Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
1
Stage-specific gene and transcript dynamics in 1
human male germ cells 2
Lara M. Siebert-Kuss,1,6 Henrike Krenz,2,6 Tobias Tekath,2 Marius Wöste,2 Sara Di Persio,1 3
Nicole Terwort,1 Margot J. Wyrwoll,3 Jann-Frederik Cremers,4 Joachim Wistuba,1 Martin 4
Dugas,2,5 Sabine Kliesch,4 Stefan Schlatt,1 Frank Tüttelmann,3 Jörg Gromoll,1 Nina Neuhaus1
5
and Sandra Laurentino1,7* 6
1 Centre of Reproductive Medicine and Andrology, Institute of Reproductive and 7
Regenerative Biology, University of Münster, Münster, Germany. 8
2 Institute of Medical Informatics, University of Münster, Münster, Germany.
9
3 Institute of Reproductive Genetics, University of Münster, Münster, Germany. 10
4 Department of Clinical and Surgical Andrology, Centre of Reproductive Medicine and 11
Andrology, University Hospital of Münster, Münster, Germany. 12
5 Institute of Medical Informatics, Heidelberg University Hospital, Heidelberg, Germany. 13
6 The authors consider that the first two authors should be regarded as joint first authors 14
7 Lead Contact 15
* Correspondence: Sandra.Laurentino@ukmuenster.de 16
.CC-BY-NC-ND 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(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity.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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
2
Abstract 17
Cell differentiation processes are highly dependent on cell stage-specific gene expression, 18
including timely production of alternatively spliced transcripts. One of the most 19
transcriptionally rich tissues is the testis, where the process of spermatogenesis, or 20
generation of male gametes, takes place. To date, germ cell-specific transcriptome dynamics 21
remain understudied due to limited transcript information emerging from short-read 22
sequencing technologies. To fully characterize the transcriptional profiles of human male 23
germ cells and to understand how the human spermatogenic transcriptome is regulated, we 24
compared whole transcriptomes of men with different types of germ cells missing from their 25
testis. Specifically, we compared the transcriptomes of testis lacking germ cells (Sertoli cell-26
only phenotype; SCO; n=3), with an arrest at the stage of spermatogonia (SPG; n=4), 27
spermatocytes (SPC; n=3), and round spermatids (SPD; n=3), with the transcriptomes of 28
testis with normal and complete spermatogenesis (Normal; n=3). We found between 839 and 29
4,138 differentially expressed genes (DEGs, log2 fold change
≥
1) per group comparison, 30
with the most prevalent changes observed between SPG and SPC arrest samples, 31
corresponding to the entry into meiosis. We detected highly germ cell-type specific marker 32
genes among the topmost DEGs of each group comparison. Moreover, applying state-of-the-33
art bioinformatic analysis we were able to evaluate differential transcript usage (DTU) during 34
human spermatogenesis and observed between 1,062 and 2,153 genes with alternatively 35
spliced transcripts per group comparison. Intriguingly, DEGs and DTU genes showed 36
minimal overlap (< 8%), suggesting that stage-specific splicing is an additional layer of gene 37
regulation in the germline. By generating the most complete human testicular germ cell 38
transcriptome to date, we unravel extensive dynamics in gene expression and alternative 39
splicing during human spermatogenesis. 40
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
3
Introduction 41
Human male germ cell differentiation is a complex process requiring cell type-specific 42
transcriptome regulation. Disturbances in spermatogenesis causing male infertility range 43
from maturation arrest at different germ cell stages to complete lack of germ cells, known as 44
Sertoli cell-only (SCO) phenotype. Although an increasing number of male infertility cases 45
can be attributed to pathogenic variants in genes involved in spermatogenesis 46
(Houston et al., 2021), the number of causative pathogenic variants identified so far remains 47
small (Tüttelmann et al., 2018). The identification and understanding of genetic causes for 48
male infertility is hindered by the lack of data regarding transcriptomic dynamics during 49
human spermatogenesis. 50
In order to obtain testicular cell-type specific gene expression profiles, previous studies took 51
advantage of samples with distinct histological phenotypes of male infertility using samples 52
matched by cellular composition (Winge et al., 2018) or by performing comparative 53
microarray analyses of samples differing in the presence of one specific germ cell-type (von 54
Kopylow et al., 2010; Chalmel et al., 2012; Lecluze et al., 2018). For example, comparing 55
testicular tissues with SCO and spermatogonial arrest phenotypes, which only differ in the 56
presence of spermatogonia, von Kopylow et al., (2010) were able to identify transcripts 57
specifically expressed by spermatogonia. The authors identified the spermatogonial markers 58
FGFR3 and UTF1, which are currently considered specific markers for different 59
spermatogonial subpopulations (Guo et al., 2018; Sohni et al., 2019; Di Persio et al., 2021). 60
Chalmel et al. (2012) expanded on this approach by including samples from different 61
developmental stages and arrest phenotypes, thereby extracting the transcriptional profiles 62
of additional germ cell types. These studies demonstrated that the comparison of distinct 63
arrest phenotypes allows the identification of transcripts expressed at specific stages of germ 64
cell differentiation during normal spermatogenesis (von Kopylow et al., 2010; 65
Chalmel et al., 2012). Technological developments such as RNA sequencing (RNA-seq) now 66
enable an unbiased and more comprehensive analysis of the transcriptome. Specifically, 67
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
4
single cell RNA-sequencing (scRNA-seq) of human testicular tissues has revolutionized 68
germ cell-specific RNA profiling by allowing the identification of cell type-specific gene 69
expression patterns (Guo et al., 2018; Hermann et al., 2018; Wang et al., 2018; Sohni et al., 70
2019; Di Persio et al., 2021). However, scRNA-seq results in sparser data compared to 71
conventional bulk RNA-seq and, by sequencing from the poly-A tail of transcripts, generates 72
limited information on transcriptional isoforms (Tekath and Dugas, 2021). Total RNA-seq 73
therefore results in the most complete capture of the transcriptome, including all transcripts 74
obtained through post-transcriptional processing. The testis presents unusual high levels of 75
these post-transcriptional events, including alternative splicing (AS) (Kan et al., 2005). AS 76
enables the production of different transcripts and proteins from a single gene, thereby also 77
constituting a crucial regulatory mechanism for gene expression. For example, storage of 78
immature mRNAs allows protein synthesis at transcriptionally silent stages of mouse 79
spermatogenesis (Iguchi et al., 2006; Naro et al., 2017). During human male germ cell 80
differentiation, AS events have so far been understudied, with the exception of the 81
association between hormone receptor genes splice site variants and human male infertility 82
(Song et al., 2002; Bruysters et al., 2008). Knowledge of the changes in isoforms that result 83
from AS during human spermatogenesis would open a new avenue for identifying so far 84
unknown causes for male infertility. 85
In this study, we aimed at generating the most complete human testicular germ cell 86
transcriptome to date. Combining the advantages of scRNA-seq data and total RNA-seq of 87
distinct pathological phenotypes, and using sophisticated bioinformatic analyses, we unveiled 88
the transcriptional profiles of male germ cell types and determined the changes in AS 89
patterns during human male germ cell differentiation. 90
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
5
Materials and Methods 91
Ethical approval 92
Male infertility patients included in this study underwent surgery for microdissection testicular 93
sperm extraction (mTESE; n=15) or to rule out a suspected malignant tumor (n=1) at the 94
Department of Clinical and Surgical Andrology of the Centre of Reproductive Medicine and 95
Andrology, University Hospital of Münster, Germany. Each patient gave written informed 96
consent (ethical approval was obtained from the Ethics Committee of the Medical Faculty of 97
Münster and the State Medical Board no. 2008-090-f-S) and one additional testicular sample 98
for the purpose of this study was obtained. Tissue proportions were snap-frozen or fixed in 99
Bouin’s solution. 100
Patient selection 101
In this study, we included testicular biopsies with a homogenous histological phenotype in 102
both testes from men showing SCO (SCO-1/ M1045, SCO-2/ M911, SCO-3/M1742), 103
spermatogenic arrests at the spermatogonial (SPG-1/ M1570, SPG-2/ M1575, SPG-3/ 104
M1072, SPG-4/ M2822), spermatocyte (SPC-1/ M1369, SPC-2/ M799, SPC-3/ M921), and 105
round spermatid stage (SPD-1/ M2227, SPD-2/ M1311, SPD-3/ M1400) (Table I). We 106
excluded patients with germ cell neoplasia and a history of cryptorchidism as well as acute 107
infections. For complete representation of the spermatogenic process, samples with 108
qualitatively and quantitatively normal spermatogenesis were included in this study 109
(Normal-1/M1544, Normal-2/M2224, Normal-3/M2234) obtained from patients with 110
obstructive azoospermia, e.g. due to congenital bilateral absence of the vas deferens 111
(CBAVD; Normal-1), anorgasmia (Normal-2) or due to suspected tumor that was not 112
confirmed (Normal-3). Prior to surgery, all patients underwent physical evaluation, hormonal 113
analysis of luteinizing hormone (LH), follicle stimulating hormone (FSH), and testosterone 114
(T), and semen analysis (World Health Organization, 2010). In addition to conventional 115
karyotyping and screening for azoospermia factor (AZF) deletions, whole exome sequencing 116
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
6
(WES) was performed for all patients, except for SPG-4 (who had undergone chemotherapy 117
because of leukemia) and one with normal spermatogenesis (Normal-3). WES data were 118
generated within the Male Reproductive Genomics (MERGE) study as previously published 119
(Wyrwoll et al., 2020) and were screened for variants in 230 candidate genes that have at 120
least a limited level of evidence for being associated with male infertility according to a recent 121
review (Houston et al., 2021). We also included a screening in the recently published genes 122
ADAD2, GCNA, MAJIN, MSH4, MSH5, RAD21L1, RNF212, SHOC1, STAG3, SYCP2, 123
TERB1, TERB2, and TRIM71, which are associated with non-obstructive azoospermia 124
(Riera-Escamilla et al., 2019; Krausz et al., 2020; Schilit et al., 2020; Hardy et al., 2021; 125
Salas-Huetos et al., 2021; Torres-Fernández et al., 2021; Wyrwoll et al., 2021). We screened 126
for rare (minor allele frequency [MAF] in gnomAD database < 0.01), possibly pathogenic 127
variants (stop-, frameshift-, and splice site variants) with a read depth > 10x, that were 128
detected in accordance with the reported mode of inheritance. 129
Histological evaluation of the human testicular biopsies 130
After overnight fixation in Bouin`s solution, the tissues were washed in 70% ethanol, 131
embedded in paraffin, and sectioned at 5 µm. AppiClear (Applichem, Cat# A4632.2500) was 132
used to dewax the tissue section. The cellular composition of all testicular biopsies (n=16) 133
was histologically examined on two periodic acid-Schiff (PAS)-stained sections from two 134
independent biopsies per testis. For PAS staining, the sections were first incubated with 1% 135
PA (Sigma-Aldrich, Cat# 1.005.240.100) and then in Schiffs reagent (Sigma-Aldrich, Cat# 136
1.090.330.500). Cell nuclei were counterstained with Mayer`s hematoxylin solution (Sigma-137
Aldrich, Cat# 1.092.490.500). After washing in tap water and dehydration through increasing 138
ethanol concentrations and AppiClear, slides were closed with Merckoglas (Sigma-Aldrich, 139
Cat# 1.039730.001). The slides were scanned using the Precipoint Viewpoint software 140
(Precipoint, Freising, Germany). The biopsies were evaluated based on the Bergmann and 141
Kliesch scoring method (Bergmann and Kliesch, 2010), which assigns a score from 0 to 10 to 142
each patient according to the percentage of tubules containing elongated spermatids. 143
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
7
Furthermore, the percentage of the seminiferous tubules with round spermatids, 144
spermatocytes or spermatogonia as the most advanced germ cell type was assessed, as 145
well as seminiferous tubules with SCO or hyalinized tubules (tubular shadows) (Table I). 146
RNA extraction from testicular tissues 147
We extracted total RNA from snap-frozen testicular tissues from all biopsies using the Direct-148
zolTM RNA Microprep kit (Zymo Research, CA, USA) according to manufacturer’s protocol. 149
Quantity and quality of isolated RNA were evaluated using RNA ScreenTape and the 150
TapeStation Analysis software 3.1.1 (Agilent Technologies, Inc., CA, USA). All samples had 151
intact ribosomal 18S and 21S bands. Samples with an RNA integrity number (RIN) >3.6 were 152
included in the analysis (Suntsova et al., 2019). 153
Library preparation and sequencing 154
Next-generation sequencing was performed by the service unit Core Facility Genomics of the 155
medical faculty at the University of Münster. Libraries were prepared according to the 156
NEBNext Ultra RNA II directional Library Prep kit (New England Biolabs, MA, USA) after 157
NEBNext rRNA depletion (New England Biolabs, MA, USA). The NextSeq HO Kit (Illumina 158
Inc., CA, USA) with 150 cycles was used for paired end sequencing on the NextSeq 500 159
system (Illumina Inc., CA, USA) with ~400 Million single reads per run. 160
Data processing 161
We processed the raw sequence data with the nextflow analysis pipeline nf-core/rnaseq 2.0 162
(Ewels et al., 2020) and annotated the transcripts with GENCODE release 36 genome 163
annotation based on the GRCh38.p13 genome reference (Frankish et al., 2019). Gene 164
expression counts were estimated using Salmon (Patro et al., 2017). 165
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
8
Differential gene expression analysis 166
All data were analyzed within the R Statistical Environment (RCoreTeam, 2020). We used 167
DESeq2 (Love et al., 2014) for analyzing differentially expressed genes (DEGs) following the 168
standard workflow for Salmon quantification files. DESeq2 uses a generalized linear model 169
based on estimated size factors and dispersion to calculate the log2 fold changes for each 170
gene (Love et al., 2014). Annotation was performed using the biomaRt R package 171
Normalization was performed using DESeq2 with the median of ratios method 172
(Love et al., 2014). Genes with a total count > 10 were considered for further analysis. DEGs 173
were calculated for each group comparison, i.e. SCO vs. SPG, SPG vs. SPC, SPC vs. SPD, 174
and SPD vs. Normal. P-values are calculated based on Wald test and adjusted with 175
Benjamini-Hochberg. Genes with a false discovery rate (FDR) < 0.05 and a log2 fold change 176
(FC)
≥
1 were considered DEGs. Dispersion of samples was visualized using DESeq2’s 177
PCAplot function for the top 500 genes with a total count > 10. 178
To evaluate gene expression of selected genes of interest at single-cell level, we generated 179
uniform manifold approximation and projection (UMAP) plots (McInnes et al., 2020) based on 180
our previously published dataset (Di Persio et al., 2021) using the tool Seurat 181
(Stuart et al., 2019; Hao et al., 2021). 182
Differential transcript usage analysis 183
For computing differential transcript usage (DTU) we employed the R package DTUrtle 184
(Tekath and Dugas, 2021), following the vignette workflow for human bulk RNA-seq analysis. 185
As for the DEG analysis, we annotated the transcripts with GENCODE release 36 genome 186
annotation. We calculated DTU genes for each group comparison (i.e. SCO vs. SPG, SPG 187
vs. SPC, SPC vs. SPD, and SPD vs. Normal) with the run_drimseq function. DTUrtle 188
conducts statistical analyses based on DRIMSeq (Nowicka and Robinson, 2016), i.e. a 189
likelihood ratio test is used on the estimated transcript proportions and precision parameter 190
(Tekath and Dugas, 2021). To increase the statistical power of the analysis, we filtered out 191
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
9
transcripts with low impact, i.e. less than 5% usage for all samples or a corresponding total 192
gene expression of less than 5 counts for all samples before the statistical testing. Also, only 193
genes with at least two high impact transcripts were considered. From the analysis we 194
obtained genes with an overall significant change in transcript usage as well as the 195
corresponding transcripts that drive the change in usage in those genes (both with overall 196
FDR < 0.05). 197
To decrease the number of analyzed transcripts per DTU genes, a post-hoc filtering was 198
applied, i.e. transcripts whose proportional expression deviated by less than 10% between 199
samples were excluded. In this study, we decided to only include transcripts, which fulfill the 200
criterion that all samples of one group must have a higher transcript usage compared to all 201
samples of the other group. 202
Pathway Analysis 203
Molecular function of DEGs and DTU genes were assessed via the Ingenuity Pathway 204
Analysis software (IPA; Qiagen, Hilden, Germany). A Benjamini-Hochberg multiple testing 205
correction P-value (FDR) <0.01 was used as threshold for significant molecular functions in 206
IPA. We selected the top 20 significant terms for molecular functions. 207
Statistical analysis 208
Statistical analysis was conducted as described in sections for differential gene expression 209
analysis, differential transcript usage analysis, and pathway analysis. 210
Results 211
Clinical characteristics of the study cohort 212
Hormonal evaluation revealed that patients with normal spermatogenesis had FSH values 213
within the reference range, whereas most patients with spermatogenic arrests had elevated 214
FSH levels (Table I). Other than patient SPD-3, who had a low grade XXY mosaicism 215
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
10
(47,XXY[2]/46,XY[28]), no patients showed chromosomal abnormalities. By analyzing WES 216
data of our patients with unknown reasons for infertility, we did not identify any likely high 217
impact pathogenic variants in known male infertility candidate genes. 218
Testicular phenotypes are recapitulated at RNA level 219
To obtain whole transcriptome expression profiles, we sequenced total RNA of human 220
testicular biopsies with SCO, SPG, SPD, as well as normal spermatogenesis (n=16) (Fig. 221
1A). Prior to sequencing, a careful histological examination (Fig. 1B) ensured that both testes 222
presented comparable phenotypes, and no sperm was found via mTESE, except in the 223
normal samples (Table I). Following total RNA-seq, principal component analysis (PCA) 224
organized the spermatogenic arrest samples in consecutive order (Fig. 1C), mirroring their 225
sequential spermatogenic phenotypes. 226
Comparative analysis reveals germ cell-specific transcriptional profiles 227
We aimed at generating germ cell-specific expression profiles to study transcriptional 228
changes throughout spermatogenesis. To this end we performed differential gene expression 229
analysis between groups of different cellularity: SCO versus SPG (comparison 1), SPG 230
versus SPC (comparison 2), SPC versus SPD (comparison 3) and SPD versus normal 231
(comparison 4). This revealed between 839 and 4,138 DEGs in the four comparisons 232
calculated (FDR < 0.05 and absolute log2 FC
≥
1, Fig. 2). In the SCO versus SPG 233
comparison, most transcriptional changes were due to the increased expression of 2,073 234
genes in SPG samples (Fig. 2A, Supplementary Table SI). Co-expression of DEGs among 235
all samples revealed the level of gene expression remained high in the other groups 236
containing spermatogonia (SPC, SPD, Normal), indicating that most of these transcripts 237
originate from the presence of spermatogonia. Indeed, among the highly expressed genes 238
were well-known spermatogonial genes such as MAGEA4 and FGFR3 (Supplementary 239
Table SII). The most prominent changes in gene expression were found when comparing 240
SPG with SPC samples (Fig. 2B, Supplementary Table SIII). The 2,886 genes that were high 241
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
11
in expression included spermatocyte-specific genes like AURKA and OVOL1 242
(Supplementary Table SII). The same genes also showed high expression in SPD and 243
normal samples and low to absent expression in SPG and SCO. This indicates that these 244
genes are specific to spermatocytes, rather than the result of gene expression alterations in 245
other cell types. When comparing SPC with SPD samples we found 2,345 highly expressed 246
genes in SPD samples (Fig. 2C, Supplementary Table SIV), including spermiogenesis 247
marker genes TNP1 and PRM1 (Supplementary Table SII). These genes also showed higher 248
expression in normal samples and lower expression in samples lacking spermatids (SPC, 249
SPG, SCO), in accordance with their spermatid-specific expression pattern. The most subtle 250
changes in gene expression were detected when comparing SPD with samples showing 251
normal spermatogenesis (Supplementary table V), in which the presence of elongated 252
spermatids is the only histological difference. Genes with increased expression in normal 253
samples (776) showed lower expression levels in the spermatogenic arrest samples (SPD, 254
SPC, SPG, SCO) (Fig. 2D) and, among others, included genes associated with the sperm 255
flagellum like CATSPER3 and TEKT2 (Supplementary Table SII). 256
Novel germ cell-specific marker genes and their expression at single cell resolution 257
To identify novel germ cell-specific marker genes, we focused on the top 100 DEGs per 258
group comparison with elevated expression in SPG, SPC, SPD, and normal samples. After 259
evaluating the expression of all genes for their germ cell-specificity in our published scRNA-260
seq dataset of 3 patients with normal spermatogenesis (Di Persio et al., 2021) (Fig. 3A), we 261
show 3 genes per group comparison as examples. Accordingly, from the SCO vs SPG 262
comparison, we selected the leucine zipper protein 4 gene (LUZP4), testis specific protein Y-263
linked 4 (TSPY4), and anomalous homeobox (ANHX), which showed increased expression 264
in SPG samples (Fig. 3B). Importantly, at single cell level, the expression of these genes was 265
specific for spermatogonia (Fig. 3C). Based on the SPG vs SPC comparison we selected the 266
proline rich acidic protein 1 (PRAP1), ferritin heavy chain like 17 (FTHL17) and synaptogyrin 267
4 (SYNGR4) (Fig. 3D). The spermatocyte-specific expression of these genes was confirmed 268
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
12
in the single cell dataset (Fig. 3E). For SPD samples, genes with high expression were 269
proline rich 30 (PRR30), actin like 7A (ACTL7A), and high mobility group box 4 (HMGB4) 270
(Fig. 3F). Based on the expression patterns at single cell level, PRR30, ACTL7A and 271
HMGB4 were expressed in early and late spermatids (Fig. 3G). TP53 target 5 (TP53TG5), 3-272
oxoacid CoA-transferase 2 (OXCT2), and hemogen (HEMGN) were the highest expressed 273
genes in normal samples in comparison to SPD samples (Fig. 3H), and also at single-cell 274
level their expression was specific for late spermatids (Fig. 3I). 275
Alternative splicing is uncoupled from gene expression 276
To study alternative splicing, we performed a DTU analysis between all four group 277
comparisons. DTU analysis calculates and compares the proportional contributions (referred 278
to as ‘usage’) of transcripts to the overall expression of a gene. A gene has a DTU event, i.e. 279
is a DTU gene, when at least two of its transcripts are differentially used between two 280
groups. We found between 1,062 and 2,153 DTU genes in each of the four comparisons 281
(Supplementary Tables SVI-SIX). By comparing DTU genes to DEGs, we found an overlap 282
of less than 8% in all four comparisons, indicating that the expression of most genes is 283
regulated either at the pre- or the post-transcriptional level (Fig. 4), and that only few genes 284
are regulated at these two levels. Furthermore, we found that the proportion of DEGs to 285
DTUs in all group comparisons was 2:1 (Fig. 4A-C), except for SPD vs Normal, where this 286
ratio was inversed with more DTU genes than DEGs (Fig. 4D). 287
DEGs and DTU genes are involved in different biological pathways 288
We used IPA to evaluate the molecular functions of the DEGs and DTU genes at the 289
different germ cell stages. In line with the small overlap between the DEG and DTU gene 290
sets, we found minor overlaps between the top 20 significantly enriched molecular functions 291
of DEGs and DTU genes in all four groups (Fig. 5). Both gene sets contained genes involved 292
in organization of cytoskeleton/cytoplasma, microtubule dynamics, apoptosis, necrosis, and 293
segregation of chromosomes. IPA analysis on DEGs highlighted functional enrichment 294
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
13
annotations that can be attributed to the most advanced germ cell type in each group 295
comparison (e.g. development of stem cells, segregation of chromosomes) (Fig. 5A). 296
In comparison to the functional annotations of DEGs, 26% of molecular functions of the DTU 297
genes overlapped across the four group comparisons (Fig. 5B). Among the overlapping 298
terms were microtubule dynamics, organization of cytoplasm, and cytoskeleton. More 299
general biological functions (e.g. RNA metabolism, cell survival) were enriched among the 300
DTU genes in each group comparison. 301
Stage-specific splicing is an additional layer of gene regulation in the germline 302
To study alternatively spliced transcripts, we investigated the transcript biotypes of selected 303
DTU genes. In comparison to the proportional distribution of transcript biotypes annotated in 304
GENCODE (Frankish et al., 2019) we found that most of the DTU events, regardless of the 305
group comparison, result in protein coding transcripts (Fig. 6A). In the comparison between 306
SPD arrest and normal, two protein-coding isoforms of actin like 6A (ACTL6A) displayed 307
differential usage (Fig. 6B). While ACTL6A-202 (ENST00000429709.7) was the predominant 308
isoform, with an average usage of 52% in SPD samples, normal samples predominantly 309
used the ACTL6A-203 isoform (ENST00000450518.6), which has an alternative 5’ splice site 310
(Fig. 6B). In comparison, spermatogenesis associated 4 (SPATA4) also showed a switch in 311
usage for its protein coding isoforms SPATA4-201 (ENST00000280191.7) and SPATA4-203 312
(ENST00000515234.1) in the comparison of SPC versus SPD samples (Fig. 6C). SPC 313
samples showed a significantly decreased usage of SPATA4-201 and a significantly 314
increased usage of SPATA4-203, whereas SPD samples exclusively used the SPATA4-201 315
isoform (Fig. 6C). These two isoforms use alternative transcriptional start and stop sites. In 316
contrast to ACTL6A, SPATA4 was also a DEG in this group comparison and had a higher 317
expression level in SPD samples (Supplementary Fig. S1A and S1B). Intriguingly, the 318
second largest group of biotypes with DTU events were retained introns (Fig. 6A). For 319
synaptonemal complex protein 3 (SYCP3), we found a significantly increased usage of the 320
retained intron isoform SYCP3-204 (ENST00000478139.1) in SPG samples, whilst SPC 321
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
14
samples had an increased usage of the protein coding isoform SYCP3-202 322
(ENST00000392924.2) (Fig. 6D). In this group comparison, SYCP3 showed increased 323
expression in SPC samples (Fig. S1C). A switch in usage from coding to non-coding 324
transcripts was also observed for marker of proliferation Ki-67 (MKI67) (Fig. 6E), which did 325
not show changes in gene expression (Supplementary Fig. S1D). However, the protein 326
coding isoform MKI67-202 (ENST00000368654.8) was lower expressed in SPC samples in 327
comparison to SDP samples. In contrast, the processed transcript isoform MKI67-205 328
(ENST00000484853.1) showed significantly increased usage in SPC samples and 329
decreased usage in SDP samples. 330
Discussion 331
The study of expression patterns in testis is developing rapidly, however a complete picture 332
of the transcriptome of human germ cells remained unexplored. Here, we demonstrate that 333
the progression of human male germ cell differentiation is accompanied by major transcript 334
dynamics, including germ cell-type dependent transcription and splicing events. The latter 335
resulting in stage-specific transcript isoforms. We found that alternative splicing is mainly 336
uncoupled from the level of gene expression and facilitates a crucial layer of gene regulation 337
in germ cells, especially in the late stages of spermatogenesis. 338
The differentiation of male germ cells requires cell-specific transcriptional regulation 339
(Guo et al., 2018; Hermann et al., 2018; Di Persio et al., 2021). Previous bulk microarray 340
studies demonstrated that the use of homogeneous human testicular tissues with stage-341
specific germ cell-arrests allows for the identification of germ cell-specific transcript profiles, 342
thus allowing the unbiased analysis of germ cell populations in their cognate environment 343
(von Kopylow et al., 2010; Chalmel et al., 2012). Due to the use of microarrays in previous 344
studies, the full spectrum of transcriptome profiles, including isoform information, remained 345
largely unknown.
346
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
15
Our systematic analysis of total RNA from testicular biopsies with well-defined, distinct germ 347
cell compositions revealed significant changes in gene expression (839 to 4,138 DEGs; 348
Supplementary Tables SI, SIII, SIV, SV). Most changes were detected between samples with 349
spermatogonial and spermatocyte arrest, indicating that the entry into meiosis results in a 350
peak of transcriptional activity. Transcripts expressed at this stage are known to be stored for 351
translation at later differentiation stages (Paronetto and Sette, 2010; Wang et al., 2020). 352
Among the topmost expressed genes for spermatogonia (2,073), spermatocytes (2,886), 353
round spermatids (2,345) and elongated spermatids (776), we found highly germ cell-specific 354
genes, which to our knowledge were not previously associated with the respective germ cell 355
stages in humans (Supplementary Table SX). 356
The transcriptional output of a gene depends not only on the level of RNA expression but 357
also on post-transcriptional processing of RNA transcripts, for instance through AS, which 358
allows a single gene to originate different transcripts and potentially different proteins 359
(Baralle and Giudice, 2017). Although it is well known that the testis is an organ with high 360
transcriptome diversity, AS is still understudied in human spermatogenesis. Making use of a 361
powerful bioinformatic technique, the DTU analysis, we were able to study for the first time 362
transcriptome dynamics during human spermatogenesis. Several studies observed 363
discontinuous patterns of transcription throughout murine and human spermatogenesis 364
(Jan et al., 2017; Vara et al., 2019). In our study, we further characterized the ongoing 365
transcriptional changes during human spermatogenesis by identifying between 1,062 and 366
2,153 genes whose transcripts were alternatively spliced at different germ cell stages 367
(Supplementary Tables SVI, SVII, SVIII, SIX). Our results indicate that alternative splicing 368
extends the transcriptome diversity in germ cells, which already present high transcriptional 369
activity, as we found that alternative splicing events are more prevalent between the 370
premeiotic and meiotic germ cell stages. As we identified more alternatively spliced genes 371
than changes in gene expression between the spermatid arrest and normal samples, we 372
hypothesize that in the final stage of spermiogenesis transcriptome diversity arises primarily 373
from alternative splicing rather than by changes in gene transcription. In line with this idea 374
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
16
are studies in mice showing that genes required for spermiogenesis are already expressed at 375
the beginning of meiosis (da Cruz et al., 2016) and that transcription in elongated spermatids 376
is decreased due to the highly compacted chromatin structure (Sassone-Corsi, 2002). Even 377
in the absence of transcriptional activity in the nucleus, stored unprocessed transcripts can 378
maintain translational activity in late stages of germ cell differentiation (Wang et al., 2020). 379
Our study demonstrates that alternative splicing is uncoupled from the level of gene 380
expression during human spermatogenesis, as only a minority of genes were both 381
differentially expressed and differentially spliced at each respective germ cell stage. Data on 382
the comparison of DEG and DTU genes in other tissues also revealed that these two gene 383
sets hardly overlap and different molecular functions are enriched (Solovyeva et al., 2021). 384
Interestingly, we found that DEGs were enriched for germ cell-specific processes, whereas 385
DTU genes were involved in more general biological processes, suggesting that during 386
human spermatogenesis these functions are predominantly regulated at transcriptional and 387
post-transcriptional level, respectively. We suggest that general processes are uncoupled 388
from the level of gene expression, as these need to be maintained even in transcriptionally 389
silent cells such as later germ cells. By looking more precisely into four DTU genes, we 390
demonstrate the importance of our dataset for further research in the field of male infertility. 391
For example, we were able to reveal that SPD and normal samples express different protein 392
coding transcripts of ACTL6A, something that would have been overlooked by conventional 393
DEG analysis. It is also relevant to understand which gene products with potentially different 394
functionality are produced by AS, as it has been shown that this may play a role in the 395
etiology of several diseases (Scotti and Swanson, 2016) such as cancer (Wiesner et al., 396
2015; Vitting-Seerup and Sandelin, 2017). Whether alterations in alternatively spliced 397
transcript expression also plays a role in the pathology of infertility remains to be assessed. 398
We showed that some crucial spermatogenic genes such as SYCP3 appear to be regulated 399
at both the transcriptional and post-transcriptional levels. SYCP3 is already expressed as an 400
immature non-coding transcript in SPG samples, whereas the mature transcript is 401
predominantly expressed in SPC samples. A previous study indicated that spermatogonia 402
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
17
already express genes required for meiosis (Jan et al., 2017), however the mechanism 403
behind this observation was not addressed. In murine spermatogenesis, intron retention 404
ensures timely and stage-depended gene expression (Naro et al., 2017). Our data supports 405
the hypothesis that the expression of spermatogenic stage-specific genes might be 406
functionally regulated through alternative splicing by intron retention during human 407
spermatogenesis. Our data strongly highlights the need to further analyze the splicing 408
machinery in human germ cells. 409
Our whole transcriptome analysis provides an unbiased evaluation of transcriptome 410
dynamics during human spermatogenesis for novel and/or germ cell-specific genes. By not 411
only focusing on protein coding exons but capturing the presence of all alternative transcripts 412
at different stages of human spermatogenesis, our dataset allows to study the role of non-413
coding pathogenic variants, e.g. in splice sites, by pinpointing the expression and splice 414
isoforms of germ cell-specific transcripts, thereby prospectively improving the genetic 415
diagnosis of male infertility. 416
Author´s roles 417
Study conception and design: J.G., F.T., N.N. and S.L.; Supervision: N.N., S.L.; Acquisition 418
and evaluation of clinical data: J.F.C., S.K., F.T.; Lab work: N.T., S.D.P., J.W.; Data and 419
bioinformatic analyses: L.M.S.-K., H.K., M.W., T.T., M.D.; Exome analyses/ evaluations: 420
M.J.W., F.T. Writing Original Draft; L.M.S-K., H.K., S.S, N.N., S.L.; All authors were involved 421
in editing, read and approved the final version of the manuscript. 422
Acknowledgements 423
We thank Heidi Kersebom and Elke Kößer for histological evaluation of testicular tissues and 424
we also thank Sabine Forsthoff for excellent support in endocrinological measurements. We 425
thank the service unit Core Facility Genomik of the medical faculty from the University of 426
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
18
Muenster for performing the next-generation sequencing. Schematic figure 1A was created 427
with BioRender.com. 428
Funding 429
This work was funded by the German research foundation (CRU362) (grants to N.N. (NE 430
2190/3-1, NE 2190/3-2), S.L. (LA 4064/3-2) F.T. (TU 298/4-1, 4-2, 5-1, 5-2, 7-1), J.G. (GR 431
1547/24-2) and a pilot project to H.K.) and by institutional funding by the CeRA.
432
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
19
Conflict of Interest 433
The authors declare no competing interests. 434
Data availability 435
The testicular RNA-Seq data of all patients in this study has been deposited in the European 436
Genome-Phenome Archive and is available under EGAS00001006135. 437
References 438
Baralle FE, Giudice J. Alternative splicing as a regulator of development and tissue identity. 439
Nat Rev Mol Cell Biol 2017;18:437–451. 440
Bergmann, M., Kliesch, S. Testicular biopsy and histology. In: Nieschlag E., Behre H.M., and 441
Nieschlag S. (eds) Andrology. 2010; Springer: Berlin, Heidelberg. 442
Bruysters M, Christin-Maitre S, Verhoef-Post M, Sultan C, Auger J, Faugeron I, Larue L, 443
Lumbroso S, Themmen APN, Bouchard P. A new LH receptor splice mutation 444
responsible for male hypogonadism with subnormal sperm production in the 445
propositus, and infertility with regular cycles in an affected sister. Human 446
Reproduction 2008;23:1917–1923. 447
Chalmel F, Lardenois A, Evrard B, Mathieu R, Feig C, Demougin P, Gattiker A, Schulze W, 448
Jégou B, Kirchhoff C, et al. Global human tissue profiling and protein network 449
analysis reveals distinct levels of transcriptional germline-specificity and identifies 450
target genes for male infertility. Human Reproduction 2012;27:3233–3248. 451
Cruz I da, Rodríguez-Casuriaga R, Santiñaque FF, Farías J, Curti G, Capoano CA, Folle GA, 452
Benavente R, Sotelo-Silveira JR, Geisinger A. Transcriptome analysis of highly 453
purified mouse spermatogenic cell populations: gene expression signatures switch 454
from meiotic-to postmeiotic-related processes at pachytene stage. BMC Genomics 455
2016;17:294. 456
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
20
Di Persio S, Tekath T, Siebert-Kuss LM, Cremers J-F, Wistuba J, Li X, Meyer zu Hörste G, 457
Drexler HCA, Wyrwoll MJ, Tüttelmann F, et al. Single-cell RNA-seq unravels 458
alterations of the human spermatogonial stem cell compartment in patients with 459
impaired spermatogenesis. Cell Reports Medicine 2021;2:100395. 460
Ewels PA, Peltzer A, Fillinger S, Patel H, Alneberg J, Wilm A, Garcia MU, Di Tommaso P, 461
Nahnsen S. The nf-core framework for community-curated bioinformatics pipelines. 462
Nat Biotechnol 2020;38:271–271. 463
Frankish A, Diekhans M, Ferreira A-M, Johnson R, Jungreis I, Loveland J, Mudge JM, Sisu 464
C, Wright J, Armstrong J, et al. GENCODE reference annotation for the human and 465
mouse genomes. Nucleic Acids Research 2019;47:D766–D773. 466
Guo J, Grow EJ, Mlcochova H, Maher GJ, Lindskog C, Nie X, Guo Y, Takei Y, Yun J, Cai L, 467
et al. The adult human testis transcriptional cell atlas. Cell Res 2018;28:1141–1157. 468
Hao Y, Hao S, Andersen-Nissen E, Mauck WM, Zheng S, Butler A, Lee MJ, Wilk AJ, Darby 469
C, Zager M, et al. Integrated analysis of multimodal single-cell data. Cell 470
2021;184:3573-3587.e29. 471
Hardy JJ, Wyrwoll MJ, Mcfadden W, Malcher A, Rotte N, Pollock NC, Munyoki S, Veroli MV, 472
Houston BJ, Xavier MJ, et al. Variants in GCNA, X-linked germ-cell genome integrity 473
gene, identified in men with primary spermatogenic failure. Hum Genet 474
2021;140:1169–1182. 475
Hermann BP, Cheng K, Singh A, Roa-De La Cruz L, Mutoji KN, Chen I-C, Gildersleeve H, 476
Lehle JD, Mayo M, Westernströer B, et al. The mammalian spermatogenesis single-477
cell transcriptome, from spermatogonial stem cells to spermatids. Cell Reports 478
2018;25:1650-1667.e8. 479
Houston BJ, Riera-Escamilla A, Wyrwoll MJ, Salas-Huetos A, Xavier MJ, Nagirnaja L, 480
Friedrich C, Conrad DF, Aston KI, Krausz C, et al. A systematic review of the 481
validated monogenic causes of human male infertility: 2020 update and a discussion 482
of emerging gene–disease relationships. Human Reproduction Update 2021;0:15. 483
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
21
Iguchi N, Tobias JW, Hecht NB. Expression profiling reveals meiotic male germ cell mRNAs 484
that are translationally up- and down-regulated. Proceedings of the National Academy 485
of Sciences 2006;103:7712–7717. 486
Jan SZ, Vormer TL, Jongejan A, Röling MD, Silber SJ, Rooij DG de, Hamer G, Repping S, 487
Pelt AMM. van. Unraveling transcriptome dynamics in human spermatogenesis. 488
Development 2017;144:3659–3673. 489
Kan Z, Garrett-Engele PW, Johnson JM, Castle JC. Evolutionarily conserved and diverged 490
alternative splicing events show different expression and functional profiles. Nucleic 491
Acids Research 2005;33:5659–5666. 492
Kopylow K von, Kirchhoff C, Jezek D, Schulze W, Feig C, Primig M, Steinkraus V, Spiess A-493
N. Screening for biomarkers of spermatogonia within the human testis: a whole 494
genome approach. Human Reproduction 2010;25:1104–1112. 495
Krausz C, Riera-Escamilla A, Moreno-Mendoza D, Holleman K, Cioppi F, Algaba F, Pybus 496
M, Friedrich C, Wyrwoll MJ, Casamonti E, et al. Genetic dissection of spermatogenic 497
arrest through exome analysis: clinical implications for the management of 498
azoospermic men. Genet Med 2020;22:1956–1966. 499
Lecluze E, Jégou B, Rolland AD, Chalmel F. New transcriptomic tools to understand testis 500
development and functions. Molecular and Cellular Endocrinology 2018;468:47–59. 501
Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-502
seq data with DESeq2. Genome Biol 2014;15:550. 503
McInnes L, Healy J, Melville J. UMAP: Uniform Manifold Approximation and Projection for 504
dimension reduction. arXiv:180203426 [cs, stat] [Internet] 2020;Available from: 505
http://arxiv.org/abs/1802.03426. 506
Naro C, Jolly A, Di Persio S, Bielli P, Setterblad N, Alberdi AJ, Vicini E, Geremia R, De la 507
Grange P, Sette C. An orchestrated intron retention program in meiosis controls 508
timely usage of transcripts during germ cell differentiation. Developmental Cell 509
2017;41:82-93.e4. 510
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
22
Nowicka M, Robinson MD. DRIMSeq: a Dirichlet-multinomial framework for multivariate 511
count outcomes in genomics. F1000Research 2016;5:1–25. 512
Paronetto MP, Sette C. Role of RNA-binding proteins in mammalian spermatogenesis: RNA-513
binding proteins and germ cells. International Journal of Andrology 2010;33:2–12. 514
Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast and bias-aware 515
quantification of transcript expression. Nat Methods 2017;14:417–419. 516
RCoreTeam. R: A language and environment for statistical computing. Vienna, Austria: R 517
Foundation for Statistical Computing 2020; 518
Riera-Escamilla A, Enguita-Marruedo A, Moreno-Mendoza D, Chianese C, Sleddens-Linkels 519
E, Contini E, Benelli M, Natali A, Colpi GM, Ruiz-Castañé E, et al. Sequencing of a 520
‘mouse azoospermia’ gene panel in azoospermic men: identification of RNF212 and 521
STAG3 mutations as novel genetic causes of meiotic arrest. Human Reproduction 522
2019;34:978–988. 523
Salas-Huetos A, Tüttelmann F, Wyrwoll MJ, Kliesch S, Lopes AM, Goncalves J, Boyden SE, 524
Wöste M, Hotaling JM, GEMINI Consortium, et al. Disruption of human meiotic 525
telomere complex genes TERB1, TERB2 and MAJIN in men with non-obstructive 526
azoospermia. Hum Genet 2021;140:217–227. 527
Sassone-Corsi P. Unique chromatin remodeling and transcriptional regulation in 528
spermatogenesis. Science 2002;296:2176–2178. 529
Schilit SLP, Menon S, Friedrich C, Kammin T, Wilch E, Hanscom C, Jiang S, Kliesch S, 530
Talkowski ME, Tüttelmann F, et al. SYCP2 translocation-mediated dysregulation and 531
frameshift variants cause human male infertility. The American Journal of Human 532
Genetics 2020;106:41–57. 533
Scotti MM, Swanson MS. RNA mis-splicing in disease. Nat Rev Genet 2016;17:19–32. 534
Sohni A, Tan K, Song H-W, Burow D, Rooij DG de, Laurent L, Hsieh T-C, Rabah R, 535
Hammoud SS, Vicini E, et al. The neonatal and adult human testis defined at the 536
single-cell level. Cell Reports 2019;26:1501-1517.e4.
537
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
23
Solovyeva EM, Ibebunjo C, Utzinger S, Eash JK, Dunbar A, Naumann U, Zhang Y, Serluca 538
FC, Demirci S, Oberhauser B, et al. New insights into molecular changes in skeletal 539
muscle aging and disease: Differential alternative splicing and senescence. 540
Mechanisms of Ageing and Development 2021;197:111510. 541
Song GJ, Park Y-S, Lee YS, Lee CC, Kang IS. Alternatively spliced variants of the follicle-542
stimulating hormone receptor gene in the testis of infertile men. Fertility and Sterility 543
2002;77:499–504. 544
Stuart T, Butler A, Hoffman P, Hafemeister C, Papalexi E, Mauck WM, Hao Y, Stoeckius M, 545
Smibert P, Satija R. Comprehensive integration of single-cell data. Cell 546
2019;177:1888-1902.e21. 547
Suntsova M, Gaifullin N, Allina D, Reshetun A, Li X, Mendeleeva L, Surin V, Sergeeva A, 548
Spirin P, Prassolov V, et al. Atlas of RNA sequencing profiles for normal human 549
tissues. Sci Data 2019;6:36. 550
Tekath T, Dugas M. Differential transcript usage analysis of bulk and single-cell RNA-seq 551
data with DTUrtle. Bioinformatics 2021;btab629. 552
Torres-Fernández LA, Emich J, Port Y, Mitschka S, Wöste M, Schneider S, Fietz D, Oud MS, 553
Di Persio S, Neuhaus N, et al. TRIM71 deficiency causes germ cell loss during 554
mouse embryogenesis and is associated with human male infertility. Front Cell Dev 555
Biol 2021;9:658966. 556
Tüttelmann F, Ruckert C, Röpke A. Disorders of spermatogenesis: Perspectives for novel 557
genetic diagnostics after 20 years of unchanged routine. medgen 2018;30:12–20. 558
Vara C, Paytuví-Gallart A, Cuartero Y, Le Dily F, Garcia F, Salvà-Castro J, Gómez-H L, Julià 559
E, Moutinho C, Aiese Cigliano R, et al. Three-Dimensional Genomic Structure and 560
Cohesin Occupancy Correlate with Transcriptional Activity during Spermatogenesis. 561
Cell Reports 2019;28:352-367.e9. 562
Vitting-Seerup K, Sandelin A. The landscape of isoform switches in human cancers. Mol 563
Cancer Res 2017;15:1206–1220. 564
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
24
Wang M, Liu X, Chang G, Chen Y, An G, Yan L, Gao S, Xu Y, Cui Y, Dong J, et al. Single-565
cell RNA sequencing analysis reveals sequential cell fate transition during human 566
spermatogenesis. Cell Stem Cell 2018;23:599-614.e4. 567
Wang Z-Y, Leushkin E, Liechti A, Ovchinnikova S, Mößinger K, Brüning T, Rummel C, 568
Grützner F, Cardoso-Moreira M, Janich P, et al. Transcriptome and translatome co-569
evolution in mammals. Nature 2020;588:642–647. 570
Wiesner T, Lee W, Obenauf AC, Ran L, Murali R, Zhang QF, Wong EWP, Hu W, Scott SN, 571
Shah RH, et al. Alternative transcription initiation leads to expression of a novel ALK 572
isoform in cancer. Nature 2015;526:453–457. 573
Winge SB, Dalgaard MD, Belling KG, Jensen JM, Nielsen JE, Aksglaede L, Schierup MH, 574
Brunak S, Skakkebæk NE, Juul A, et al. Transcriptome analysis of the adult human 575
Klinefelter testis and cellularity-matched controls reveals disturbed differentiation of 576
Sertoli- and Leydig cells. Cell Death Dis 2018;9:586. 577
World Health Organization (2010). WHO Laboratory Manual for the Examination and 578
Processing of Human Semen. 5th ed.: World Health Organization. 579
Wyrwoll MJ, Temel
Ş
G, Nagirnaja L, Oud MS, Lopes AM, Heijden GW van der, Heald JS, 580
Rotte N, Wistuba J, Wöste M, et al. Bi-allelic mutations in M1AP are a frequent cause 581
of meiotic arrest and severely impaired spermatogenesis leading to male infertility. 582
The American Journal of Human Genetics 2020;107:342–351. 583
Wyrwoll MJ, Walree ES van, Hamer G, Rotte N, Motazacker MM, Meijers-Heijboer H, Alders 584
M, Meißner A, Kaminsky E, Wöste M, et al. Bi-allelic variants in DNA mismatch repair 585
proteins MutS Homolog MSH4 and MSH5 cause infertility in both sexes. Human 586
Reproduction 2021;37:178–189. 587
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
25
Table 1: Clinical characteristics of the patient groups. 588
Patient
groups Karyotype Histological paramet ers of tubules Hormonal parameters (normal range) Sperm
mTESE
Score % ES % RS % SC % SG % SCO % TS FSH (1-7U/l) LH (2-10U/l) T(>12nmol/l)
SCO
(n=3) 46,XY 0 0 0 0 0 98.7 (± 1.5) 1.3 (± 1.5) 13.3 (± 4.2) 5.8 (± 2.6) 13.7 (± 3.4) No
SPG
(n=4)
SPG-1, SPG-2,
SPG-3: 46,XY;
SPG-4: n.d. 0 0 0 0 31.0 (±34.6) 34.3 (± 20.7) 35.0 (± 20.6) 20.4 (± 14.2) 13.4 (± 9.7) 16.2 (± 6.9) N o
SPC
(n=3) 46,XY 0 0 0 8 9.3 (± 11.0) 4.7 (±4.6) 1.0 (± 1.0) 5.3 (± 5.5) 5.7 (± 1.3) 5.7 (± 4.5) 9.9 (± 2.4) No
SPD
(n=3) SPD-1, SPD-2:
46,XY; SPD-3:
a
0 0 28.3 (± 2.3) 59.3 (± 18.0) 3.0 ( ±2.0) 1.7 (± 2.9) 8.7 (± 14.2) 7.4 (± 0.9) 3. 7 (± 0.5) 18.7 (± 5.7) No
Normal
(n=3) 46,XY 8-10 87.3 (± 8.6) 3.3 (± 2.5) 8.7 (± 5.7) 0 0 1.0 (± 1.0) 2.5 (± 1.3) 2. 6 (± 1.0) 24.7 (± 2.2) Yes
b
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
26
Figure and table legends 589
Table 1: Clinical characteristics of the patient groups. 590
Data are presented as mean ± standard deviation. Percentage of tubules with elongated 591
spermatids (%ES), round spermatids (%RS), spermatocytes (%SPC), spermatogonia 592
(%SPG), Sertoli cell-only phenotype (%SCO), and tubular shadows (%TS). Score refers to 593
Bergmann and Kliesch score (Bergmann and Kliesch, 2010). Hormonal parameters for 594
follicle stimulating hormone (FSH), luteinizing hormone (LH) and testosterone (T). aPatient 595
SPD-3 had a low number of XXY karyotype mosaicism (47,XXY[2]/46,XY[28]). bTESE 596
results: N-1 had 100/100 sperm, N-2 had an average of 89/100 sperm; No TESE result 597
available for N-3 due to consultation to exclude a malignant tumor. SCO – Sertoli cell-only; 598
SPG – spermatogonial arrest; SPC – spermatocyte arrest; SPD – spermatid arrest; Normal – 599
normal spermatogenesis; n.d. – not determined. 600
Figure 1: Cellular composition of the human testicular biopsies. 601
(A) Schematic illustration depicts the cellular composition of the testicular biopsies with 602
Sertoli cell-only (SCO) arrest at the spermatogonial (SPG), spermatocyte (SPC) and 603
spermatid (SPD) stage as well as samples with normal spermatogenesis (Normal). (B) 604
Stacked barplots represent the proportion of round seminiferous tubules and their most 605
advanced germ cell-type in each sample group. The cellularity of samples from one group is 606
averaged. (C) A principal component analysis (PCA) plot depicts clustering of the total RNA 607
sequenced samples based on the top 500 genes. 608
Figure 2: Co-expression of the DEGs among all samples. 609
(A-D) Heatmaps display the normalized expression counts of the DEGs (rows) of the (A) 610
SCO vs. SPG, (B) SPG vs. SPC, (C) SPC vs. SPD, and (D) SPD vs. Normal group 611
comparisons across all samples (columns) scaled via a row Z-score. Red = increased; blue = 612
decreased. 613
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
27
Figure 3: Examination of germ cell-type specific gene expression at single cell level. 614
(A) UMAP plot depicts 15,546 cells integrated from three patients with obstructive 615
azoospermia and normal spermatogenesis. Sertoli cell, spermatogonia, spermatocyte, early 616
and late spermatid clusters are color-coded, respectively. (B, D, F, H) Vulcano plots of the 617
increased and decreased genes in samples with (B) spermatogonial arrest, (D) 618
spermatocyte, (F) and spermatid arrest, as well as in (H) normal spermatogenesis. (C, E, G, 619
I) Feature plots show the expression of three genes selected for (C) spermatogonia, (E) 620
spermatocytes, (G) round spermatids, (I) and elongated spermatids at single-cell level. 621
Figure 4: Comparison of DEG and DTU gene numbers in all four group comparisons. 622
(A-D) Venn-diagrams display number and proportion of genes that are differentially 623
expressed, have a DTU event, or both in the (A) SCO vs. SPG, (B) SPG vs. SPC, (C) SPC 624
vs. SPD, and (D) SPD vs. Normal group comparisons. Yellow = DEGs, blue = DTU genes. 625
Figure 5: Molecular functions of DEG and DTU genes. 626
Heatmaps of color-coded –log10 p-values display the molecular functions of (A) DEGs and 627
(B) DTU genes per group comparison. Top 20 molecular functions with p-values <0.01 are 628
included. (*) Molecular functions enriched in both, the DEG and DTU gene sets. 629
Figure 6: Transcript biotypes with DTU events. 630
(A) Relative amount of different transcript biotypes with DTU events in each of the four group 631
comparisons in comparison to the transcript biotype annotation from the GENCODE release 632
36 genome annotation based on the GRCh38.p13 genome reference (Frankish et al., 2019). 633
(B-E) Schematic illustration of the exons (grey bars) of the transcript isoforms, which 634
predominantly contribute to the relative change in isoform usage (box plots) in (B) ACTL6A, 635
(C) SPATA4, (D) SYCP3 and (E) MKI67. P-values refer to specific transcripts that 636
significantly drive the change in isoform usage in genes with an overall significant change in 637
transcript usage. * = <0.05, ** = <0.01, *** = <0.001 638
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
28
Supplementary figures and tables 639
Figure S1: Levels of gene expression for selected DTU genes. (Related to figure 6). 640
P-values: *** = <0.001 641
Table SI: List of DEGs of the SCO vs. SPG group comparison. 642
Table SII: Well-known germ cell markers and related publications. 643
Table SIII: List of DEGs of the SPG vs. SPC group comparison. 644
Table SIV: List of DEGs of the SPC vs. SPD group comparison. 645
Table SV: List of DEGs of the SPD vs. Normal group comparison. 646
Table SVI: List of DTU genes of the SCO vs. SPG group comparison. 647
Table SVII: List of DTU genes of the SPG vs. SPC group comparison. 648
Table SVIII: List of DTU genes of the SPC vs. SPD group comparison. 649
Table SIX: List of DTU genes of the SPD vs. Normal group comparison. 650
Table SX: Novel germ cell marker genes and related publications. 651
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint
.CC-BY-NC-ND 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 March 22, 2022. ; https://doi.org/10.1101/2022.03.22.485121doi: bioRxiv preprint