatogenesis is provided by continuous action of SSCs, which undergo a fate decision to either renew their population or produce transit-amplifying progenitor spermatogonia that proliferate before differentiating in response to retinoic acid (RA) and eventually entering meiosis as spermatocytes. The most widely accepted kinetic model of SSC self-renewal and differentiation in mammalian testes was first proposed in 1971, and links cell fate to spermatogonial clone length (Huckins, 1971; Oakberg, 1971). As isolated Asingle (As) SSCs divide, they can produce two As spermatogonia or become linked via an ICB and form a clone (or chain) of Apaired (Apr) spermatogonia. These Apr spermatogonia proliferate to form longer clones of Aaligned (Aal) transit-amplifying progenitors that are increasingly committed to differentiation. These original morphology-based predictions of cellular fate have been confirmed by recent fate marker, regeneration and transplantation analyses (Chan et al., 2014; Grisanti et al., 2009; Hara et al., 2014; Helsel et al., 2017).
ICBs presumably form after incomplete cytokinesis in germ cells expressing TEX14, which interacts with ‘centrosomal protein 55’ (CEP55) and blocks cell abscission (Greenbaum et al., 2006; Iwamori et al., 2010; Wu et al., 2003). ICBs are retained throughout the remainder of spermatogenesis, and grow wider in meiotic spermatocytes and post-meiotic spermatids to form an open ring-shaped TEX14+ structure (Greenbaum et al., 2006). ICBs permit passage of X-linked macromolecules (mRNAs and proteins) between X- and Y-chromosome-bearing haploid spermatids (Braun et al., 1989; Morales et al., 2002; Ventelä et al., 2003). However, sharing of macromolecules via ICBs has not been demonstrated in spermatogonia or spermatocytes.
In addition to ICBs, scarce classical and contemporary evidence reveals mammalian spermatogonia may have greater degrees of interconnectedness. Indeed, spermatogonia are not round or ovoid, but exhibit multiple finger-like projections (Fig. 1A; Sertoli, 1877). These projections appear longer and thinner than short cylindrical ICBs, which are 1-1.5 μm wide×0.5-1 μm long (Dym and Fawcett, 1971; Weber and Russell, 1987). In recent reports, these projections are visible following immunostaining for membrane-associated proteins, although they are only mentioned briefly in one study (Abid et al., 2014; Gassei and Orwig, 2013; Grisanti et al., 2009; Nakagawa et al., 2010; Niedenberger et al., 2015; Suzuki et al., 2009; Tokuda et al., 2007).
Recent reports reveal that spermatogonia also have cellular projections in lower organisms. In locusts, male germline stem cells (GSCs) use filopodia to interact with somatic apical cells (Dorn and Dorn, 2011). In Drosophila testes, short microtubule-based nanotubes extend from the GSCs into the hub to facilitate ligand-receptor interactions between GSCs and terminally differentiated somatic supportive hub cells and cyst stem cells (Lin, 2002).
In this study, we provide the first detailed examination of mammalian spermatogonial cytoplasmic projections. Using a variety of approaches, we find that these projections are distinct from ICBs. Observed in spermatogonia and their precursors (prospermatogonia/gonocytes), these projections are thin dynamic cytoplasmic elements that either terminate or connect multiple adjacent and distant spermatogonia. We found that both spermatogonial cytoplasmic projections and ICBs allow the transfer of proteins between interconnected As spermatogonia. Taken together, these results provide clear evidence for intercellular communication between developing spermatogonia in vivo.
RESULTS AND DISCUSSION
Cytoplasmic projections are distinct from ICBs
Spermatogonial cytoplasmic projections were first described by Enrico Sertoli. We recently translated this work into English for the first time (Sertoli, 2018); Sertoli noted ‘interesting [in spermatogonia] is the presence of the projections, which give the cells their characteristic star-shaped form’ (Fig. 1A) (Sertoli, 1877). These projections are visible after immunostaining whole-mounted testis cords or tubules with antibodies against membrane-associated proteins (Abid et al., 2014; Gassei and Orwig, 2013; Grisanti et al., 2009; Nakagawa et al., 2010; Niedenberger et al., 2015; Suzuki et al., 2009; Tokuda et al., 2007), although they have never been characterized.
For facile visualization of these structures, we used transgenic mice expressing enhanced green fluorescent protein (EGFP) in spermatogonia (Id4-eGfp; Chan et al., 2014). EGFP epifluorescence intensity is linked to spermatogonial fate in developing testes: ID4-EGFPbright spermatogonia represent SSCs, whereas ID4-EGFPdim and ID4-EGFP– populations are progenitor and differentiating spermatogonia, respectively (Chan et al., 2014; Helsel et al., 2017).
We first examined the extensive network of cytoplasmic projections in live ID4-EGFP+ spermatogonia; numerous thin projections of varying lengths were straight, branched and/or had enlargements (Fig. 1B). Projections were preserved in both ID4-EGFPdim and ID4-EGFPbright fixed cells, and appeared to connect spermatogonia up to 30 μm apart (Fig. 1C,D). Projections were readily discerned by electron microscopy (EM), coursing between adjacent Sertoli cells (Fig. 1E). Projections were present in nearly all ID4-EGFP+ spermatogonia, and 58% appeared to interconnect adjacent spermatogonia. Moreover, nearly all apparently interconnected spermatogonia exhibited similar ID4-EGFP intensity (89% appeared to connect bright-bright or dim-dim, and the remaining 11% were not apparently connected, Fig. 1F), suggesting a common fate (Chan et al., 2014; Helsel et al., 2017). We expect to have underestimated the frequency and numbers of projections; those extending towards or away from the visualized plane went undetected. In addition, we likely underestimated the lengths of projections at oblique angles to the visualized plane.
Spermatogonial projections appeared too long and narrow to represent short cylindrical ICBs (Weber and Russell, 1987). We employed three approaches to discriminate between cytoplasmic projections and bona fide ICBs. First, we stained intact cords from Id4-eGfp testes with an antibody recognizing TEX14, an integral component of all germ cell ICBs (Greenbaum et al., 2006; Iwamori et al., 2010). The vast majority (86%) of all cytoplasmic projections over 4 μm long were TEX14−, indicating longer projections were not ICBs, even in KIT+ differentiating spermatogonia (Fig. 1G, Fig. S1). Second, we averaged dimensions of TEX14+ ICBs and TEX14− cytoplasmic projections. Although sizes varied somewhat, TEX14+ connections were significantly (P<0.01) shorter and wider (2.8 μm×1.2 μm) than TEX14− projections (8.1 μm×0.68 μm). Third, we compared cytoplasmic projections in intact testis cords from wild-type and Tex14 KO mice by immunostaining for CDH1, which marks most spermatogonia at this stage (Niedenberger et al., 2015; Tokuda et al., 2007). Tex14 KO mice lack ICBs (Greenbaum et al., 2006; Iwamori et al., 2010; Kim et al., 2015), yet exhibited multiple projections appearing to connect adjacent spermatogonia, as in wild-type littermates (Fig. 1H,I). Spermatogonia in Tex14 KO testes were more likely to be apparently connected by projections (72.5% in Tex14 KO, 57.9% in wild type, P<0.05). The average number of projections per cell did not differ between genotypes (3.4 in Tex14 KO, 2.8 in wild type, P=0.22). Taken together, these data reveal most cytoplasmic projections, especially long thin ones, were not TEX14+ ICBs. These results highlight the necessity for researchers to determine spermatogonial clone length by staining with TEX14, rather than making assumptions based on proximity alone.
Spermatogonial projections are present in precursor prospermatogonia, in both undifferentiated and differentiating spermatogonia, but not in meiotic spermatocytes
We next assessed the temporal appearance of projections during spermatogenesis. We first stained neonatal quiescent precursor prospermatogonia (Vergouwen et al., 1991; Western et al., 2008) for GFRA1, which is expressed in all prospermatogonia at this stage (Niedenberger et al., 2015). Projections extended from, and apparently connected, adjacent postnatal day (P)1 prospermatogonia (Fig. 2A). These appeared shorter and fewer than those on spermatogonia at P6 that were undifferentiated (GFRA1+, Fig. 2B,D) or differentiating (KIT+, Fig. 2C,E). The apparent connectedness and numbers of projections were similar in both GFRA1+ and KIT+ spermatogonia. As models of primate SSC renewal and differentiation differ from rodents (Fayomi and Orwig, 2018), we tested whether these spermatogonial projections were also present in primates. We stained intact testis cords (containing only spermatogonia) from a juvenile baboon using anti-GFRA1 and observed projections emanating from spermatogonia (Fig. 2F). In mice, projections were not observed after the spermatogonial stage in meiotic HIST1H1T+ spermatocytes (Inselman et al., 2003), which are also EGFP+ in Id4-eGfp mice (Fig. 2G,H).
A tenet of the As stem cell model is that diminution of stem cell capacity occurs with increasing length of ICB-connected spermatogonial clones (Huckins, 1971; Oakberg, 1971). As ICBs theoretically provide cytoplasmic continuity within a clone, it has been suggested that they coordinate clonal spermatogonial fate via sharing macromolecules such as RNAs and proteins. We examined this concept in Tex14 KO mice (Fig. S2) and found that absence of ICBs only slightly altered ratios of spermatogonia that were undifferentiated (GFRA1+, from ∼12% in wild type to ∼9% in KO) but not differentiating (KIT+/STRA8+). We speculate that, even in the absence of ICBs in Tex14 KO mice, cytoplasmic projections coordinate proper spermatogonial proliferation and differentiation. This would explain the earlier observation that spermatogenesis in Tex14 KO testes did not arrest until meiosis (Greenbaum et al., 2006; Sironen et al., 2011). As spermatocytes lack these projections (Fig. 2G,H), they presumably lack a means of sharing cytoplasmic contents.
Cytoplasmic projections and ICBs provide a physical connection between adjacent spermatogonia for rapid diffusion of macromolecules (such as EGFP)
We next assessed whether these thin projections could mediate cytoplasmic exchange between adjacent spermatogonia. We first used EM on over 200 serial sections from P6 testes and observed, at high resolution, apparently continuous cytoplasm connecting neighboring spermatogonia via ICBs (Fig. 3A) and cytoplasmic projections (Fig. 3B). We next used fluorescence recovery after photobleaching (FRAP) to definitively determine whether cytoplasmic projections and ICBs represented patent connections between adjacent live spermatogonia. We photobleached ID4-EGFP+ spermatogonia (to 10% original fluorescence) in various configurations (As, Apr and As apparently connected by projections) and quantified fluorescence recovery over a short interval (<12 min, Fig. 3C). Although recovery could theoretically result from synthesis and accumulation of nascent EGFP molecules, this generally takes several hours (Kourtis and Tavernarakis, 2009). We first photobleached As ID4-EGFPbright spermatogonia exhibiting no apparent connections; they did not regain fluorescence (Fig. 3D-G,P), revealing that was indeed too short an interval for accumulation of detectable EGFP. We next photobleached individual ID4-EGFPbright spermatogonia that were apparently connected to adjacent ID4-EGFPbright spermatogonia via cytoplasmic projections. These As spermatogonia (not ICB-connected) rapidly recovered ∼40% of their original fluorescence, while their non-photobleached partners lost ∼40% of their original fluorescence (Fig. 3H-K,P). Finally, we photobleached ID4-EGFPbright spermatogonia in clear Apr configurations (connected by ICB) and found recovery was rapid (≤10 min), while the non-photobleached partner became ∼50% less intense (Fig. 3L-P). The linear kinetics of the redistribution of EGFP over this brief interval supported rapid diffusion of EGFP molecules between cells. The percentage of EGFP shared between cells formed the ‘mobile fraction’ (Fig. 3Q). This is the first report, to our knowledge, documenting passage of macromolecules (here, EGFP, MW ∼27 kDa) between adjacent spermatogonia.
Cytoplasmic projections are dynamic structures connecting related spermatogonia
Our current understanding of spermatogonial behavior during development is largely based on static observations using fixed tissues and isolated macromolecules. Here, we performed time-lapse imaging of live testis cords from Id4-eGfp mice maintained in situ to determine whether spermatogonial cytoplasmic projections represented static (stable) or dynamic (transient) structures. To distinguish between these possibilities, images were captured every 5 min from 15 different sites over extended incubations and combined into movie clips (see Movie 1, still images in Fig. S3). ID4-EGFPdim and ID4-EGFPbright spermatogonia repeatedly extended and retracted multiple projections over relatively short time periods (∼30 min) over the ∼14 h imaging period. In Movie 1, two ID4-EGFPbright spermatogonia extended projections to the cord periphery (∼0-17 s, Fig. S3A), became rounded and retracted their projections (∼18-20 s, Fig. S3B), and then nearly simultaneously underwent mitosis (∼20-22 s, Fig. S3C). The newly formed spermatogonial pairs appeared to separate but then interact multiple times over the remainder of the video (∼23-41 s, Fig. S3D-F). The top pair remained ID4-EGFPbright, while the bottom pair became ID4-EGFPdim, suggesting they adopted separate fates (SSC and progenitor/differentiating, respectively). In addition, numerous spermatogonia moved in and out of the viewing plane, suggesting movement along the testis cord.
Finally, we examined whether projections formed between related progeny following division or between unrelated cells. We employed Brainbow R26R-Confetti transgenic mice (JAX #013731). In this model, Cre-mediated deletions or inversions in the Brainbow transgene activate nuclear GFP (green), cytoplasmic RFP (red) or YFP (yellow), or membrane-tethered CFP (blue) (Fig. S4; Cai et al., 2013; Livet et al., 2007). We crossed Brainbow with Ddx4-Cre transgenic mice, which begin to express Cre recombinase in the male germline around E15 (Gallardo et al., 2007). Continuous Cre activity would cause a predicted evolution of deletions and inversions in the Brainbow transgene, creating spermatogonia singly expressing GFP, RFP, YFP or CFP or combined GFP/YFP or RFP/CFP (the latter two from repeated inversions, see Fig. S4). We found spermatogonia were either YFP+, RFP+, CFP+ or RFP+/CFP+, but never GFP+. According to the donating investigator (H. Clevers, Hubrecht Institute, Utrecht, the Netherlands www.jax.org), weaker Cre expression was correlated with fewer GFP+ cells. Based on potential combinations from continuous inversions (Fig. S4), YFP+ spermatogonia arose from different precursors than RFP+, CFP+ or RFP+/CFP+ spermatogonia. YFP+/RFP+ or YFP+/CFP+ spermatogonia would indicate sharing of fluorescent protein across projections between unrelated spermatogonia, which we never observed (Fig. 4). Therefore, we conclude that projections physically connect spermatogonia arising from common precursors. These observations resemble those from the Capel lab; only fetal prospermatogonia from wild-type×GFP chimeric mice of similar origin (WT/WT or GFP+/GFP+) were observed to interact via ICBs (Mork et al., 2012).
Here, we report that mammalian spermatogonia are connected by an extensive network of fine cytoplasmic projections in vivo. These projections are dynamic and transient, and provide patent connections between adjacent spermatogonia that, in addition to ICBs, allow for passage of macromolecules (here, EGFP) between spermatogonia of similar fates. The full functionality of these cytoplasmic projections will likely be elucidated in future studies, but we can certainly speculate on their potential in vivo role(s) based on known spermatogonial behavior. First, some may be precursors to ICB formation; it has been assumed, but not shown, that they form due to incomplete cytokinesis. Second, it is clear that spermatogonia move within mammalian seminiferous tubules (Hara et al., 2014; Yoshida et al., 2007), and these dynamic projections may function as filopodia to direct this movement. Third, as each spermatogonium is surrounded by numerous somatic cells, it is likely that these projections allow for ligand-receptor interactions in a similar manner to those described in Drosophila testes, in which MT-nanotubes relayed BMP signaling exchanges between GSCs and the nearby hub (Lin, 2002). Indeed, mammalian spermatogonial fate is regulated in large part by ligand-receptor interactions, including GDNF-GFRA1/RET, FGF-FGFR, RA-RAR/RXR and KITL-KIT (reviewed by Busada and Geyer, 2015; Mark et al., 2015; Oatley and Brinster, 2012; Yang and Oatley, 2014). Undifferentiated and differentiating spermatogonia generally express differing levels of these receptors, which poises them to respond to localized signals that direct or maintain their fate. We predict that further investigation into the specific roles played by these projections will significantly enhance our understanding of mammalian spermatogonial biology.
MATERIALS AND METHODS
Animal procedures were approved by the Animal Care and Use Committees of East Carolina University (Assurance #A3469-01) and Baylor College of Medicine (Assurance #A3823-01). Testes were obtained by necropsy from a 29-month-old baboon at Southwest National Primate Research Center via Texas Biomedical Research Institute (Assurance #A3082-01). All procedures followed guidelines outlined in the National Research Council Guide for the Care and Use of Laboratory Animals.
Whole-mount indirect immunofluorescence
Testes were dissected and fixed in 4% paraformaldehyde in 1×PBS at 4°C for 2 h to overnight, depending on age. After an overnight soak in 1×PBS at 4°C, testes were cut into ∼1 mm3 pieces, permeabilized by incubation in 0.1% Triton X-100 for 30 min at room temperature, and then blocked for 1 h at room temperature with 3% BSA in 0.1% Triton X-100 in 1×PBS. Primary antibodies were diluted with 3% BSA in 0.1% Triton X-100 in 1×PBS and incubated with tissues overnight at 4°C. Tissues were then washed three times for 30 min each in 1×PBS containing 0.1% Triton X-100. Fluorescently conjugated secondary antibodies and phalloidin (to visualize F-actin) were diluted with 3% BSA in 0.1% Triton X-100 in 1×PBS, then incubated with tissue overnight at 4°C. Tissues were then washed three times for 30 min each in 0.1% Triton X-100 in 1×PBS. Tissues were mounted in Fluoroshield with DAPI (Abcam) and imaged on an Olympus FV1000 confocal microscope. Multiple z-stacked images were acquired (10 when using the 60× objective, five when using the 60× objective with 2× digital zoom) that were 2 μm apart, and merged in ImageJ to generate maximum intensity projections. Details for primary antibodies and labeling reagents are provided in Table S1.
Images used for quantitation were five-slice maximum intensity projections taken with the 60× objective with 2× digital zoom. Nine images were taken totaling 79 projections from n=4 mice. These projections were measured in Adobe Photoshop using an image-calibrated ruler tool. Between 6-13 projections were measured from each image.
Images of whole-mount testis cords and tubules were captured using an Olympus FluoView FV1000 confocal laser scanning microscope or a Zeiss LSM 880 with Airyscan. The determination of cells staining positive or negative for markers such as KIT and GFRA1 were performed using ImageJ software, with thresholds set at 195-255 for ID4+ cells and 40-200 for ID4− cells. Nonspecific staining was removed from images of TEX14-stained sections in ImageJ with the Particle Remover tool with the threshold set to 55-255. Size was set to 0.0-0.7 μm2 and holes were included. The Subtract Background tool was used to reduce background with rolling ball radii set to 10 pixels.
Two transgenic mice were used in the Brainbow experiments, from different litters. PFA-fixed testes were cut into ∼2 mm3 pieces, mounted on a slide with Vectastain containing DAPI, and imaged using a 40× objective. Images consisted of two stacks with 43-68 1 μm slices. Each slice was carefully examined for connected cells to determine whether spermatogonia expressing different fluorophore combinations were connected by projections. Z-stack images were captured on the Olympus FV-1000 as well as the Zeiss LSM 880 with Airyscan. Time-lapse experiments were repeated 15 times (15 movies were recorded) in two separate experiments using testes from eight different Id4-eGfp mice. FRAP experiments were repeated five times for each type of cell type (As, ICB-connected Apr and projection-connected) on at least 10 mice.
Serial electron microscopy
Whole testes were incubated in 2% glutaraldehyde for 1 h at 4°C. Epididymides were removed and testes cut into several pieces. Fixation was carried out in 2% glutaraldehyde overnight at 4°C, followed by three successive washes in 0.1 M cacodylate buffer. For wild-type testes, post-fixation was performed in 1% osmium tetroxide at room temperature for 1 h, followed by three 15 min washes with sodium phosphate buffer. Testes were then dehydrated through a graded ethanol series (25%, 50%, 70%, 95%, 100%) at room temperature. Following fixation, testes were embedded in Spurr's medium (Electron Microscopy Sciences) in flat tissue tear-away containers that were baked overnight at 70°C. Serial sections were captured using a tape-collecting device coupled with an ultramicrotome (Kasthuri et al., 2015). A consecutive series of 237 sections was cut at a thickness of 40 nm and collected on Kapton tape. The sections were imaged using the backscatter mode of a Zeiss Sigma field emission scanning electron microscope as previously described (Terasaki et al., 2013). A 61 μm2 area was imaged at a resolution of 5 nm/pixel (12,233×12,233 pixels). Images were aligned using the Linear Stack Alignment with SIFT algorithm in ImageJ.
Fluorescence recovery after photobleaching
Testes were dissected from P6 Id4-eGfp mice, detunicated and transferred to a glass-bottomed Petri dish. Testis cords were gently teased apart using a pair of 30-gauge needles, then covered with a slab of agarose dissolved in Fluorobrite DMEM supplemented with 4 mM L-glutamine (Thermo Fisher Scientific). All imaging was performed at 37°C with 5% CO2 in a humid chamber on an Olympus FV1000 confocal microscope equipped with Olympus Fluoview software. FRAP imaging was performed with the following parameters: (z-stack, four slices of 2 μm each), 2% laser intensity, 800×800 pixels, 4 μs/pixel scan speed and a fully-open aperture. Photobleaching was carried out with 100% laser intensity at 100 μs/pixel. Specimens were imaged 80 and 40 s prior to photobleaching, and then a total of 25 times every 40 s after photobleaching.
Measurements of fluorescence intensity were carried out in ImageJ with Time Series Analyzer V3.0. The intensity was normalized at each timepoint using Microsoft Excel with the following steps: (1) we subtracted the average intensity of the background (a dark area with no visible cells) from the average intensity from a non-photobleached cell; (2) the resulting value for each time point was then divided by the resulting value for the initial timepoint; (3) the average background intensity over the course of the experiment was then subtracted from the average intensity of a frapped cell for each timepoint; (4) the resulting number [calculated in (3)] was then divided by the number calculated in (2) and (5). The result of (4) for each timepoint was divided by (4) at the initial timepoint, then multiplied by 100 to calculate percentage change in average intensity, which was then analyzed using Sigma Plot 13.0.
Time-lapse live imaging
Tissues were mounted as above for FRAP, and imaged at 37°C with 5% CO2 in a humid chamber on a Zeiss LSM 700 laser scanning confocal microscope equipped with Zeiss Zen Black software using the 40× objective. Five z-stacked images were acquired with spacing of 2 μm every 5 min for a total of 14 h. This experiment was performed twice, and a total of 15 different sites were imaged. Maximum intensity projections were generated using Zen Black, and sequential images were combined to produce a short video clip.
Statistical analyses were carried out using Student's t-test, and significance was set at P<0.05. Statistical analyses of FRAP data was performed using Sigma Plot 13.0, which determined that data assumed a normalized distribution (Shapiro-Wilk, P=0.372) and equal variance (Brown-Forsythe, P=0.282). Statistical significance was determined using a two-tailed t-test.
We thank Dr Jon Oatley (Washington State University, Pullman, WA, USA) for Id4-eGfp mice, Dr Eva Johannes (North Carolina State University, Raleigh, NC, USA) for technical assistance with microscopy, and Ms Katya Harris (graduate student, East Carolina University School of Art and Design, Greenville, NC, USA) for illustrations.
Conceptualization: N.D.S., K.A.L., M.T., B.P.H., C.B.G.; Methodology: B.A.N., E.K.V., J.E.A., K.A.L., M.T., C.B.G.; Software: C.B.G.; Validation: C.B.G.; Formal analysis: B.A.N., E.K.V., K.A.L., C.B.G.; Investigation: B.A.N., K.C., V.B., N.D.S., C.B.G.; Resources: J.E.A., M.T., B.P.H., M.M.M., C.B.G.; Data curation: C.B.G.; Writing - original draft: B.A.N., C.B.G.; Writing - review & editing: M.T., B.P.H., M.M.M., C.B.G.; Visualization: C.B.G.; Supervision: C.B.G.; Project administration: C.B.G.; Funding acquisition: B.P.H., M.M.M., C.B.G.
Research was supported by grants from the Eunice Kennedy Shriver National Institute of Child Health and Human Development (1R01HD090083 to C.B.G. and P01HD087157 to M.M.M.). This investigation used resources supported by the Southwest National Primate Research Center grant (P51 OD011133) from the National Institutes of Health's Office of Research Infrastructure Programs. Deposited in PMC for release after 12 months.
The authors declare no competing or financial interests.