Gene Expansion and Regulatory Rewiring Shape Sex-Biased Evolution of the Mouse Submandibular Gland Secretome

Abstract

Mammalian saliva plays essential roles in digestion, immunity, and host-microbiome interactions, yet its protein composition varies across species and sexes. The evolutionary mechanisms underlying this molecular diversity remain poorly understood. Here, we compared mouse and human salivary gland secretomes at genomic, transcriptomic, and proteomic levels to understand how saliva composition evolves. We performed RNA-seq analysis of the major mouse salivary glands (parotid, submandibular, and sublingual), liver and pancreas from both sexes, compared them with reanalyzed previously published human salivary gland transcriptomes, and integrated them with proteome data of mouse and human whole saliva. We found that evolution of gene expression in mouse salivary glands is driven by rapid gene turnover and sexual dimorphism. In the submandibular and sublingual glands, respectively, 68% and 73% of expression from genes encoding secreted proteins derives from lineage-specific genes that lack one-to-one human orthologs. Mouse submandibular gland shows striking sexual dimorphism, with 1537 tissue specific sex-biased genes, five times higher than in the liver, a classic model of sex-biased expression. These genes cluster in regions shaped by recent gene duplication, such as the kallikrein gene cluster, a mouse-specific expansion that accounts for ~16.4% of male-biased submandibular expression. Our analyses suggest that this bias arises through regulatory changes that are expanded by gene duplication, including the spread of a testosterone-associated regulatory motif and the expansion of a shared chromatin domain that promotes coordinated gene regulation. Our results reveal how lineage-specific gene duplication and regulatory rewiring drive rapid, sex-specific evolution of the mammalian salivary gland secretome.

INTRODUCTION

Diet and pathogen exposure are major environmental drivers of mammalian evolution (Price et al. 2012; Enard et al. 2016). Because the oral cavity represents the body’s first interface with the external environment, saliva plays a central role in mediating these pressures, balancing protection against pathogens with responsiveness to dietary variation. Saliva also contributes to oral homeostasis by lubricating and cleansing the mouth, protecting tooth enamel, enabling taste perception, and initiating digestion (Ruhl 2012; Pajic et al. 2022). These functions are carried out by a complex mixture of proteins secreted primarily by the three major salivary glands, parotid, submandibular, and sublingual glands (Dawes et al. 2015; Chibly et al. 2022). Importantly, the salivary gland secretome, i.e., the entirety of proteins secreted by the salivary glands, reflects adaptive tuning to species-specific ecological contexts, with protein repertoires shaped by dietary specialization (Pajic et al. 2019) and pathogen exposure (Xu et al. 2016). Together, the integration of immune, digestive, and sensory functions makes saliva an exceptional system for investigating how environmental pressures drive the evolution of mammalian genomes.

Previous studies suggested that saliva-related genes undergo rapid evolution, mainly to accommodate environmental and dietary changes which accompany speciation (Pajic et al. 2019). Recent locus-specific comparative genomic studies showed that genes expressed in the salivary glands that encode secreted salivary proteins differ substantially between mice and humans (Pajic et al. 2025). For instance, analyses of the secretory calcium-binding phosphoprotein (SCPP) locus indicated that multiple human salivary genes lack orthologs in mice. Proteomic analyses also have revealed that the composition of saliva proteins differs markedly between humans and mice (Blanchard et al. 2015; De Sousa-Pereira et al. 2015), and comparative studies further show substantial interspecies differences in their major components (Karn et al. 2013). These findings suggest that human and mouse salivary glands differ greatly in both gene content and expression profiles, supporting the view that the salivary gland secretome evolves rapidly in response to environmental and pathogenic pressures. Despite these insights, a systematic, secretome-wide investigation of the evolutionary dynamics and regulatory mechanisms underlying these differences has not yet been conducted.

In contrast to humans, the salivary glands in many other mammalian species are different between sexes. Sexual dimorphism in salivary glands has been well documented in rodents (Ikemoto and Matsushima 1984), and pronounced sex differences have been shown in salivary glands of mice (Chai et al. 1993), which are not known to occur in humans. Generally, sex-biased genes across multiple tissues in mammals are poorly conserved across species (Naqvi et al. 2019). In particular, the mouse submandibular gland exhibits striking sex-specific differences in cellular composition, secretory activity, and hormonal regulation (Hatakeyama et al. 1987; Konttinen et al. 2010; Mukaibo et al. 2019). Although substantial progress has been made in understanding saliva composition and its role in oral biology in both species, the evolutionary basis of lineage- and sex-specific variation in the salivary gland secretome remains underexplored.

Here, we investigate the mechanisms underlying the rapid diversification of the salivary gland secretome. Using comparative analyses between humans and mice, we assess the conservation of genes highly expressed in salivary glands and determine the contribution of lineage-specific genes to the salivary gland secretome. We further compare male and female mice to identify the genomic basis of sex-biased expression in the submandibular gland. By integrating transcriptomic, proteomic, and genomic data, we delineate the evolutionary processes driving salivary diversity and sexual dimorphism.

RESULTS

The highest proportion of gene expression in mouse submandibular and sublingual glands derives from lineage-specific genes

To investigate differences in salivary gene expression across commonly used mouse models and compare them to humans, we performed transcriptomic profiling of the parotid, submandibular, and sublingual glands, along with the pancreas and liver as reference tissues, in CD1 and C57BL/6 strains (n = 3 per sex per strain; Tables S1). Both CD1 and C57BL/6 strains are widely used in laboratory studies of salivary glands, and were included as a representative of an inbred and outbred mouse strain, respectively, to strengthen the broader applicability of our findings (Nguyen et al. 2009; Gervais et al. 2015; Hsieh et al. 2017). In parallel, we conducted mass spectrometry-based proteomic profiling of whole saliva from ten C57BL/6 mice, five males and five females (Table S2). Principal component analysis (PCA) of the transcriptomic data shows that samples cluster strongly by tissue of origin, with clear separation among the three salivary glands (Figure S1), supporting the robustness of the dataset.

To characterize gland-specific expression patterns and contribution to saliva proteome, we calculated mean transcript abundance herein estimated in transcripts per million (TPM). Each gland displayed a distinct transcriptional signature, typically dominated by a small subset of highly expressed, gland-specific genes that code for secreted proteins, consistent with previous work (Gao et al. 2018). We then identified the 30 most highly expressed genes encoding secreted proteins in each mouse gland (see Methods; Figure 1a, Table S3). These genes account for 69.3%, 64.9%, and 35.6% of total transcript abundance in the parotid, submandibular, and sublingual glands, respectively. When considering only transcripts encoding secreted proteins, these top 30 genes explain an even larger fraction of expression: 95.7%, 95.9%, and 87.0% in the respective glands.

Figure 1.

Most mouse genes coding for secreted proteins in salivary glands are lineage-specific. (a) Bubble plots showing the top 30 most highly transcribed genes that code for secreted proteins in each major mouse salivary gland. Bubble size represents mean transcript abundance herein estimated in transcripts per million (TPM), and colors denote the type of orthology: one-to-one (green), lineage-specific (maroon), and other orthologous relationships (including many-to-many and one-to-one) (gray). (b) Same analysis as in A, here for human salivary glands. (c) Proportion of total gene expression (mean TPM) of genes that code for secreted proteins attributed to each orthology category in mouse and human. (d) Saliva proteins separated by SDS-PAGE and stained with Coomassie blue and Periodic acid Schiff’s stain for pooled whole saliva of males (M, n=5) and females (F, n=5) for mice and humans. Pink bands are glycosylated proteins revealed by periodic acid Schiff stain. Note the shift in gelelectrophoretic mobility of the high-molecular-weight glycoprotein band from 150 kDa in males to 130 kDa in females. MW denotes molecular weight markers. (e) Correlation between gene expression (Log₁₀(1+Mean TPM)) and protein abundance (Log₁₀(1+Mean iBAQ)) for genes that code for secreted proteins in the submandibular gland. The right panel highlights the top 100 most highly expressed genes, for which RNA and protein expression are more concordant. ρ denotes Spearman’s correlation. *denotes p<0.05.

Next, we categorized mouse salivary gland genes based on their evolutionary relationships with human genes, either as one-to-one orthologs, other orthologous relationships (i.e., one-to-many or many-to-many; see Methods), or lineage-specific genes (Figure 1a). We applied the same classification to human salivary gland transcriptomes from our previous study (Saitou et al. 2020), enabling direct comparison between species (Figure 1b, Table S4). The proportion of expression originating from one-to-one orthologs that code for secreted proteins are remarkably low for sublingual and submandibular glands, approximately 8% and 11%, respectively (Figure 1c). In contrast, the parotid gland displays a higher proportion of expression of one-to-one orthologs (77.58%), suggesting greater evolutionary conservation in its core secretory functions.

We recently hypothesized, based on a locus-specific analysis, that salivary glands evolve faster than other tissues because of shifting dietary and pathogenic evolutionary pressures (Pajic et al. 2025). Here, we tested this hypothesis more comprehensively at the transcriptome-wide level by comparing the expression levels of one-to-one orthologs between human and mouse, which are expected to maintain comparable expression levels across species (Kryuchkova-Mostacci and Robinson-Rechavi 2016). This expectation was met for genes that code for secreted proteins in pancreas and liver, which showed strong correlation of expression between human and mouse transcriptomes (Spearman’s ρ = 0.64, p = 1.68 × 10⁻¹⁴⁷ and ρ = 0.78, p = 8.55 × 10⁻²⁶⁷, respectively). In contrast, genes that code for secreted proteins in the salivary glands exhibited weaker mouse-human correlations (parotid: ρ = 0.575, p = 5.90 × 10⁻¹¹²; submandibular: ρ = 0.58, p = 2.71 × 10⁻¹¹⁴; sublingual: ρ = 0.606, p = 2.27 × 10⁻¹²⁷; Figure S2).

To determine whether these transcriptomic differences for genes encoding secreted proteins are also reflected at the proteomic level, we performed gel electrophoresis of whole saliva (Figure 1d; Figure S3). The resulting protein and glycoprotein profiles revealed striking qualitative and quantitative differences in banding patterns between human and mouse saliva, with only a few bands appearing to match between the two species. Particularly noteworthy is the observed shift in gel electrophoretic mobility of the high-molecular-weight glycoprotein band from 150 kDa in males to 130 kDa in females. This protein band was identified by mass spectrometry (MS/MS) analysis as Muc19 (Table S5).

In parallel, we asked the extent to which the transcriptomic variation observed translates into variation in the salivary proteome. To assess this, we compared gene expression in the salivary glands of each mouse strain with the abundance of the corresponding proteins in their saliva for 467 gene-protein pairs (see Methods). We observed a significant positive correlation for each gland (parotid: ρ = 0.362, p = 7.29 × 10⁻¹⁶; sublingual: ρ = 0.268, p = 3.76 × 10⁻⁹; submandibular: ρ = 0.304, p = 1.81 × 10⁻¹¹) (Figure S4). Notably, correlation strength increased among the most highly expressed genes that code for secreted proteins, with stronger associations observed when restricting the analysis to the top 100 transcripts per gland (parotid: ρ = 0.492, p = 2.06×10⁻⁷; sublingual: ρ = 0.528, p = 1.65×10⁻⁸; submandibular: ρ = 0.582, p = 2.10×10⁻¹⁰) (Figure 1e). These protein-level analyses support the notion that transcript abundance in the salivary glands directly affects protein levels in saliva, and that highly expressed genes encoding secreted proteins have an outsized influence on shaping the oral proteome.

Our combined transcriptomic and proteomic analyses support a model in which mammalian salivary glands function as specialized organs that produce a limited set of highly abundant secreted proteins. Our data further support the hypothesis that the genes coding for secreted proteins in the salivary glands show faster turnover than those in liver and pancreas, a phenomenon more pronounced among the top expressed genes encoding secreted proteins. Transcriptomic data from individual salivary glands accurately predict their relative contributions to the oral proteome, consistent with previous observations in humans (Saitou et al. 2020). Overall, our results reveal fundamental differences in expression profiles of genes coding for secreted proteins in the salivary glands of humans and mice, consistent with a rapid evolution of the salivary gland secretome.

The submandibular gland is the major contributor to sexual dimorphism in saliva

Previous studies have documented sexual dimorphism in the postnatal development and cellular composition of the mouse submandibular gland, most notably the male-specific expansion of granular convoluted tubules (ducts) (Caramia 1966; Pardini and Taga 1996). These morphological differences are known to influence salivary protein output and gene expression (Treister et al. 2005; Mukaibo et al. 2019). Thus, we hypothesize that the submandibular gland contributes to sexual dimorphism in the salivary proteome. Analysis of our mass spectrometry dataset identified 368 and 152 proteins with significantly male- and female-biased protein abundances, respectively (p_adj<0.05 and |log₂FoldChange| > 0.5, Table S6). To determine the transcriptional basis of this protein-level dimorphism, we next examined whether sex-biased gene expression across individual salivary glands contributes to the observed differences in salivary protein abundance.

To systematically address this hypothesis, we performed comparative transcriptomic analyses of sex-biased gene expression across mouse salivary glands, liver, and pancreas (Table S7). The submandibular gland exhibited the largest number of tissue-specific differentially expressed genes (DEGs, FoldEnrichment>1, p_adj<0.05) between sexes (1,537 genes) with other salivary glands exhibiting markedly fewer tissue-specific DEGs (Figure 2a), consistent with previous reports (Mukaibo et al. 2019). What distinguishes our analysis is the direct cross-organ comparison. We found that the submandibular gland, analyzed within the same individuals, harbors more than fivefold sex-biased DEGs than liver and ninety-sixfold more than pancreas. These results place the submandibular gland among the most sexually dimorphic tissues in mice, clearly exceeding even the liver, which is widely regarded as the canonical model for studying sexual dimorphism in mammalian gene expression (Lai et al. 2016; Nikkanen et al. 2022). Further, a direct comparison of the extent of sexual differentiation in submandibular transcriptomes shows a dramatic difference between mice and humans (Figure 2b, Table S8).

Figure 2.

The submandibular gland is the major contributor to sexual dimorphism of the salivary gland secretome. (a) Upset plot showing that the submandibular (SM) gland harbors more exclusive DEGs between mouse sexes than other major salivary glands (parotid [PAR], sublingual [SL]) or non-salivary tissues (liver [LIV], pancreas [PANC]). Sex-linked genes (Ddx3y, Eif2s3y, Kdm5d, Uty, Xist) that are differentially expressed in all tissues are indicated with an arrow. (b) Volcano plot of differentially expressed genes in the mouse (left) and human (right) submandibular gland. Arrows indicate the total number of significantly upregulated genes in females (pink) and males (blue). Pink dots represent genes significantly upregulated in females, and blue dots represent genes significantly upregulated in males. Significance is measured in Log₁₀(p_adj). (c). Correlation plot of protein abundances in males and females quantified in Log₁₀(1+Mean iBAQ). Each dot represents a differentially abundant protein. Female protein abundance is shown on the y-axis and male protein abundance on the x-axis. The dashed line indicates one-to-one correspondence. Colored points denote sexually dimorphic proteins that are also differentially expressed genes in each gland. The inset panel shows the number of differentially abundant proteins whose genes are also differentially expressed in each gland, with the highest contribution observed in the submandibular gland. (d) Immunofluorescence staining showing differences in expression patterns between female (top) and male (bottom) submandibular glands. NGF (red). DAPI marks nuclei (blue). (e) Sex-biased expression of the sialyltransferase St3gal1 is associated with differences in Muc19 glycosylation. Left: Boxplot showing transcript abundance of St3gal1 (TPM) in male and female mice. One female outlier (TPM = 443) is shown as a triangle and was kept for statistical calculations. Right: SDS-PAGE and lectin blot analysis of Muc19 in pooled male (M) and female (F) saliva samples. PNA detects unsialylated core O-glycans, whereas MAL-II detects terminal α2,3-linked sialic acids. Differences in lectin binding and electrophoretic mobility indicate higher sialylation of female Muc19, consistent with increased St3gal1 expression.

Next, we tested if most sexually dimorphic saliva proteins originate from genes with sex-biased expression in the submandibular gland. Protein abundance in saliva is influenced by many factors, including RNA abundance, translational efficiency, posttranslational modifications, and peptide degradation. Despite this complexity, we found that 70 of the 507 sexually dimorphic salivary proteins are encoded by genes that are differentially expressed in the submandibular gland, compared with only one gene in the parotid gland and one in the sublingual gland (Figure 2c). These results show that the submandibular gland is the primary contributor to the sexual dimorphism observed in the protein composition of mouse saliva.

To further validate these results, we used immunohistochemical staining with well-characterized antibodies to examine the expression of nerve growth factor (Ngf), one representative sex-biased protein, which showed pronounced sexual dimorphism in both submandibular RNA-seq and salivary proteome analyses (Bie et al. 2010; Wong et al. 2019). Examination of submandibular gland sections from male and female mice revealed clear immunohistochemical differences between sexes (Figure 2d), supporting that pronounced sexual dimorphism at the gene expression and secreted protein levels, are also manifested at the cellular level.

In addition to transcriptional differences, post-translational modifications, such as glycosylation, may further contribute to sex-specific differences in the biochemical and functional properties of salivary proteins. In particular, sialic acids are negatively charged terminal glycan residues on most mammalian glycan chain and are commonly present on glycans of secreted glycoproteins, most prominently on mucins, where they influence the biophysical and tribological properties of these molecules and play important roles in their interaction with microbes and viruses (Varki and Schauer 2009; Hiremath and Wagner 2026). Consistent with this possibility, we observed differential expression of several genes involved in sialylation pathways, including the sialyltransferases St3gal1, St3gal4, St3gal5, St6galnac1, and St6galnac2, between males and females (Figure S5). Notably, in contrast to mice, in humans none of these sialyltransferases are differentially expressed between sexes in the submandibular gland.

To test whether these transcriptional differences correspond to differences in protein glycosylation, we investigated the prominent glycoprotein band in mouse saliva that shifts in electrophoretic mobility from ~150 kDa in male saliva to ~130 kDa in female saliva (see Figure 1d). Analysis by MS/MS of the major protein present in these bands excised from the gel identified it as mucin Muc19. Lectin blotting using PNA, revealed strong binding to the band in male saliva, but lack of binding to the band in females, whereas lectin blotting with MAL-II showed strong binding to the band in females and weak binding to the band in male saliva. Together, these results indicate presence of unsialylated core O-glycans (Galβ1-3GalNAc) on male Muc19 and a higher proportion terminal α2,3-linked sialylation on female Muc19 (Figure 2e). The higher degree of Muc19 sialylation in female mice is consistent with the observed electrophoretic shift of the Muc19 band from ~150 kDa in male saliva to ~130 kDa in female saliva due to the added electronegativity conferred by the higher number of negatively charged sialic acid molecules on the female mucin glycan chains. These results suggest that, in addition to sex-specific differences affecting the transcription of genes encoding secreted proteins in the submandibular gland directly, sex-specific post-translational modifications also contribute substantially to the sexual dimorphism observed in mouse saliva, and are caused by sex-specific transcription differences in modifying enzymes, thereby affecting secreted protein products indirectly in a sex-specific manner.

Together, these findings demonstrate that the submandibular gland, by far the most sexually dimorphic tissue in our dataset, is the primary contributor to the strong sexual dimorphism observed in the salivary proteome, with additional contributions from sex-specific post-translational modifications.

Klk gene family expansion underlies sex-biased expression in mice

The profound sexual dimorphism in the mouse submandibular gland is not recapitulated in humans (Figure 2b). This observation motivates a closer examination of the evolutionary processes shaping sex-biased gene expression in the mouse submandibular gland. In particular, we asked whether this extensive sex bias might reflect coordinated regulation of clusters of genes within specific genomic regions or a broad, genome-wide effect.

To address this, we first searched for physically clustered genes that may be co-regulated to express in a sex-specific manner. Specifically, we performed a genomic clustering analysis of sex-biased DEGs in the mouse submandibular gland using 100-kb windows (see Methods; Table S9). We found that the Klk gene family forms the largest cluster of sex-biased genes in the genome and is a significant outlier relative to other clusters (p < 0.001; Figure 3a). Fourteen members of the kallikrein (Klk) family (Klk15, Klk1b8, Klk1b1, Klk1b9, Klk1b11, Klk1b26, Klk1b27, Klk1b21, Klk1b22, Klk1b16, Klk1b24, Klk1b3, Klk1b4, and Klk1b5) exhibit strong male-biased expression, whereas one member (Klk1) shows female-biased expression. Altogether, Klk genes make up ~16.4% of total gene expression in male submandibular glands, compared with only 2.5% in female ones (Figure 3b).

Figure 3.

The Klk gene family forms the largest cluster of sex-biased transcripts and proteins in mouse submandibular glands, independent of differences in cell abundance. (a) Genomic distribution of genes encoding for sex-biased proteins (top, green) and sex-biased transcripts (bottom, magenta) along mouse chromosome 7 in the submandibular gland. Points represent differentially abundant proteins or differentially expressed genes, plotted by genomic position and log2 fold difference between sexes. The shaded region highlights the Klk locus. Density tracks below each scatter plot show kernel-smoothed spatial density of significant genes or proteins along the chromosome 7. Blue and pink points indicate genes whose sex bias at the transcript level is also supported by proteomic differences in males and females, respectively. (b) Relative expression of Klk genes in male and female submandibular glands. Boxes show distribution across biological replicates; points represent individual samples. ** indicates p < 0.001. (c) Comparison of female versus male Klk gene expression levels. All Klk1b paralogs show >5-fold higher expression in males, indicating that sex-biased expression is not explained by differences in glandular cell composition. (d) Immunofluorescence staining of submandibular glands from female (top) and male (bottom). Klk1 protein is shown in green; nuclei are counterstained with DAPI (blue), illustrating stronger Klk1 expression in females.

We then asked whether the clustering of sex-biased genes can also be observed at the proteomic level. Of the Klk genes showing sex-biased expression, 11 out of 15 also display concordant sex-biased differences in protein abundance in saliva. Using genomic coordinates of genes encoding differentially abundant proteins, we observed that the Klk genes exhibiting sex-biased protein abundance are clustered on chromosome 7 (Figure 3a), consistent with the transcriptomic clustering pattern. Together, these results indicate that the kallikrein locus represents a genome-wide outlier cluster of coordinated, sex-biased regulation at both the transcriptomic and proteomic levels.

Previous studies have already shown that Klk genes exhibit male-biased expression in the mouse submandibular gland (Stopka et al. 2016; Mukaibo et al. 2019) and in saliva (Penschow et al. 1991; Karn and Laukaitis 2011; Blanchard et al. 2015; Moustardas et al. 2021). Single-cell transcriptomic analyses further indicate that Klk genes are specifically expressed in granular convoluted tubule (GCT) and acinar cells (Horeth et al. 2021), which are approximately 5- and 1.5-fold more abundant in males, respectively (Pardini and Taga 1996). This raises the question of whether differences in cell abundance between male and female submandibular glands alone can explain the observed sex-biased expression. If expression per cell were equivalent between sexes, differences in cell-type abundance could, in principle, account for higher Klk expression in males. However, when we quantified overall expression levels, we found that male-female differences exceeded by at least 5.5-fold for all differentially expressed Klks, indicating that the observed expression differences cannot be explained by cell-type proportions alone (Figure 3c).

One remarkable exception is that the observed male-biased pattern for Klk1b genes is not observed in Klk1. In fact, we found that Klk1 is expressed significantly higher in female submandibular glands despite the fact that Klk1b genes were shown to be duplicates of Klk1 (Lawrence et al. 2010; Pavlopoulou et al. 2010), and that Klk1 is also known to be expressed in acinar and GCT cells which are much more abundant in males. To verify this transcriptional level observation at the protein and cellular levels, we performed immunohistochemical staining of both male and female submandibular gland sections for the female-biased Klk1 gene, one of the few Klk family members for which a robust antibody is available (Figure 3d). Consistent with the gene expression pattern, protein abundance was significantly higher in female granular convoluted tubules (Figure S6), resulting in greater overall Klk1 expression in females despite the higher abundance of these cells in males. This observation suggests evolution of complex regulatory mechanisms in the Klk locus, distinct for Klk1 and its paralogs.

Together, these results indicate that differences in cellular composition alone are insufficient to explain sex-biased Klk expression and suggest additional genetic or epigenetic regulation that drive this dimorphism.

Sex-specific Klk expression is shaped by hormonal regulation, and lineage-specific 3D genomic architecture

To investigate the genetic basis of male-biased expression at the Klk locus, we first generated a gene tree incorporating gene family expansion and contraction estimates inferred by CAFE 5 and obtained from Ensembl (Mendes et al. 2021). We found that placental mammals, including other rodents, harbor up to 15 Klk genes, whereas mice and rats harbor approximately 17-25, indicating a burst of duplications in the murid lineage (Figure S7). Specifically, there are 13 additional tandemly duplicated copies in mice compared to the human genome (Figure 4a), all duplicated from Klk1 (Lawrence et al. 2010; Pavlopoulou et al. 2010) and with male-biased expression.

Figure 4.

Expansion of the mouse Klk locus is associated with sex-biased expression and reorganization of local 3D genome structure. (a) Schematic comparison of the KLK/Klk locus organization in humans (top) and mice (bottom). Circles represent individual KLK/Klk genes ordered by genomic position and transcriptional orientation. Genes showing female- or male-biased expression are indicated in pink and blue, respectively. Orange bars denote putative transcription factor binding motifs shared across the expanded mouse cluster. The mouse lineage shows extensive tandem duplication of Klk genes compared with the more compact human locus. (b) Hi-C contact map of the mouse Klk region generated from B-cell lymphoma cell line, revealing two adjacent topologically associating domains (TADs). The left TAD predominantly contains non–sex-biased genes, whereas the right TAD encompasses the expanded cluster of sex-biased Klk paralogs. Gene annotations are shown below the contact map together with chromatin accessibility and regulatory marks (ATAC-seq, H3K4me3, CTCF, and RAD21), highlighting active promoters and architectural features associated with TAD boundaries. (c) Hi-C contact maps of the mouse Klk locus reconstructed from male Sertoli cells (top) and female granulosa cells (bottom). Dashed vertical lines indicate shifts in TAD boundaries between the two cell types. Genes located within these alternative domain configurations are highlighted, suggesting that differences in higher-order chromatin organization may influence sex-specific regulation of subsets of Klk paralogs. (d) Conceptual model illustrating how tandem gene duplication at the Klk locus may have been accompanied by expansion and restructuring of the surrounding TAD landscape.

Next, we asked whether the large cluster of differentially expressed genes containing Klk genes in the submandibular gland share common regulatory sequences that could underlie their coordinated male-specific expression. To address this, we performed a de novo motif enrichment analysis on this cluster, identifying overrepresented sequence motifs located near genes (see Methods). We found an enrichment (p = 1 × 10⁻⁴²) for a conserved CTGATCCTGTTC motif located within 30 bp of the transcription start site in all sex-biased Klk genes except Klk15 (Figure 4a; see Methods; Tables S10 and S11). Previous work showed that this motif is a binding site for Ghrl2, a developmental transcription factor whose activity is sex-biased and correlated with testosterone levels (Paltoglou et al. 2017; Reese et al. 2019). Consistent with this, kallikreins are well documented as androgen-regulated genes in the mouse submandibular gland (Van Leeuwen et al. 1987; Kurabuchi et al. 2001; Lawrence et al. 2010). In contrast, the motif is absent from Klk genes that do not show sex-biased expression. Moreover, none of the human KLK genes present this motif at the orthologous locus (Table S11). Given that the Klk1b paralogs arose through lineage-specific duplications of an ancestral Klk1 gene (Lawrence et al. 2010; Pavlopoulou et al. 2010), our results support an evolutionary scenario in which the CTGATCCTGTTC motif emerged in the ancestral murine Klk1 gene before subsequent gene expansion. Subsequent duplications may have increased gene expression through this regulatory mechanism, thereby amplifying male-biased expression in the submandibular gland.

Previous work in humans identified two distinct topologically associating domains (TADs) that structure the regulatory landscape of the KLK gene family, both demarcated by CTCF binding sites (Khoury et al. 2020). We therefore asked whether the lineage-specific duplications in mice resulted in the formation of a new or expanded the existing topologically associating domain (TAD). To address this, we analyzed publicly available Hi-C, ATAC-seq, and ChIP-seq datasets, including histone modification profiles, from the mouse B-cell lymphoma cell line (see Methods). Consistent with the human architecture, we identified two TADs at the mouse Klk locus of which the second TAD is markedly expanded, encompassing 15 genes, including all those that show sex-biased expression (Figure 4b). Chromatin accessibility (ATAC-seq), histone modification (H3K27ac), and transcriptional anchor protein binding (Ctcf, Rad21) profiles delineate the borders of these TADs.

Although the TAD boundaries in mice largely mirror those observed in humans and encompass all sex-biased Klk genes, it remains unclear how Klk1 retains both the regulatory motif and its position within the same TAD. To explore whether sex-specific chromatin organization might contribute to this regulatory difference, we compared TAD boundaries between male and female cells. Using publicly available Hi-C data from male Sertoli cells and female granulosa cells, we reconstructed TAD structures and compared their boundaries (Figure 4c). This analysis revealed a slight boundary shift that coincides with the region showing male-specific expression, suggesting that Klk1 in females may be regulated within a different TAD context. These findings indicate that regulatory control at this locus is more complex than previously appreciated, and that the data presented here are insufficient to fully explain the expression pattern of Klk1.

Together, these results suggest that lineage-specific Klk gene duplications have reshaped the ancestral TAD structure shared with humans (Figure 4d), expanding it in mice in a sex-specific manner.

DISCUSSION AND CONCLUSION

In this study, we investigated the evolutionary forces shaping the composition of the salivary gland secretome in mice. We show that most highly expressed genes encoding secreted saliva proteins in mice are lineage-specific and lack direct human orthologs. We further identify the submandibular gland as the focal point of sex-specific expression. The kallikrein (Klk) gene family in mice accounts for ~16.4% of male-biased expression, and we discuss how lineage-specific gene family expansion, emergence of novel regulatory motifs, and topological clustering contribute to the pronounced sex differences in expression. Together, the findings illustrate how gene duplication and regulatory rewiring drive sex-biased gene expression and contribute to the rapid evolution of the mammalian salivary gland secretome in different lineages.

When we examined sex differences in mouse saliva, we found that the submandibular gland is the main contributor to the protein differences observed between males and females, consistent with previous results (Treister et al. 2005; Mukaibo et al. 2019). This male bias in submandibular expression is particularly notable when compared with the liver, the most studied tissue for sex differences in mice (Nikkanen et al. 2022), where the degree of dimorphism is lower. We further found that this bias can be traced to the lineage-specific gene duplication clusters identified in our analysis, with the kallikrein (Klk) cluster standing out as a single gene family accounting for ~16.4% of male-biased expression in the submandibular gland. The sex-biased expression of kallikrein genes is well documented in the literature, aligning with the patterns we observe here (Karn and Laukaitis 2011; Stopka et al. 2016; Moustardas et al. 2021). Specifically, our results support a model in which the expression of a hypothetical ancestral Klk gene in the murine lineage was subsequently amplified through a burst of tandem duplications within the same topologically associating domain (TAD), a process that could have both accompanied and facilitated developmental shifts in the male submandibular gland. This mechanistic framework suggests a broader evolutionary trend: rapid regulatory change across multiple genes can emerge through de novo motif evolution at a single locus, followed by its propagation via gene duplication within the same 3D genomic domain (Figure 5). A similar trend for motif evolution has recently been observed at the primate amylase locus (Karageorgiou et al. 2025)

Figure 5.

Proposed model for the evolutionary trajectory of mouse kallikrein (Klk) genes in salivary glands. Gene duplication created multiple Klk paralogs, which subsequently acquired new regulatory motifs. These regulatory changes enabled functional diversification, including interaction with Ghrl2. Further duplications occurred within a conserved topologically associating domain (TAD), promoting coordinated regulation of the expanded gene set. This combination of gene family expansion and regulatory rewiring led to increased male-biased expression in the submandibular gland.

The pronounced male-biased expression of kallikreins in saliva raises questions about their phenotypic impact in male mice. A prevailing hypothesis proposes that the presence of sex-specific proteins in saliva in male mice is related to improved healing through wound licking behavior after male-male aggression (Hutson et al. 1979; Stopka et al. 2016). While this hypothesis may be valid for the higher expression of wound healing related saliva proteins in male mice, such as epidermal growth factor and fibroblast growth factor (Hirata and Orth 1979; Brown et al. 2020), it may not extend to the kallikrein gene family. An alternative hypothesis has been posited by Karn and Laukaitis where sex-biased expression of proteins in rodent saliva is explained by either signaling male dominance among rival males or mate identification through pheromone-like mechanisms (Karn and Laukaitis 2011). Our results cannot confirm or falsify either of these hypotheses, but they may provide a mechanistic base for formulating testable hypotheses approaching the questions from a molecular genetic level.

While we propose a hypothesis of co-regulation among male-biased Klk genes, our results do not fully explain the expression pattern of the only female-biased Klk gene (Klk1). Previous studies suggest that additional regulatory mechanisms contribute to the regulation of Klk transcripts, including microRNA-mediated silencing (Kurihara et al. 2011), highlighting the regulatory complexity of this locus. Moreover, Klk1 has been reported to be regulated by estrogen in mice through mechanisms distinct from those described for other Klk family members (Iwasaki et al. 2023). Future studies integrating genomic, regulatory, and three-dimensional genome architecture data from salivary glands of the same individuals ideally using a single-cell resolution will be necessary to validate our proposed model of co-evolution among regulatory motifs, chromatin organization, gene duplications, and cell-type composition.

The biological functions and physiological relevance of the extensive species- and sex-specific differences in salivary gene and protein expression remain largely unknown. While addressing these functions was beyond the scope of the present study, our results provide a comprehensive catalogue of candidate genes and proteins that can now be used to test hypotheses about the roles of salivary gland-derived proteins in physiology, immunity, and behavior. One novel observation is the sex-specific glycosylation of the Muc19 protein in mouse saliva, suggesting that differences in glycosylation machinery, particularly the differential expression of sialyltransferases between males and females, contribute to sex-specific modification of secreted mucins. This stands in contrast to humans, where MUC19 expression is low in salivary glands and no clear sexual dimorphism has been reported. Notably, genetic variation in human MUC19 has been linked to complex introgression patterns in hominins and signatures of local selection in Indigenous American populations (Villanea et al. 2025), marking this gene as a hotspot of adaptive evolution. Together, our observations provide a foundation for future work aimed at dissecting the regulatory basis of species-specific expression, the molecular mechanisms underlying sex-specific glycosylation, and the evolutionary processes shaping Muc19 variation across mammals.

Of importance for biomedical research, our findings raise more fundamental questions about the suitability of the mouse as a standard model for studying human salivary gland biology, a concern that has also been noted previously (Maruyama et al. 2019). Focusing on salivary glands and their secreted products, we show that mice and humans differ fundamentally in both the expression patterns and composition of salivary proteins, including their post-translational modifications. With respect to disease relevance in humans, we found that several highly expressed Sjögren’s-related genes (Horeth et al. 2021), including Dcpp1, Scgb2b26, and Scgb2b27, are mouse-specific and lack human counterparts, thus highlighting the importance of validating translational studies involving mice salivary glands and saliva in human models. In addition, our results underscore the importance of explicitly accounting for sex as a biological variable in studies of oral and salivary gland biology, particularly in mouse models.

Overall, this study, although limited to a comparison between two species, contributes to the emerging view that saliva exhibits considerable evolutionary plasticity, enabling accelerated evolution under selective pressures (Ruhl 2012; Pajic et al. 2019; Pajic et al. 2025) and offers novel additional mechanistic hypotheses to explain these trends. Furthermore, our results are consistent with the hypothesis of developmental systems drift (True and Haag 2001), whereby similar phenotypic traits across species evolve through distinct genetic or gene-regulatory mechanisms. Although saliva performs broadly conserved functions across mammals, primarily related to digestion, protection of oral tissues, microbial homeostasis, and pathogen defense, our results suggest that these functions may be achieved through different regulatory architectures across species. In this context, the murine Klk locus provides a useful model for investigating how gene duplication, regulatory motif evolution, and 3D genome organization can shape the evolution of sex-biased gene regulation and the emergence of novel sexually dimorphic traits. Expanding similar analyses across multiple mammalian species will be crucial for uncovering general principles governing the evolution of salivary gland gene regulation and function.

METHODS

Sample collection for RNA-seq

A total of 12 mice (3 months old) were used, consisting of three biological replicates for each strain-sex combination: 3 female CD1, 3 male CD1, 3 female C57BL/6, and 3 male C57BL/6 (Figure S8). CD1 and C57BL/6 mice were selected as representative outbred and inbred strains, respectively, as both are widely used in laboratory studies of salivary glands (Nguyen et al. 2009; Gervais et al. 2015; Hsieh et al. 2017). Mouse samples were provided by UCSF. Animals were euthanized following the approved protocol AN201280-00I. From each mouse, we collected five glands: the three major salivary glands (parotid, sublingual, and submandibular glands), as well as the pancreas and liver, which were used as reference tissues.

RNA-sequencing and analysis of mouse transcriptomes and differential analysis

Total RNA was extracted from each tissue, and RNA-seq libraries were prepared by Azenta Life Sciences (GENEWIZ) using their Standard/Strand-Specific RNA-Seq service with poly(A) selection for mRNA enrichment, followed by sequencing on an Illumina^® NovaSeq^™ platform (2×150 bp). Raw reads were processed with Trim Galore (Krueger et al. 2023), which uses cutadapt for adapter trimming and FastQC for quality assessment, to remove low-quality bases and adapter sequences. Cleaned reads were then aligned to the Mus musculus reference genome (GRCm39) using HISAT2 (Kim et al. 2019), and gene-level quantification was performed with featureCounts (Liao et al. 2014), which assigns aligned reads to annotated genes and produces a matrix of raw read counts per gene.

Six parotid samples (two male and one female CD1, and three female C57BL/6) exhibited high expression of genes associated with muscle function and contraction, which is unusual in parotid gland transcriptomes. These samples were excluded from downstream analyses, as their expression profiles suggested possible contamination during tissue dissection. Next, we measured differential gene expression between sexes for each tissue using DESeq2 (Love et al. 2014), applying a significance threshold of p_adj< 0.05 and |log2 fold change| > 1. DESeq2 models raw read counts for each gene and compares expression levels between sexes, estimating differential expression while accounting for biological variability. Counts are internally normalized by library size and sequencing depth to allow accurate comparisons across samples, and significantly differentially expressed genes (DEGs) were identified for each tissue. To assess sample quality and confirm proper clustering by biological variables, we performed principal component analysis (PCA) across all glands and separately for the three major salivary glands (Figure S1).

Orthology and Secreted Gene Identification

We first sought to determine for all genes in each tissue the human orthologs and which were mouse specific. For this, we used an orthology reference table provided by The Jackson Laboratory (JAX) (https://www.informatics.jax.org/downloads/reports/HMD_HumanPhenotype.rpt), which lists mouse-human orthologous genes. Using this table, we grouped genes based on the number of orthologous relationships between species. Specifically, genes were classified as one-to-one when a single mouse gene mapped to a single human gene; one-to-many when a gene in one species mapped to multiple genes in the other; and many-to-many when multiple genes in both species were linked through orthology relationships. Genes without any annotated orthologs in the other species were classified as mouse- or human-specific.

Next, we aimed to characterize the contribution of highly expressed genes to salivary secretion by identifying secreted proteins and building an expression profile for each gland. Transcript abundance was calculated in transcripts per million (TPM) using the standard formula (Wagner et al. 2012). TPM values were computed for all genes in the three major salivary glands (parotid, sublingual, and submandibular). All genes in each gland were annotated as secreted or non-secreted based on the UniProt mouse proteome (https://www.uniprot.org/proteomes/UP000000589) (UniProt Consortium 2025) and human from Human Protein Atlas (proteinatlas.org) (Uhlén et al. 2019). For the mouse proteome, specific annotations for salivary proteins were not available; therefore, we classified all genes as secreted or non-secreted based on general proteome annotations from UniProt. Combined with the orthology annotations described above, these data were used to generate Figure 1. We identified the 30 most highly expressed genes that code for secreted proteins in each salivary gland and plotted their expression levels (in TPM) to visualize gland-specific expression patterns using packcircle (https://mbedward.r-universe.dev/packcircles).

Analysis of human transcriptomes

For humans, we reanalyzed publicly available RNA-seq data from salivary glands (parotid, submandibular, and sublingual) generated by Saitou et al. (2020). Raw sequencing reads were processed as described above for mouse, using the human reference genome (hg38). Expression levels were quantified as transcripts per million (TPM), and mean TPM values were calculated for each gland. For liver and pancreas tissues, transcript data in TPMs was downloaded from GTEx and filtered for the same genes present in our dataset for salivary glands. TPMs were then recalculated for only those genes, and mean TPMs were calculated using all individuals. For orthology classification, please see the description of orthology classification above. Differential expression analysis was performed using DESeq2 (Love et al. 2014) using raw gene counts only for submandibular gland samples.

Correlation analysis of human-mouse transcriptomes

To assess the similarity of gene expression patterns between mouse and human salivary glands, we performed cross-species correlations of gene expression (mean TPM values) for each gland, focusing on one-to-one orthologs for genes with TPM>2 in each tissue. Because genes encoding secreted proteins and non-secreted proteins are subject to distinct evolutionary pressures, and secreted proteins are often expected to evolve and diverge more rapidly (Winter et al. 2004; Feyertag et al. 2016), we focused our analysis on genes encoding secreted proteins, which constitute the primary functional components of saliva. For the mouse, we used the annotated transcriptomes generated in this study for all glands. For the human data, we used the transcriptomic profiles from Saitou et al. (2020) for the major salivary glands (parotid, submandibular, and sublingual) and from GTEx for the pancreas and liver. Spearman correlation coefficients (ρ) were calculated for each comparison. Finally, we note that these analyses involve data from different cohorts and species and are therefore subject to potential batch effects; as such, these results should be interpreted with caution. Because of these technical limitations, we avoided direct comparisons of human and mouse expression data as a primary analytical approach. Nevertheless, we identified several one-to-one orthologs that exhibit dozens- to hundreds-fold differences in expression between human and mouse salivary glands (Table S3 and S4), differences that are unlikely to be explained solely by batch effects and therefore warrant further investigation.

Mouse saliva collection

Saliva samples were collected from 4-month-old male (n=5) and female (n=5) C57BL/6 mice. Mice were anesthetized with isoflurane and administered the sialogogue pilocarpine dissolved in sterile saline via intraperitoneal (i.p.) injection. Pilocarpine was delivered at a final dose of 4.53mg/kg (concentration 0.68mg/mL; injection volume 200uL per 30g body weight). Saliva collection commenced 3 minutes post-injection and continued for 10 minutes using a pipette placed in the oral cavity. Samples were immediately snap-frozen and stored at −80 degrees Celsius for subsequent protein analysis.

Human saliva collection

Saliva samples were collected from humans (male and female, n=5 each) using the passive drooling technique as previously described (Thamadilok et al. 2020).

SDS-PAGE

The 8-16% gradient pre-cast Novex Tris-glycine gel (LifeTechnologies; catalog no. XP08162BOX) was used to study the protein profiling of saliva samples based on the molecular size of the protein. Meanwhile, the samples were prepared for gel electrophoresis by mixing with a 2X sample buffer followed by boiling at 100°C for 10 min. The samples (~15 μg of protein/well) were loaded into the wells, and the gel was run on an electrophoresis system (XCell SureLock Mini-Cell , Invitrogen) at 170 V for approximately 40 minutes. Subsequently, the gel was prepared for gel staining.

Coomassie blue-PAS (Periodic acid Schiff) Staining

Protein bands were first stained using the Coomassie blue staining procedure. In brief, the procedure included three steps: a) fixation using fixing solution for 15 min twice, b) staining in Coomassie stain blue (R-250) solution for 1 hr., and c) destaining for 30 min thrice. After the destaining step, the gel was stained subsequently using the PAS staining procedure as described (Thamadilok et al. 2020). Stained gels were digitized using the GE Healthcare ImageScanner III.

Protein identification from excised gel bands by LC-MS/MS

Protein bands of interest were excised from Coomassie blue-stained SDS-PAGE gels corresponding to the ~150 kDa (male) and ~130 kDa (female) regions using a clean surgical blade and processed for protein identification by mass spectrometry. Gel slices were subjected to in-gel digestion with trypsin following standard protocols. Peptides were extracted and analyzed by liquid chromatography coupled to tandem mass spectrometry (LC-MS/MS). The resulting spectra were processed using Spectronaut (Biognosys AG, version 20.0) and searched against the UniProt Mus musculus protein database. Protein identifications were filtered using a false discovery rate (FDR) threshold of 1% at both the peptide and protein levels.

Lectin Blotting

Proteins were separated by SDS-PAGE under reducing conditions using 8-16% gradient gels and subsequently transferred to nitrocellulose membranes. For lectin detection, membranes were blocked in Tris-buffered saline containing 0.1% Tween-20 (TBST) supplemented with 2% Polyvinyl alcohol (PVA) for 1 h at room temperature. Membranes were then incubated with biotinylated lectins diluted in blocking buffer. Peanut agglutinin (PNA) was used to detect exposed Galβ1–3GalNAc (T-antigen) O-glycan structures, while Maackia amurensis lectin II (MAL-II) was used to detect α2,3-linked sialic acid residues. Following lectin incubation, membranes were washed extensively with TBST and incubated with Alexa fluor-488-conjugated neutravidin. Lectin binding was detected by enhanced chemiluminescence and imaged with a digital imaging system.

Protein digestion

Saliva specimens were processed following the E3 protocol described recently (Martin et al. 2024). In brief, around 20 μL of crude saliva was first mixed with 100 μL SDS buffer that contained 5% (w/v) sodium dodecyl sulfate (SDS), 100 mM Tris-HCl solution (pH 8.0) and final concentration of 10 mM tris(2-carboxyethyl) phosphine (TCEP) and 40 mM 2-chloroacetamide (CAA), and then boiled at 95 °C for 10 min. Next, the lysate was mixed with 4x volume of 80% acetonitrile (CAN) aqueous solution and transferred to E3filters (CDS Analytical, Oxford, PA). Following a centrifugation step, the E3filters were washed with 200 μL of 80% ACN aqueous solution three times. For protein digestion, 1 mg of trypsin (Promega, WI) in a 200 μL of 50 mM triethylammonium bicarbonate (TEAB) buffer was added to E3filters and incubated at 37 °C for 16-18 h with gentle shaking (300 rpm). To elute peptides, the E3filters were first centrifuged at 3,000 rpm and then flushed with two elution buffers, 200 ml of 0.2% formic acid in water, and 200 ml of 0.2% formic acid in 50% ACN, respectively. The eluents were dried in SpeedVac, and then subjected to desalting cleanup using C18 based StageTips (CDS Analytical, Oxford, PA). Dried peptide samples were stored under −80 °C until further analysis.

LC-MS/MS analysis

The LC-MS/MS analysis was performed using an UltiMate 3000 RSLCnano system coupled with an Orbitrap Eclipse mass spectrometer and a FAIMS Pro Interface (Thermo Scientific). The dried peptides were resuspended into 13 μl of LC buffer A (0.1% formic acid in water) and 10 μl of it was then loaded onto a trap column (PepMap100 C18, 300 μm × 2 mm, 5 μm; Thermo Scientific) followed by separation on an analytical column (PepMap100 C18, 50 cm × 75 μm i.d., 3 μm; Thermo Scientific). A linear LC gradient flowing at 250 nL/min was applied from 1% to 25% mobile phase B (0.1% formic acid in ACN) over 125 min, followed by an increase to 32% mobile phase B over 10 min. The column was washed with 95% mobile phase B for 5 min, followed by equilibration with mobile phase A for 15 min. For the ion source settings, the spray voltage was set to 1.9 kV, funnel RF level at 50%, and heated capillary temperature at 275°C. The MS data were acquired in Orbitrap at 120K resolution, followed by MS/MS acquisition in data-independent mode following a protocol described previously (Elsayyid et al. 2025). The MS scan range (m/z) was set to 380-985, maximum injection time was 246 ms, and normalized AGC target was 100%. For MS/MS analysis, the spectra were acquired with Orbitrap at resolution of 30K. The isolation mode was Quadrupole, the isolation window was 8 m/z, and the window overlap was 1 m/z. The collision energy was 30%, AGC Target was 400K, and normalized AGC target was 800%. For FAIMS analysis, a 3-CV experiment (−40|−55|−75) was applied.

Proteome quantitation and data analysis

The MS data were processed using Spectronaut software (version 20.0), a library-free directDIA+ workflow, and the UniProt protein databases (17,141 sequences; reviewed only). Briefly, the settings for Pulsar and library generation include: Trypsin/P as specific enzyme; peptide length from 7 to 52 amino acids; allowing 2 missed cleavages; toggle N-terminal M turned on; Carbamidomethyl on C as fixed modification; Oxidation on M and Acetyl at protein N-terminus as variable modifications; FDRs at PSM, peptide and protein level all set to 0.01; Quantity MS level set to MS2, and cross-run normalization turned on.

Differential protein abundance between male and female samples was assessed using the built-in statistical framework in Spectronaut (version 20.0). The protein LFQ method was set to Automatic (MaxLFQ); Quantity MS level was set to MS2, and the Quantity type was Area. The Candidates output from Spectronaut was used for quantitation analysis. Proteins were considered significantly differentially abundant between sexes if they met the thresholds of |fold change| > 1.5 and q-value < 0.05.

Transcriptome-Proteome Integration and Correlation Analysis

Proteomic data were filtered to retain only proteins with a single corresponding gene annotation to avoid ambiguity in gene-protein assignments. For each protein, mean intensity-based absolute quantification (iBAQ) values were calculated separately for males, females, and across all samples. For visualization and correlation analyses, expression values were transformed as log₁₀(1 + mean iBAQ). To enable transcriptome-proteome comparisons, we first filtered the transcriptomic dataset to genes annotated as secreted proteins (see description above). Proteins whose corresponding genes were present in this filtered transcriptomic dataset were then retained, resulting in 467 matched gene-protein pairs. Spearman correlation analyses were then performed using these matched protein-transcript pairs. It is important to note that each salivary gland contributes only partially to the overall proteome composition of mixed whole saliva, the oral fluid analyzed in this study. Therefore, strong correlations between glandular transcriptomes and the whole-saliva proteome are not necessarily expected, given the distinct spectrum of proteins secreted by each gland.

Furthermore, several highly expressed salivary gland transcripts were not detected by mass spectrometry, consistent with known technical limitations of proteomic assays for certain salivary proteins (Malaker 2024). Conversely, some proteins detected in saliva at high abundance, such as serum albumin, are not synthesized by salivary glands but instead enter whole saliva through plasma exudation in the oral cavity, as reported previously (Oppenheim and Hay 1972; Ruhl 2012).

Immunohistochemistry analysis

Submandibular glands were manually dissected and immediately fixed in 4% paraformaldehyde (PFA) at 4°C overnight. Following fixation, tissues were cryoprotected using a graded sucrose approach (10% sucrose w/v followed by 25% sucrose w/v in PBS). Tissues were then embedded in optimal cutting temperature (OCT) compound (Fisher Scientific, Cat # 4585) and cryosectioned at a thickness of 16μM. Sections were permeabilized with 0.5% Triton X-100 in PBS for 20 minutes at room temperature. Following permeabilization, sections were blocked for 1 hour at room temperature in a buffer containing 10% donkey serum (Jackson ImmunoResearch, Cat# 017-000-121) and 5% BSA (Fisher BioReagents, Cat# BP9700100) in PBS.

Sections were incubated with primary antibodies, Kallikrein 1 polyclonal antibody (1:300; Invitrogen, Cat# PA5-115481) and Nerve Growth Factor (NGF) polyclonal antibody (1:300; Sigma-Aldrich, Cat# N6655), diluted in antibody diluent overnight at 4°C. Following washing with 0.05% Tween-20 in PBS, antibodies were detected using Cy2-conjugated secondary Fab fragment antibodies (1:300; The Jackson Laboratory) incubated for 2 hours at room temperature. Nuclei were counterstained with Hoechst 33342 (1:1000; Anaspec Inc. Cat# AS-83218). Slides were mounted using Fluoromount-G (SouthernBiotech, Cat# 01100-01) and imaged using a Zeiss LSM 900 confocal microscope with a 40 X oil immersion objective.

Quantification of KLK-1 and NGF expression was performed using Fiji software. Regions of interest (ROIs) were manually delineated around salivary gland ductal structures identified by distinct morphology. The Mean Fluorescence Intensity (MFI) was calculated for each ductal ROI. Results are reported as MFI (a.u.).

Clustering of DEGs, and motif enrichment analysis

We hypothesized that co-regulated gene clusters contribute to the differences in gene expression observed between sexes. To test this, we focused on the submandibular gland for sex differences, as this tissue showed the highest numbers of DEGs. Using bedtools, we grouped DEGs into genomic clusters defined by a 100 kb window, such as the following example:

bedtools cluster –i SM_sex_DEGs.sorted.bed –d 100000 > SM_sex_DEGs.clustered_100kb.bed

To evaluate whether DEGs were more clustered than expected by chance, we performed a permutation analysis. Specifically, we randomly sampled 10,000 subsets of 1615 genes (matching the observed DEG count) from the total set of expressed genes and applied the same clustering procedure to each subset. From these permutations, we built an expected distribution of the number of genes per cluster size and compared it to the observed data (Table S9).

Next, we focused on the kallikrein (Klk) gene cluster in the mouse submandibular gland. Promoter regions for genes within the cluster were extracted and analyzed for motif enrichment using HOMER (Hypergeometric Optimization of Motif EnRichment) (Heinz et al. 2010). HOMER performs de novo motif discovery by identifying sequence motifs that are significantly enriched in a target set of sequences relative to a background set using a hypergeometric framework, and compares discovered motifs to known transcription factor binding motifs in curated databases (including HOMER and JASPAR libraries). For each gene in the cluster, sequences spanning −400 bp upstream to +100 bp downstream of the transcription start site (TSS) were analyzed. De novo motif discovery was performed using motif lengths of 8, 10, and 12 bp (Table S10), according to the following code

findMotifs.pl “$gene_list_in cluster” mouse “out_folder/” \

     –start -400 -end 100 \

     –len 8,10,12 \

     –p 4

To further confirm the motifs identified in the enrichment analysis and evaluate whether they are present across other Klk genes in mouse and in their human orthologs, we searched for these motifs across all Klk gene sequences in both species. Gene sequences for all annotated Klk genes were downloaded from the GRCm39 mouse genome assembly and the corresponding human genome assembly in FASTA format. Extracted sequences included the full gene regions encompassing coding sequences as well as 5′ and 3′ untranslated regions (UTRs). The short motif sequence identified by HOMER (CTGATCCTGTTC) was queried against these gene sequences using BLASTN with the blastn-short task, which is optimized for short query sequences, according to the following example:

blastn task blastn-short -query $motif.fa -subject $genes.fa -outfmt 6 -out out.blastn.out

This approach allowed us to identify exact or near-exact matches of the motifs within Klk gene sequences (Table S11).

Evolution of the Klk genes, chromatin conformation and Topological Associated Domains Visualization

To investigate the evolution of the Kallikrein genes and gene expansions-contractions, we used CAFE 5 (Mendes et al. 2021). To investigate whether the sex-biased expression of Kallikrein (Klk) genes could be explained by differences in genome architecture and transcriptional regulation, we analyzed publicly available epigenomic datasets from the ENCODE project. Specifically, we visualized the Klk locus (chr7:43650000-44300000, mm10) using Hi-C contact maps at 25kb resolution generated from B-cell lymphoma cell line (https://www.encodeproject.org/files/ENCFF076GYR/) using HiCExplorer (Wolff et al. 2020). We further identified topologically associating domains (TADs) from the Hi-C data and overlaid them onto the figure. The code used for TAD calling and visualization is provided below.

hicFindTADs \

 --matrix ENCFF651SGY.mcool::/resolutions/25000 \

 --chromosomes chr7 \

 --minDepth 75000 \

 --maxDepth 250000 \

 --step 50000 \

 --correctForMultipleTesting fdr \

 --thresholdComparisons 0.05 \

 --outPrefix TADs_heart_chr7_25kb_Local

WIN2=“chr7:43650000-44300000”

hicPlotTADs --tracks 25kbTADs_selec.ini --region $WIN2 \ --outFileName 25kbTADs_selec.svg --dpi 300

To explore potential regulatory features underlying the observed expression patterns, we overlaid ATAC-seq data (https://www.encodeproject.org/experiments/ENCSR451NAE/) representing chromatin accessibility, chromatin immunoprecipitation sequencing (ChIP-seq) data for the architectural protein Ctcf (https://www.encodeproject.org/experiments/ENCSR000ERM/) and for histone modification H3K27me3 (https://www.encodeproject.org/experiments/ENCSR000CGK/), and for Rad21 (https://www.encodeproject.org/experiments/ENCSR000ERK/) from the same cell line.

To investigate sex-specific differences in chromatin organization at the kallikrein (Klk) locus, we analyzed publicly available Hi-C datasets generated from granulosa (female) and Sertoli (male) cells (Lindeman et al., 2021). These datasets were obtained from the 4DN Data Portal (https://data.4dnucleome.org/) under accession 4DNEXDK2R8M5 and 4DNEX99DGLEF. Multi-resolution cooler (.mcool) files were downloaded for each dataset (Sertoli: 4DNFIB7RFFBB; granulosa: 4DNFIZ59STGB). These cell types were selected because they represent sex-specific gonadal cell lineages, providing a model to examine potential sex-dependent chromatin architecture. Hi-C contact maps were extracted at 25 kb resolution, and topologically associating domains (TADs) were independently identified for each dataset using HiCExplorer as described above. The resulting TAD boundaries were then compared between male and female datasets to assess differences in higher-order genome organization at the Klk locus.

DATA AVAILABILITY

RNA-seq data generated in this study will be deposited in the NCBI Gene Expression Omnibus (GEO) prior to publication, and accession numbers will be provided upon acceptance. Human salivary gland RNA-seq data were obtained from Saitou et al. 2020 and are publicly available under the corresponding accession numbers. Proteomic data generated in this study will be deposited in an appropriate public repository (e.g., ProteomeXchange/PRIDE) prior to publication. All scripts used for data processing and analysis will be made publicly available upon acceptance.