Adult human 46, XY dermal fibroblast to Sertoli cell trans-differentiation strategy produces Sertoli like cells (SLC) that exhibit a significant change in morphology

We utilized the predictive software Mogrify, to predict the TFs necessary to direct differentiation of fibroblasts into Sertoli cells. Mogrify is a comprehensive atlas of human cell conversions that utilizes gene expression data with regulatory network information from ~ 300 different cell and tissue FANTOM5 data sets to compute a ranked list of transcription factors that cumulatively exert up to 98% differential regulatory influence on the target cell type in comparison to starting cell type [37]. Mogrify predicted that transgenic expression of eight TFs, FGF2, GATA6, GATA4, MXI1, JUNB, NFYB, NR5A1 and EBF1, was required for human dermal fibroblast to Sertoli cell trans-differentiation. It also predicted that just six out of those eight, excluding NR5A1 and EBF1, were enough to achieve a cumulative 95% Sertoli influence on dermal fibroblasts (Fig. 1A). Fgf2 and Gata4 are differentially expressed between male and female gonads of mice at developmentally important stages, as revealed by RNA-seq analyses [44, 45]. NR5A1 is a well-known human DSD gene [46].

Fig. 1

46,XY fibroblast derived SLCs exhibit significant change in morphology over the 1-month trans-differentiation and show subcellular expression of Sertoli markers. A Mogrify-predicted transcription factors required for dermal fibroblasts to Sertoli cell trans-differentiation, # = not required to reach 95% coverage of the required genes. B Outline of the trans-differentiation strategy employed throughout this report. C Representative images of live cells belonging to the indicated groups expressing GFP across the 1-month trans-differentiation culture of 46, XY dermal fibroblasts. Morphometric analysis of cells in (C) showing shape factor (D) and area quantification (E) (N = 3, n = 50–60 for each group). *Represent p values calculated from Mann Whitney statistical tests conducted between the indicated groups

In a pilot study, we tested a strategy for trans-differentiation of 46, XY normal adult human dermal fibroblasts using 16 different conditions (Additional file 1: Fig. S1A) stably expressing 8 or 6 TFs (and GFP) at different multiplicity of infections with different transduction and trans-differentiation media. The resulting cells were imaged and harvested to measure expression of Sertoli markers by qPCR. Considering optimal transduction efficiency, cell survival/proliferation, morphological appearance, and expression of Sertoli markers SOX9, PGDS and BMP4, conditions 6 and 5 were chosen to be pursued (and will be called “6TF” and “8TF” respectively below) (Fig. 1B).

One of the defining features of cells is their morphology. Of the two cell types relevant to this study, fibroblasts exhibit an elongated spindle-shaped morphology, while Sertoli cells are known to assume squamous epithelial morphology. Sertoli cells grown in adherent cultures have been reported to appear large, have a polygonal shape and exhibit long cytoplasmic extensions or filamentous processes of irregular lengths radiating out of the cell body, along with irregularly shaped nucleus [47,48,49,50]. Cells were imaged weekly under green fluorescence for the duration of the one-month culture (Fig. 1C). The size and shape of cells were investigated by morphometric analysis of cell images with recording of the surface area and perimeter of the individual cells.

Shape factor measurement revealed that while there was no significant difference in the shape between GFP transduced control and 6TF, 8TF SLCs in the first week, the shape of 6TF and 8TF SLCs had diverged significantly from GFP control starting at the 2nd week (Fig. 1D). While the GFP control cells remained elongated in shape, characteristic of fibroblasts, 8TF cells started to assume more polygonal appearance with elongated and irregular cytoplasmic processes, typical of Sertoli cells. These irregular processes were even more accentuated in 6TF cells, as observed visually and quantified by a lower shape value in 6TF SLCs than the more circular or polygonal 8TF SLCs (Fig. 1D). Additionally, the GFP control and 6TF, 8TF cells appeared to be similar in size in the first week but the experimental cells started flattening, spreading and growing in size in comparison to GFP control beginning in the second week and this difference became even more enhanced with each passing week (Fig. 1C, E). All the cells, including GFP controls exhibited significant growth in size at the end of 1 month in comparison to when they started (Fig. 1E). Together, these morphometric evaluations showed that the 1 month of trans-differentiation culture gradually pushed fibroblasts to change in size and shape and to start exhibiting some morphological features associated with Sertoli cells.

SLCs exhibit Sertoli molecular signatures

To investigate the Sertoli features of transdifferentiated cells further we compared their transcriptome with that of a published adult Sertoli cell (aSC) transcriptome [48], and curated lists of Sertoli marker genes (Additional file 1: Table S1).

Principal component analysis (PCA) of RNA seq transcriptomes revealed that replicates of 6TF SLC cluster away from GFPC, while 8TF did not (Fig. 2A). Replicates of each category of samples grouped together and away from samples of other category (Fig. 2A).

Fig. 2

46,XY SLCs show Sertoli marker gene expression: Whole genome transcriptional profiling of HDFa SLCs. A Principal Component Analysis of all the transcriptomes under consideration in this analysis. B Venn diagrams displaying differentially expressed genes (DEGs, fold change > 1.5, p adjusted < 0.05) between GFPC (N = 3) vs adult Sertoli cells (aSC, in brown) (N = 2) and GFPC (N = 3) vs 8TF (red) and C 6TF (blue) SLCs (N = 3). Intersection areas show co-differentially expressed genes (co-DEGs). D Heat map showing 6TF and 8TF vs. GFPC log2 fold change for indicated gonadal marker genes measured by RNAseq (n = 3). E–H Representative immuno-fluorescence images showing subcellular expression and quantification of percentage of cells showing such expression for 6TF SLCs, 8TF SLCs, and GFP controls (N = 3, n = 50–60 for each group) for (E) SOX9 (F) DMRT1 (G) AMH and (H) DHH

To normalize potential differences owing to their different origins, all three groups (6TF, 8TF and aSC) were first compared with 46,XY-derived GFP control (GFPC) to determine the set of differentially expressed genes (DEGs). Of the 1870 genes differentially expressed between 6TF and control GFPC, 2/3 were in common with the aSC DEGs (Fig. 2C). In 8TF cells, this percentage was almost 3/4 (502/681 genes) (Fig. 2B). Such high percentages of co-differentially expressed genes between the transcriptomes of SLCs and adult Sertoli cells indicated that our experimental SLCs had significant Sertoli-like gene expression patterns.

Gene ontology (GO) analysis of the co-DEGs determined above identified several terms expected to be associated with Sertoli cell as organizing center of male reproductive system structure and seminiferous tubule development (Additional file 1: Fig. S2C, D). The top twenty GO biological processes list included developmental process, organ development, urinogenital system development, tube morphogenesis, external encapsulating structure organization, ECM structure and matrix organization, tube development, with reproductive system development, and reproductive structure development further down the list. It also showed terms for positive and negative regulation of lipid biosynthetic and metabolic processes, congruent with the known lipid metabolism and lipid droplet formation in Sertoli cells [50, 51].

Next, we compared the gene expression profile of the 1253 co-DEG in 6TF SLC, GFPC and aSC in a heatmap (Additional file 1: Fig. S2A). [A similar heatmap looking at expression of the 502 8TF-GFPC co-DEG genes is shown in Additional file 1: Fig. S2B]. Each 6TF or 8TF SLC replicate appears quite different from GFPC and notably, resemble aSC expression to a large extent.

In order to investigate the molecular phenotype, a gene list of immature, mature, and all-stage Sertoli markers was drawn up from literature (Additional file 1: Table S1). The list also includes markers for non-Sertoli male gonadal cell types such as Leydig, myeloid, and germ cells, as well as for granulosa cells, the female homologue of Sertoli cells.

6TF SLC exhibited an increase in expression of bipotential gonad markers GADD45G, ZFPM; immature Sertoli markers KRT18, INHBA, NACAM1/2, TGFA; mature Sertoli markers IL1A, CX43/GJA1; and all-stage-Sertoli markers SOX9, KITLG, GDNF, BMP4 over GFPC (Fig. 2D, Additional file 1: Fig. S2E).

In contrast, myeloid (ACTA2) and granulosa (FOXL2) cell markers did not show an increase; Leydig cell markers CYP11A1 and STAR decreased (Fig. 2D). (Expression of germ cell marker DDX4 was undetected; data not shown). This indicates that the six transcription factors GATA4, GATA6, MXI1, JUNB, NFYB and FGF2 were sufficient to push dermal fibroblasts towards a Sertoli fate with predominant expression of immature and all-stage markers.

8TF SLC exhibited a significant increase in the expression of bipotential gonad marker GADD45G, immature Sertoli markers CYP26B1, PTGDS, KRT18, and NCAM1/2, and all-stage Sertoli markers SOX9 and BMP4 over GFPC (Additional file 1: Fig. S1E). 8TF SLC also exhibited a similar pattern for mature markers (CTSL, IL1A and CX43/GJA1) with lower fold increase than for immature markers. While 8TF SLCs also didn’t show an increase in expression of myeloid marker ACTA2 or granulosa marker FOXL2, they did exhibit a strong increase in expression of Leydig markers HSD3B2, CYP11A and STAR (Fig. 2D, Additional file 1: Fig. S1E). This suggests that 6TF is a better condition for producing SLCs as 8TF additionally produce Leydig (or adrenal) cell characteristics.

Expression of several markers was further validated by qPCR, which confirmed the RNASeq findings. All-stage Sertoli marker SOX9 and BMP4 showed 46-fold and 3.5-fold increases in 6TF SLCs over GFPC, respectively (Additional file 1: Fig. S1F). Immature marker PTGDS showed a twofold increase on linear scale. 6TF SLC showed no change in germ cell marker DDX4 or in granulosa marker FOXL2 and myeloid marker ACTA2. (Similar results were found for 8TF; Additional file 1: Fig. S1F).

Finally, the molecular phenotype was ascertained at the protein level. Sertoli cells express the transcription factors SOX9 and DMRT1 in the nucleus across all stages [52]. Sertoli cells also exhibit cytoplasmic expression of signaling molecules AMH during the immature stage and DHH in all stages (Additional file 1: Table S1). Immunofluorescent imaging revealed that all these four markers exhibited Sertoli-specific sub-cellular expression in 6TF SLCs as well as in 8TF SLC (Fig. 2E–H). The percentage of the cell population showing these appropriate expression patterns was higher in 6TF than in 8TF confirming that 6TF is likely the better transduction condition for Sertoli-like cells.

Trans-differentiation strategy applied to dermal fibroblasts of different 46, XY DSD genetic backgrounds and 46,XX produced SLCs that exhibit morphological changes

The method described here was able to trans-differentiate dermal fibroblasts derived from a 46, XY adult male into Sertoli-like cells. To test the robustness of the method, it was applied to commercially available dermal fibroblasts from individuals with various DS including:

46, XY SRXY1 (46,XY female “sex reversal” of unknown genetic etiology. No variant in known DSD genes was found by exome sequencing (not shown).

46,XY CD: Campomelic Dysplasia. A 3.2 Mb heterozygous deletion including SOX9 was identified by whole genome sequencing (Additional file 1: Fig. S5-S4),

46, XY WT1, where a typical large deletion including the WT1 gene was demonstrated by the company.

a 46,XX healthy female.

These four cell lines were separately seeded, transduced and cultured in the same way as described previously to produce 6TF and 8TF SLC and GFPC for each line. The live cultures were imaged on Day 7 and Day 28 of the culture under the GFP filter and the images were analyzed for changes in shape (shape factor) and size (area).

The 6TF SLCs for three 46, XY DSDs showed significant changes in shape (Fig. 3A–C) and size (Additional file 1: Fig. S3A–C) over the 1-month differentiation period. In contrast, the 46, XX 6TF SLC did not exhibit a significant shape change but did show a size increase like the others (Fig. 3D, Additional file 1: Fig. S3D). [Data for 8TF are shown in Additional file 1: Fig. S3A–D].

Fig. 3

46, XY DSD and 46,XX fibroblast derived SLCs exhibit varying degrees of change in morphology after 1 month trans-differentiation. Representative images of GFPC and 6TF SLCs on Day7 and Day28; Shape factor quantifications (N = 3, n = 50–60 for each group). *Represent p values calculated from Mann Whitney statistical tests conducted between the indicated groups for (A) 46, XY; SRXY1 (B) 46,XY; CD; (C) 46,XY; WT1 and (D) 46,XX (E) Shape factor quantifications comparing GFPC and 6TF SLC among all the four genetic backgrounds

We also compared these parameters in 1-month-old experiment groups of all these DSD and female lines with that of control 46,XY. While the GFPC of the four lines had significantly different shapes at the start of the culture, these differences ebbed after the 6TF trans-differentiation process (Fig. 3E). Interestingly, this was coupled with persistent differences in sizes of XY DSD and XX-derived GFPC and 6TF cells with those of control 46,XY (Additional file 1: Fig. S3E). This showed that cells derived from DSD individuals can also undergo transdifferentiation but with different characteristics than the control 46,XY cells and each other, possibly reflecting the different etiology of the DSD condition.

Adhesion and proliferation of SLCs from XY, XX, and XY DSD fibroblast lines

A key feature of Sertoli cells is their ability to adhere to form tubules then proliferate. SLCs derived from 46,XY, 46,XY with CD and 46,XX were analysed using xCELLigence Real Time Cell Analysis (RTCA) assays to measure their adhesion and proliferation properties. 0–6 h covers the time for adhesion of SLC suspensions to culture surface, while 6–48 h covers the period for cell proliferation. The Cell Index (CI) was plotted over time (Additional file 1: Fig. S4B–F) and the slope of CI curve was used as a measure to discern the differences or similarities between different experimental groups in their adhesion and proliferation phenotypes. The 46, XY 6TF SLCs showed lower adhesion (Fig. 4A) and slower proliferation (Fig. 4B) than the GFP control. This might be indicative of cell differentiation, which usually correlates negatively with proliferation. By contrast, 46 XY 8TF SLCs exhibited higher adhesion than GFPC while showing slower proliferation (Additional file 1: Fig. S4A).

Fig. 4

xCELLigence assay shows that 6TF SLCs exhibit distinct adhesion and proliferation phenotype than GFP controls across different genetic backgrounds. Slope of 0–6 h cell index (CI) adhesion curves (A) and slope of 6–48 h cell index (CI) proliferation curves (B) for 1 month old GFPC (green) and 6TF (blue) SLC derived from fibroblasts of indicated genetic backgrounds. All the experimental readings represent an average of three biological replicates (N = 3), * represent p values calculated from one way ANOVA test conducted amongst the three groups within each graph

We also analyzed the adhesive and proliferative abilities of SLCs derived from the 46, XY with CD and 46,XX cells. The GFP control fibroblasts of all three cell lines showed similar adhesion and proliferation profiles (Fig. 4A, B) despite their genetic differences. Sex dimorphism was observed in the adhesive abilities of the 46,XY SLCs and 46,XX SLCs, with the 46,XX SLCs showing a significantly higher adhesive ability (Fig. 4A). The 46,XY;CD SLCs showed an intermediate adhesive ability, significantly higher than the 46,XY SLCs but significantly lower than the 46,XX SLCs. A trend towards sex dimorphism was also observed when comparing the proliferation of 46,XY SLCs and 46, XX SLCs although this did not reach statistical significance (Fig. 4B). The 46,XY;CD SLCs once again showed significantly higher proliferation than the 46,XY SLCs, to a comparable level with the 46,XX SLCs (Fig. 4B).

XX 46,XX and 46,XY DSD fibroblasts derived SLC exhibited altered levels of Sertoli marker gene expression

The 46,XX- and 46,XY DSD-derived SLCs were evaluated for expression of Sertoli cell markers. Principal component analysis (PCA) of transcriptomes showed that all the samples belonging to each of the four cell lines clustered away from aSC as observed earlier for 46,XY SLCs (Fig. 2A). The 6TF and 8TF samples clustered away from GFPC in 46, XY; SRY1, 46,XY; WT1 and 46, XX but they didn’t separate as much in 46,XY;CD (Additional file 1: Fig. S5-S1). The correlation plot for GFPC, 6TF SLC and 8TF SLC derived from each of the four cell lines showed that the replicate samples correlated with each other and each of the three experimental groups within each line were distinct from one another (Additional file 1: Fig. S5-2B,C,D). Together these data suggest that the experimental procedure had a high level of reproducibility.

We investigated the fold change in expression of known DSD genes and Sertoli markers.

46,XX cell line: Notably, while 46, XX SLCs expressed the expected Mogrify TFs in 6TF and 8TF SLCs (Additional file 1: Fig. S5-3D) they did not show increased SOX9 expression over GFPC, either in the RNA seq data (Fig. 5A) or in qPCR data (Additional file 1: Fig. S5-3E). Even though they didn’t express SOX9, they did express bipotential gonad markers GADD45G, ZFPM2, NR0B1 and immature SC markers KRT18, INHBA, NCAM1/2, HSD17B3 and the all-stage marker GDNF. They barely expressed any mature markers, unlike 46,XY SLCs. Even though their SC marker expression profile was subdued in comparison to 46,XY SLCs, the presence of NR5A1 and EBF1 in the transduction mix had a similar effect on 8TF XX SLC in inducing Leydig markers HSD3B2 and CYP11A1 (Additional file 1: Fig. S5-3D). Lack of induction of SOX9 in 46,XX-derived SLCs interestingly coincides with lack of significant morphometric change (Fig. 3D).

Fig. 5

46, XY DSD and 46, XX fibroblasts-derived 6TF SLCs show varying levels of Sertoli specific marker gene expression.  Heat map showing 6TF vs. GFPC log2 fold change for indicated gonadal marker genes as measured by RNAseq for each of the indicated genetic backgrounds: 46,XY (N = 3), 46,XY; SRXY1 (N = 2); 46,XY; CD (N = 3), 46,XY; WT1 (N = 3), 46,XX (N = 3)

46,XY, SRXY1 DSD cell line: 6TF and 8TF SLCs derived from 46, XY, SRXY1 DSD cells showed induction of SOX9 by both RNAseq (Fig. 5A) and qPCR (Additional file 1: Fig. S5-3E) as observed in the wild-type 46,XY-derived SLCs. These SLCs exhibited similar profiles of expression of bipotential gonad markers LHX9, GADD45G, ZFPM2; immature SC markers KRT18, INHBA/B, NCAM1 and all-stage marker GDNF. This showed that the trans-differentiation method was capable of pushing cells from DSD background genotype towards a Sertoli fate.

46,XY, campomelic dysplasia cell line: 46, XY;CD-derived 6TF SLCs showed increased SOX9 expression (Fig. 5A; 8TF shown in Additional file 1: Fig. S5-3B,E); over the respective GFP control which was notably much smaller than that observed in control 46, XY-derived SLCs. This is evident in both the qPCR data (Additional file 1: Fig. S5-3E) and in the RNAseq data (Additional file 1: Fig. S2E vs Additional file 1: Fig. S5-3B). 46, XY;CD-derived 6TF SLCs also showed a modest, not statistically significant rise in the expression of PTGDS, NCAM1/2, INHB and BMP4.

46,XY;WT1 DSD cell line: SOX9 was also expressed in 46, XY;WT1-derived 6TF and 8TF SLCs at levels comparable to control 46, XY-derived SLCs (Fig. 5A, Additional file 1: Fig. S5-3C,E). These SLCs expectedly exhibited an increase in expression of other immature SC markers like CYP26B1, PTGDS, KRT18, INHBA and NCAM2. They additionally expressed mature markers CLDN11, CTSL and another all-stage marker BMP4.

8TF SLCs derived from all the three DSD fibroblasts also exhibited higher levels of expression of Leydig markers HSD3B2 and CYP11A1 in comparison to 6TF SLCs, just like observed for control 46,XY SLCs.

The transcriptomic profiling of all these cell lines shows that the trans-differentiation method is robust enough to bring about fibroblast to Sertoli cell fate differentiation in cells belonging to different genetic backgrounds with some properties sexually dimorphic and DSD SLCs showing intermediate, distinct phenotypes.