Unveiling critical signaling pathways in the murine salivary gland and the role of midkine

Summary

Salivary gland (SG) development is a complex process involving coordinated signaling between epithelial and stromal cell populations. While some growth drivers and branching mechanisms are known, many intercellular communication pathways that underpin embryonic SG morphogenesis remain unidentified. We leveraged single cell RNA-sequencing datasets, from murine submandibular salivary glands at three embryonic stages to identify critical signaling networks. Using CellChat, we mapped global ligand-receptor interactions among different cell populations revealing an evolving signaling landscape during gland maturation. Our analysis highlighted both well-established and understudied pathways, including midkine (MDK) signaling. Functional experiments using embryonic SG explants demonstrated that MDK regulates branching morphogenesis via Rho-associated coiled coil containing protein kinase 1 (Rock1) signaling. Additionally, the key lineage driving transcription factor p63 was shown to act as primary mediator of the MDK/Rock1 axis. These findings uncover a comprehensive signaling code that guides SG development and offers new targets for follow-up studies into SG regeneration.

Graphical abstract

Highlights

• •Salivary gland (SG) development entails intricate cell-cell communication patterns
• •Dynamic ligand-receptor interactions among diverse cell types drive morphogenesis
• •Midkine (MDK) is a functionally important pathway driving SG branching development
• •The transcription factor p63 drives branching through the MDK/Rock1 signaling axis

Cell biology; developmental biology

Introduction

Salivary gland (SG) morphogenesis requires an exquisitely fine-tuned interplay of signaling pathways that operate between a myriad of cell populations in the stroma and the epithelia. The first phase of murine SG development transpires during embryogenesis when the oral epithelium invaginates into the surrounding mesenchyme at embryonic day 11 (E11), the prebud stage. Soon after, instructive mesenchymal-epithelial interactions, anchored primarily by reciprocal fibroblast growth factor (FGF) signaling propels the gland into the initial bud stage (∼E12). The mesenchyme abutting the primitive gland at this stage is enriched with plexus of endothelial cells and parasympathetic ganglion neurons that coalesce around the primary duct and nourish the growing tissue.^1,2,3,4,5 At E13.5, the bud undergoes repeated rounds of clefting marking the beginning of the pseudoglandular stage. Extracellular matrix components, such as collagen, fibronectin, and laminin play important functional roles during this stage of cleft formation and branching.^6,7,8 In addition, many growth factors and pathways regulate branching morphogenesis, including epidermal growth factor (EGF), WNT, and NOTCH signaling.⁹ Another vital signaling pathway active during this process is heparan sulfate proteoglycan (HSPG), which impacts cell survival, motility, and tissue structure by modulating ligand diffusion and local concentration of heparin-binding secreted growth factors, such as FGF in the vicinity of epithelial cells.¹⁰ The early stages of SG development are thus marked by dynamic signaling crosstalk between the stromal and epithelial cells that in turn dictate SG growth, patterning, and differentiation.

The subsequent phase of SG maturation involves hollowing out the core of the tubular network at E15.5 (canalicular stage) through apoptosis and the coordination of WNT, the ectodysplasin/ectodysplasin receptor (EDA/EDAR), HedgeHog (HH), and vasoactive intestinal peptide (VIP) signaling pathways.^{8,11,12,13,14,15} This process of lumenization allows the flow of saliva from the acini to the oral cavity. By E16.5, the outer layer of epithelial cells starts proliferating and there is increased cytodifferentiation, a process which relies on crosstalk between various cell types. For instance, acinar cell differentiation is dependent upon signals sent from neuronal cells to the end bud.¹⁶ Other supportive cell types such as endothelial cells also secrete growth factors that aid in SG patterning as the vasculature network co-develops with the epithelium.¹⁷ Finally, by the terminal bud stage at E18, the lumina connecting the ducts to the acini are completed and the SG at this stage is considered to be mostly functional. However, the gland continues to develop postnatally until a well-organized network of acini and ducts that are surrounded by various cell types and structures including blood vessels, nerves, and immune cells, are formed. This complex and dynamic developmental process is critically dependent on cell-cell communication and epithelial-mesenchymal interactions as shown by many ex vivo and in vivo studies.^4,5,18,19,20 However, despite progress in this field, in-depth knowledge about the intercellular communication and signaling networks between the distinct cell types at different stages remain relatively poorly understood. In particular, it is likely that there are, as yet, unknown signaling pathways and cell communication codes that remain to be discovered.

Recent advances in genomics, specifically single cell RNA-sequencing (scRNA-seq), allows for in-depth investigations of cellular communication events and signaling patterns during development. Salivary gland scRNA-seq datasets can be leveraged to not only define cell population characteristics at molecular levels but can also provide information on ligand-receptor interactions and cell-cell signaling networks.²¹ Toward this end, we have integrated and analyzed scRNA-seq datasets of the developing murine submandibular gland (SMG)²² across distinct stages to explore cell-cell communication events important for generating a mature and functioning gland. While many studies in the past have focused on mesenchymal to epithelial crosstalk, the importance of other cell types influencing epithelial cell development cannot be understated.^17,19,23 In order to answer these unresolved issues, we utilized CellChat, a gene expression-based modeling algorithm that can predict global cellular crosstalk mediated by ligand-receptor interactions.^24,25 Interestingly, while our analysis identified well-established pathways that contribute to the various stages of embryonic development, such as COLLAGEN, LAMININ, and NOTCH, additional pathways were also unearthed that were enriched at each developmental stage. Furthermore, our findings indicate that the midkine (MDK) signaling pathway, previously uncharacterized in the context of SG embryogenesis, may play a pivotal role in mediating intercellular communication. We validated the presence and distribution of MDK in specific cell types by immunofluorescence and examined the functional role of MDK on developing SMG morphogenesis in explant models using loss-of-function and gain-of-function approaches. Our mechanistic studies showed that MDK modulates the downstream Rock1 signaling in directing branching in the SG, whereas the lineage driving transcription factor p63 acts as an upstream regulator of the MDK/Rock1 signaling pathway to coordinate branching morphogenesis. These new findings combined with previous knowledge of developmental pathways and cell-cell communication regulating SG organogenesis can offer the requisite knowledge base to develop therapeutic strategies for SG regeneration or repair in humans.²⁶

Results

Identifying broad signaling patterns directing various stages of embryonic submandibular gland morphogenesis

Following up on prior studies that identified a role for Sfrp1 in directing branching morphogenesis at a single developmental stage, E14,²⁷ we broadened our analysis to examine potential signaling patterns that occur throughout the entire process of SMG morphogenesis. Toward this end, we leveraged a published scRNA-seq atlas of murine SMGs datasets representing key developmental time points, including E12, E14, and E16.²² We first re-analyzed the datasets using CellRangerv7.1.0, which resulted in a total of 7,070 (E12), 7,358 (E14), and 6,229 (E16) cells that were then integrated and used for downstream analysis. The re-processing of the datasets not only confirmed the presence of broad cell populations including epithelial, stromal, immune, and neuronal cell types (Figure S1) in agreement with published studies,^22,28 but additionally identified basal cells and fibroblasts at E12 and E14, respectively, which were not previously reported. Having identified distinct cell populations, we next sought to evaluate the prominent signaling pathways that are enriched during the various stages of SMG development by performing a comparative analysis using CellChat v2.²⁴ Our investigation revealed several pathways to be uniquely enriched at E12, including Adhesion G protein-coupled receptor B (ADGRB) and V-domain immunoglobulin suppressor of T cell activation (VISTA) (Figure 1). Notably, we also identified a number of additional pathways, that while not unique to E12, were nonetheless enriched during this early developmental stage (Figure 1). Of these, prominent were WNT and neuronal growth regulator (NEGR) signaling, both of which have been shown to be important for SG morphogenesis and homeostasis, respectively^14,27,29,30 (Figure 1). At the later E14 stage, we observed selective enrichment of GALANIN, tubby-like proteins (TULP), and semaphorin 5 (SEMA5) signaling pathways (Figure 1). Finally, at E16 we found enrichment of growth and differentiation factor (GDF) signaling, a known ligand belonging to the transforming growth factor-β (TGF-β) superfamily (Figure 1). Additionally, our analysis identified CEACAM (carcinoembryonic antigen cell adhesion molecule) signaling to be enriched during this developmental stage (Figure 1). Overall, these results highlight the diversity of signaling pathways associated with the various stages of SG development, some of which might play a conserved role in other glandular and secretory organs.

Figure 1.

Enriched signaling pathways predicted to play a role during the various stages of SMG development

Bar chart showing the signaling pathways based on the relative information flow between datasets. Blue, yellow, and green bars represent pathways enriched during E12, E14, and E16, respectively.

Cell-cell communication networks driving salivary gland development

Cell to cell communications are essential and vital processes required by all living organisms to ensure proper growth, development, differentiation, tissue and organ maintenance, and regeneration. To probe potential incoming and outgoing intercellular communication networks and identify inferred ligand-receptor pairs between interacting cell types during development, we constructed intercellular communication networks using CellChat v2, which utilizes an expansive database of signaling networks and ligand-receptor pairs.²⁴ We first compared the number of inferred ligand-receptor pair communications/interactions among the different cell populations at E12, E14, and E16. We found a total of 1,906 predicted interactions at E12 (Figure S2A). However, by E14, there was a significant increase in the number of predicted interactions to 3,858 and by E16 the number of interactions had grown to 4,719 (Figures S2B and S2C). This steady increase in the overall number of cell-cell communications over the course of morphogenesis fits well with the complexity of glandular morphogenesis during embryonic development. Closer examination revealed that at E12, the mesenchyme, Krt19⁺ duct, and end bud cell populations were the major outgoing cellular sources (senders) with the highest number of putative interactions. Conversely, the mesenchyme and Krt19⁺ duct were ranked as the highest putative incoming cell targets (receivers) at this developmental stage (Figure S2A). Interestingly, similar outgoing communication patterns were observed in the mesenchyme and epithelial cells including the Krt19⁺ duct, end bud, and myoepithelial cells (MECs), while putative incoming cell-cell communication patterns were seen in the mesenchyme, and epithelial cells comprising the K19⁺ duct and basal cells, over the remaining developmental stages (Figure S2). Not surprisingly, the weight and strengths of cell-cell interactions paralleled these findings (Figure S2). The overall increase in putative cell-cell communications together with the predicted involvement of the mesenchyme and different epithelial cell types, highlight the various cellular players and dynamic changes in signaling occurring during development, particularly during the later stages.

Having determined the intricate cell-cell communication networks between the various cell types, we next explored how diverse cell populations and signaling pathways synchronize to coordinate the different stages of SG development. Indeed, among the diverse cell populations, CellChat predicted the mesenchymal cells to be the major senders of outgoing signaling pathways during the various stages of development. This was followed by the epithelial cells (Figure 2A). Interestingly, by E16, in addition to the mesenchymal cells, the MECs were predicted to be the highest epithelial cell sub-type among senders of outgoing signals. Conversely, examination of the incoming signaling pathways revealed that during the early stages of morphogenesis, the endothelial cells, followed by the nerves, end bud, Krt19⁺ duct, and basal cells were the major receivers of incoming signaling (Figure 2B). At E16, while the immune cells including macrophages and monocytes were predicted to be top receivers of incoming signaling pathways, these were again closely followed by the basal, end bud, and MEC cells, suggesting the importance of epithelial cells during this critical stage of SG development (Figure 2B).

Figure 2.

Network and cell-cell communication signaling patterns during SMG development

Heatmap visualization showing the summary of the signaling pathways that contribute to outgoing (A) and incoming (B) communications based on scRNA-seq.²² The colored bars represent the relative signaling strength of a signaling pathway across cell types as indicated.

Predictably, CellChat revealed several overrepresented incoming and outgoing signaling pathways that were found to be operational during SG development. As expected, we identified several signaling pathways that have been previously shown to play important roles in SG morphogenesis, including FGF, NOTCH, LAMININ, and COLLAGEN.^7,31,32 Interestingly, for a number of these signaling pathways including FGF, LAMININ, and COLLAGEN, the cell types that were predicted to be major receivers of these signals were the epithelial cells (Figure 2B). With this in mind, we further investigated the intercellular communication networks by performing a cell-cell interaction analysis to examine the outgoing signals from non-epithelial supporting cells to the epithelial cells. Among the outgoing signaling emanating from the mesenchyme, our results predicted ligand-receptor pairs from several signaling pathways including MDK, pleiotrophin (PTN), fibronectin 1 (FN1), insulin-like growth factor (IGF), and FGF (Figure S3). Similarly, the top outgoing signaling pathways from the endothelial cells included HSPG, IGF, adhesion G protein-coupled receptors subfamily C (ADGRC), myelin protein zero (MPZ), and erythropoietin-producing hepatoma A subfamily (EPHA). Additionally, top outgoing signaling pathways from immune cells included tumor necrosis factor (TNF), TGF-β, IGF, GALECTIN, and Kallikrein-related peptidase (KLK). Finally, examination of the top outgoing signaling pathways from nerve and glial cells to the epithelial cells revealed MDK, PTN, neural cell adhesion molecule (NCAM), to be among the top signaling pathways (Figure S3). Collectively, these results highlight the complexity of the diverse signaling mechanisms that operate between the various cell populations in the embryonic salivary glands.

Among the discovered signaling pathways, MDK emerged as the top incoming and outgoing signaling pathway across the various cell populations at each of the three representative developmental stages (Figure 2). MDK is a heparin-binding cytokine that belongs to the midkine/pleiotrophin (MDK/PTN) family and has been shown to play a key role in a wide range of processes, including development, tumorigenesis, inflammation, tissue repair, and regeneration.^33,34,35,36 Given that MDK was identified as the predominant signaling pathway across multiple cell populations, particularly within epithelial lineages, and that its function in SG morphogenesis has not been previously characterized, we directed our subsequent investigations toward elucidating the mechanistic role of this pathway.

Midkine signaling in the embryonic salivary gland

Network centrality analysis of the inferred MDK signaling network in the E16 SMG, performed using CellChat, identified the mesenchyme and MECs as the top sources of MDK signaling, whereas the several other distinct cell types were predicted to act as receivers. These findings further underscored the wide involvement of this pathway during embryonic SMG morphogenesis (Figure 3A). Next, to examine cell type-specific expression of MDK, we probed the E16 SMG dataset, and as shown by the violin plot, MDK was found to be expressed at varying levels across all principal epithelial cell types as well as in fibroblast, endothelial, and nerve cells (Figure 3B). To gain additional insights into how MDK signaling might operate between the myriad cell populations of the embryonic SMG, we next performed ligand-receptor analyses. In order to identify predicted MDK specific enriched receptors in the E16 SG epithelium, we utilized MDK as the source ligand and focused on predicted outgoing signals from both the mesenchyme and MECs, as these cell types were predicted sources of MDK signaling (Figure 3C). Our targeted analysis identified 10 potential MDK receptor genes enriched in the Krt19⁺ duct, MECs, end bud, and basal epithelium which included the Integrin (ITG) and Syndecan (Sdc) family of genes, as well as Nucleolin (Ncl) (Figure 3C). Closer examination revealed that while many of the receptors showed ubiquitous expression across all the cell types, a prominent MDK receptor Sdc4, showed enriched expression in the epithelial cells of the SG as well as in a subset of immune cells (Figure 3B). To examine the spatial distribution pattern of MDK at the protein level, we next performed a series of immunofluorescent co-stains with the end bud/acinar cell marker Nkcc1 or the ductal cell marker K19, with MDK. We observed prominent epithelial expression of the MDK protein at the E13, E14, and E16 stages (Figure 3D), in agreement to the corresponding mRNA patterns in end bud and ductal cell populations of E16 SMG as depicted in the violin plot shown in Figure 3B. Interestingly, in contrast, MDK protein expression was comparatively sparce in non-epithelial cell populations at E13 or E14 but readily detectable in the mesenchymal cells of the E16 glands (Figure 3D). We posit that this intriguing dichotomy between the abundance of the MDK mRNA transcript and the corresponding protein in cells such as fibroblasts at the earlier developmental time points might be the result of the experimental conditions of the immunofluorescence or possibly due to as yet unknown, post-transcriptional or post-translational regulatory mechanisms. Taken together, these data suggest that a dynamic and complex pattern of MDK signaling operates in the developing embryonic SMG.

Figure 3.

Midkine associated signaling in the murine SMG

(A) Heatmap visualization of the MDK signaling network across various cell populations in the mouse SMG.

(B) Violin plot showing expression levels of ligands and receptors associated with MDK signaling in the different cell populations of the mouse SMG at E16.

(C) Chord diagram depicting outgoing communication probability of ligand-receptor pairs contributing to MDK signaling from mesenchyme (upper panel) and MECs (lower panel), based on results from panel A.

(D) Representative immunofluorescence images of mouse SMGs co-stained with Mdk and the acinar cell marker Nkcc1 or the ductal marker K19 at E13, E14, and E16. Scale bars, 20 μm.

Midkine signaling directs salivary gland branching morphogenesis

To evaluate the mechanisms through which MDK governs SG morphogenesis, we utilized loss-of-function and gain-of-function approaches employing an established ex vivo organ explant cell culture model system. Toward this end, E13.5 SG rudiments from wild type C57BL/6J mouse embryos were harvested and cultured ex vivo in the presence of a well-established MDK inhibitor³⁷ (iMDK) or treated with recombinant MDK (rMDK) for 72 h (E13.5 + 72 h). To determine the optimal concentrations of iMDK and rMDK for treating cultured SG rudiments, we first conducted time course and dilution analyses to assess toxicity (Figures S4 and S5, respectively). We found that E13.5 SG explants treated with 1 μM of iMDK for 72 h resulted in a pronounced reduction in overall branching as compared to the dimethyl sulfoxide (DMSO) treated control glands (Figure 4A). The branching defects were prominently visible upon whole-mount immunofluorescence staining of control and iMDK treated glands. Glands co-stained with the basal stem/progenitor cell marker Keratin 5 (K5) and the nerve cell marker Tubb3, revealed dramatic reductions to branching morphogenesis and the associated intertwined nerves within the SG epithelia (Figure 4A, lower panel). The defects in branching were further quantified by Spooner’s ratio, which revealed a significant reduction in the number of end buds in the iMDK treated glands compared to controls (Figure 4B). Interestingly, the end buds of the iMDK treated glands appeared enlarged with a striking reduction in the number of clefts revealing a broader disorder in tissue organization (Figure 4A). Conversely, gain-of-function experiments in which E13.5 explants were treated with 100 ng/ml of rMDK for 72 h, exhibited increased end buds compared to control glands based on quantification of Spooner’s ratio (Figure S5).

Figure 4.

Branching morphogenesis of embryonic E13.5 SMGs treated with midkine inhibitors is reduced in ex vivo culture

(A) Upper panel are light micrographs of wild type C57BL/6J SMGs harvested at E13.5 and cultured for 72 h in the presence of midkine inhibitor (iMDK) or DMSO control. Lower panel shows whole-mount immunofluorescence staining of DMSO control and iMDK treated glands shows expression of Keratin 5 (K5) and neuronal tubulin expressing nerves (Tubb3). Scale bars, 100 μm, (n = 3).

(B) Spooner’s ratio quantification of the number of end buds (expressed as a ratio of the number of end buds at 72 h/number at 1 h) of DMSO control (n = 10) and iMDK treated glands (n = 10).

(C) Immunofluorescence staining of paraffin embedded DMSO control and iMDK treated glands reveal decreased expression of ΔNp63 and reduced expression of K14 expressing basal stem/progenitor cells. Alterations to the acinar cell differentiation program were also observed in the iMDK treated glands as evident by reduced expression of Mist1, Nkcc1, and Aqp5. Ductal cell differentiation was also altered as depicted by decreased K7 expression in the iMDK treated glands compared to DMSO controls. Scale bars, 20 μm. Lower panel shows quantification of the p63⁺K14⁺/p63⁺ basal cells, p63⁺Sma⁺/p63⁺ MECs, and Nkcc1⁺Mist1⁺/Nkcc1⁺ acinar cell population as mean ± SD (n = 3).

(D) Quantitative RT-PCR validation of select genes in the DMSO control and iMDK treated glands. Values were normalized to the housekeeping gene Hprt. Data are represented as means ± SD (n = 3). ∗∗p < 0.01, ∗∗∗p < 0.001.

Given the robust branching defects observed in the iMDK treated explants, we performed follow-up studies on the 1 μM iMDK treated explants. Specifically, to better evaluate the cellular alterations associated with the branching defects due to MDK inhibition, we performed immunostaining analysis. DMSO control and iMDK treated explants were co-stained with the basal stem/progenitor cell markers, p63, Keratin 14 (K14), and the MEC marker, smooth muscle alpha (Sma). We found p63 expressing cells to be specifically located around the leading edges of the epithelial tissue, further accentuating the branching defects observed in the iMDK treated glands (Figure 4C). Conversely, both K14 and Sma expression were significantly reduced in iMDK treated glands compared to DMSO controls (Figure 4C). Follow-up quantification of the p63⁺/K14⁺ basal and p63⁺/Sma⁺ MEC cell markers corroborated our findings (Figure 4C lower panel). Similarly, staining for the acinar cell markers Mist1, Nkcc1, and Aqp5 and the ductal marker K7, revealed reduced expression which was accompanied by disorganized acinar and ductal structures in the iMDK glands (Figure 4C, lower panel). Quantification of the Nkcc1⁺/Mist⁺ acinar cell markers was performed to validate our results. Finally, further evaluation of the expression levels of a subset of genes by quantitative real-time PCR (RT-qPCR), supported our immunostaining findings (Figure 4D). Collectively, these complementary loss and gain-of function results suggest that MDK signaling is likely to play an important role in SG branching morphogenesis and that loss of MDK results in impaired epithelial cell differentiation.

Midkine modulates Rock1 expression in directing branching morphogenesis

Rock1 belongs to the family of serine-threonine kinases which has been shown to play important functions in the development of several branched organs including the lung and kidney.^38,39,40 Importantly, previous studies on SG development have revealed a prominent role for Rock1, as this protein is essential for cleft formation—a critical step in the branching morphogenesis program.^41,42,43 To investigate the potential mechanism through which MDK might exerts its downstream effects associated with the observed branching defects, we sought to determine the status of Rock1 in the iMDK treated glands. Whole-mount immunofluorescence staining of DMSO control and iMDK treated glands revealed reduced expression levels of Rock1 in the iMDK treated glands compared to DMSO controls (Figure 5A). In agreement, western blot analysis of total protein extracts from DMSO control and iMDK treated glands confirmed reduced Rock1 protein expression levels (Figure 5B). Additionally, we observed reduced protein expression levels of p63 and Sma (Figure 5B), in accordance with our immunofluorescence staining (Figure 4). The reduced protein expression levels of ΔNp63 in the iMDK treated glands prompted us to evaluate the expression status of Sfrp1, a WNT ligand agonist, as our previous studies have shown an important role for the ΔNp63/Sfrp1 axis in directing branching morphogenesis.²⁷ Western blot analysis showed reduced expression levels of Sfrp1, concomitant to reduced ΔNp63 levels in the iMDK treated glands. Overall, these data suggested a role for MDK in directing SG branching morphogenesis by potentially modulating Rock1 signaling and hinted at a potential cross-talk and/or regulatory feedback loop between WNT and MDK signaling that might be anchored by p63.

Figure 5.

Midkine modulates Rock1 signaling in directing branching morphogenesis

(A) Whole-mount immunofluorescence staining of wild type C57/BL/6J SMGs harvested at E13.5 and cultured for 72 h in the presence of iMDK or DMSO treated controls show a dramatic loss of Rock1 expression in the iMDK treated glands compared to controls. Scale bars, 100 μm (n = 3).

(B) Western blot showing the effects of midkine inhibition on Rock1, ΔNp63, Sma, and Sfrp1 protein expression levels compared to control treated glands as described in A above. Protein expression levels were normalized to β-actin. Data are represented as means ± SD (n = 3). ∗p < 0.05, ∗∗p < 0.01. Scale bars, 100 μm.

The transcription factor ΔNp63 directs embryonic salivary gland branching morphogenesis through regulation of the midkine/Rock1 signaling axis

In order to identify potential upstream signaling mediators of MDK gene expression, we performed Pearson correlation analysis of the integrated embryonic datasets to identify regulatory genes whose expression correlated with those of MDK within the epithelial cell populations. Focusing our analysis on transcription factors (TFs) that showed strong positive correlation with MDK, we identified a number of genes which encode for TFs that have been shown to be important in SG development in both human and mouse including Ybx1, Foxp1, Eno, Nfib, and Sox11 (Figure 6A).^22,45,46 Importantly, we identified the Trp63 gene encoding the TF p63, which similarly showed a positive correlation with Mdk expression (Figure 6A). p63, specifically the ΔNp63 isoform, is a lineage-specific master TF which has been shown by our group and others, to be critical for SG development and branching morphogenesis.^{27,47,48,49,50} Based on our findings from the correlative gene expression analysis, we next sought to determine if midkine was a potential target gene of ΔNp63. By probing our previous p63 chromatin immunoprecipitation-sequencing (ChIP-seq) dataset,⁴⁴ we identified a ΔNp63-bound region downstream of the Mdk locus (Figure 6B). The ΔNp63-bound regulatory region was deemed a likely active enhancer as demonstrated by enriched H3K27Ac and H3K4Me1 histone peaks at this site (Figure 6B). Moreover, analysis of the E16 scRNA-seq datasets revealed overlapping expression of Trp63 and Mdk in similar epithelial cell populations (Figure S6A). Taken together, these findings strongly suggested that ΔNp63 might be a direct transcriptional regulator of MDK. To test this hypothesis, we next performed follow-up ex vivo explant studies using ΔNp63 conditional knockout mouse (ΔNp63cKO) models, as previously described.^27,44 For these studies, E13.5 salivary glands from ΔNp63^fl/fl (control) and UBC^CreERT2;ΔNp63^fl/fl (ΔNp63cKO) embryos were harvested and cultured ex vivo. In order to knock down ΔNp63 expression, media was supplemented with activated tamoxifen (TAM) and ΔNp63cKO explants and control glands were grown for 72 h.^27,44,50 Loss of ΔNp63 resulted in a prominent reduction in overall levels of branching in agreement with previous findings (Figure 6C). Immunofluorescence analysis of the freshly established E13.5 explant glands revealed robust baseline expression of ΔNp63 in the epithelium and MDK expressed in both the epithelium and mesenchyme (Figure S6B). As expected, post 72 h of TAM treatment, we observed a marked reduction in ΔNp63 expression. Interestingly, this was accompanied by a parallel loss of MDK expression in both the mesenchyme and epithelium of ΔNp63cKO explants compared with controls (Figures 6C and S6B). Given the epithelial-restricted expression of ΔNp63, we surmise that non cell autonomous mechanisms are likely involved in the mesenchymal expression of MDK regulation. Additionally, we observed a significant reduction in Rock1 expression in the ΔNp63cKO explants compared to control (Figure 6C). Protein quantifications by fluorescence intensity and mRNA analysis by RT-qPCR, confirmed these findings (Figures 6D and 6E). To further investigate and gain a better appreciation of the overall architecture of the whole-mounted SMG explant tissue, we generated a 3D video rendering of confocal microscopy images. This 3D visualization reaffirmed the broad changes in the SMG tissue architecture and epithelial cell morphology following the loss of p63, along with the associated downregulation of Rock1 and MDK expression (Videos S1 and S2). Taken together, our combined genomic, genetic, and molecular studies have identified MDK/Rock1 signaling axis as an important molecular link in the genetic circuitry through which p63 directs cell-to-cell communication during SG branching morphogenesis (Figure 7).

Figure 6.

Regulation of Midkine expression by the transcription factor ΔNp63

(A) Correlation analysis highlighting the top transcription factors correlated with Midkine expression. Color scale represents correlation scores.

(B) Visualization of the ΔNp63 binding site identified upstream of the midkine genomic locus (integrated genomics viewer). Top two lines display the p63 ChIP binding site and signal enrichment in primary salivary gland epithelial cells.⁴⁴ Overlays of histone ChIP-seq marking enhancers (H3K27ac and H3Kme1) are shown.

(C) Light micrographs of control and ΔNp63cKO SMGs harvested at E13.5 and cultured for 72 h in the presence of activated TAM (left panel), scale bars, 100 μM. Whole-mount immunofluorescence staining of control and ΔNp63cKO SMGs harvested at E13.5 and cultured for 72 h in the presence of activated TAM shows a dramatic loss of midkine and Rock1 expression in the ΔNp63cKO SMGs compared to TAM treated controls (right panel). The persistent expression of ΔNp63 in the ΔNp63cKO salivary glands (purple cells) is likely due to inefficient/incomplete deletion upon TAM administration. The epithelium is outlined in white. Scale bars, 100 μm, (n = 3).

(D) Relative fluorescence intensity of whole-mount immunofluorescence staining of control and ΔNp63cKO SMGs harvested at E13.5 and cultured for 72 h in the presence of activated TAM using ImageJ. Data are represented as means ± SD (n = 3).

(E) Quantitative RT-PCR validation of ΔNp63, Mdk, and Rock1 in control and ΔNp63cKO SMGs harvested at E13.5 and cultured for 72 h in the presence of activated TAM. Values were normalized to the housekeeping gene Hprt. Data are represented as means ± SD (n = 3). ∗∗p < 0.01, ∗∗∗p < 0.001.

Figure 7.

Diagram illustrating ΔNp63-MDK-Rock1 axis

Under normal developmental conditions, the transcription factor ΔNp63 could potentially regulate the expression of midkine which influences Rock1 signaling thereby directing branching morphogenesis. Ablation of ΔNp63 results in the loss of midkine expression resulting in reduced Rock1 signaling and subsequently leading to a defect in branching morphogenesis.

Discussion

The dynamic processes that drive the development and maturation of embryonic SMGs have been well studied in the context of signaling pathways such as WNT, FGF, and NOTCH; however, the role of MDK signaling in these processes has received comparatively less attention. This is in part due to the complexity of the molecular events that are orchestrated in a dynamic fashion by the various cell populations working in both cell autonomous and non-autonomous manners. The glandular structure of the SMG forms by the primary developmental process of branching morphogenesis, which involves coordinated cell proliferation, clefting, differentiation, migration, and apoptosis, among other processes. Not unexpectedly, it is likely that these embryonic morphogenic processes are associated with both well-established and novel, yet to be discovered signaling pathways. Our comprehensive analysis of the stage-specific scRNA-seq datasets shed light on these cellular processes, particularly in the context of the signaling crosstalk that underlie the rapidly changing structural and morphological changes in the growing salivary glands of the mouse embryo. Specifically, our CellChat based analysis on the scRNA-seq datasets highlights the diverse array of intercellular incoming and outgoing communication patterns across distinct stages of embryonic development.

Many of these signaling pathways are understudied in the SG but have been shown to be important in other branching organs. Of note, GALANIN signaling which was enriched in the SMG at E14 has been shown to be important for branching morphogenesis during both puberty and pregnancy and for proper secretory function in the mouse lactating mammary gland.^51,52,53 Given the functional and structural similarities between the mammary gland and SG, it’s tempting to speculate that GALANIN may play similar secretory roles in the SG. Additionally, GDF has not directly been implicated in the context of SG development, yet its distant family member TGF-β, has been shown to be important in SG development including branching morphogenesis.^54,55 Finally, our CellChat analysis also identified CEACAM signaling to be enriched at E16. While CEACAM signaling has been associated with various biological processes including angiogenesis, cellular adhesion, and immune function,^56,57 its role in SG development remains somewhat unexplored. However, a recent report by Rheinheimer et al. has identified an Etv1 positive cell population, likely representing acinar cell precursors, to express receptors for CEACEM ligands in adult mouse parotid gland.⁵⁸ Furthermore, CEACAM1 has been shown to be functionally important in mammary gland morphogenesis and lumen formation.^59,60,61 These observations, we posit, suggest that CEACEM signaling is a strong candidate for further exploration in SG development.

As illustrated in this report, indeed we find that a panoply of signaling pathways operate reciprocally between the epithelial, mesenchymal, neuronal, and endothelial cells of the embryonic SG—befitting the dynamic cellular and tissue growth and structural reorganization that takes place during the E12-16 developmental window. One surprising finding is that these signaling pathways ply not only in the epithelial-mesenchymal corridor as has been reported in published literature but importantly, many of the support cells in the glandular stroma, such as neuronal, endothelial, and immune cells are equally active players in these processes. This is further substantiated by our CellChat results that show as the embryonic gland matures, outgoing signaling patterns of various kinds from endothelial (such as IGF and HSPG) and neural (such as NCAM and PTPR) cells are at play. Immune cells also display similar cell to cell communication patterns, with surprise candidates such as GALECTIN serving as a major contributor from macrophages. Another interesting observation from our studies is that some signaling pathways are broadly operational between different cell populations suggesting that they may contribute to general SG epithelial differentiation rather cell type-specific differentiation. This is well-illustrated from the results pertaining to outgoing signaling events sent from endothelial cells to epithelial cells. While HSPG and IGF signaling from endothelial cells impinges upon all four epithelial cell types (Krt19⁺ ducts, MECs, end buds and basal), ADGRG signaling doesn’t affect the MECs. In contrast, EPHA signaling operates between endothelial cells, MECs and the end buds, while MPZ signals between endothelial cells, Krt19⁺ ducts, and end buds. Further follow-up studies are warranted not only to validate these observations but also to elucidate the molecular mechanisms underlying the differential utilization of specific signaling pathways and their functional consequences. In particular, we posit that a renewed interest in mechanistic and genetic studies of supporting cells in the developing SG will be quite informative given their prominent instructive roles as has been described in the literature.^17,62 It will be remiss not to acknowledge a major caveat of the aforementioned CellChat-based findings namely, that the inferred cell-to-cell communication is solely derived from transcriptomic based data of ligand-receptor interactions. This shortcoming notwithstanding, it is reassuring to note that recently described studies that include clonal lineage tracing, embryonic explants, and single-cell quantitative imaging have reaffirmed the roles of signaling pathways such as NOTCH, in the SG morphogenetic programs during early embryogenesis.^32,63 Similar innovative approaches and validation through parallel examination of corresponding signaling protein translation, processing, secretion, and diffusion, as well as in vivo genetic studies, represents a critical area of future research.⁶⁴

Our focus on MDK signaling in this study was driven by its top ranking in the CellChat analysis and its broad activity in both outgoing and incoming signaling across diverse cell populations. Furthermore, although a growing body of literature links MDK to various aspects of organogenesis—including growth, proliferation, survival, migration, angiogenesis, and repair—most relevant studies been conducted in adult tissues or within cancer contexts. Consistent with the CellChat results, our loss-of-function and gain-of-function experiments using the SMG demonstrate that MDK signaling plays a crucial role in the growth and maturation of the embryonic SG. Our efforts to tease apart the molecular components of the MDK pathway demonstrate that the master lineage transcription factor p63, specifically the ΔNp63 isoform acts upstream of MDK and likely directly controls its gene expression in the embryonic epithelial cell populations. We also found that MDK signaling intersects with Rock1, whose activity is essential for initiating epithelial clefts, a key event that drives branching morphogenesis.^41,42,43 Rock1 has also been shown to alter organ shape by altering actin-myosin mediated contractility, which is required for the assembly of fibronectin in the basement membrane during both cleft progression and regulation of focal adhesion formation in the epithelium.⁶⁵ Collectively, these findings support a model in which the p63-MDK-Rock1 axis plays a crucial role in regulating key epithelial cellular processes that underpin embryonic SMG development. However, our SMG explant-based studies do not address the contributions of other cell populations, particularly the mesenchymal cells which also serve as an important source of MDK. Future studies such as those employing conditional mouse knockout models will be critical to elucidate cell-specific functions of MDK.

While our study has primarily focused on embryonic development and differentiation programs of the mouse SG, the insights obtained may have broader relevance. We posit that the global signaling networks identified during mouse embryogenesis are conserved in human submandibular glands and may be re-engaged during regenerative processes, tissue repair following injury, and other stress-induced biological responses. Furthermore, it is noteworthy that there has been renewed interest in MDK signaling, particularly regarding its role in neoplastic diseases. A recent landmark study demonstrated that MDK is upregulated during aging and contributes to breast tumorigenesis by influencing progenitor cell populations.⁶⁶ A deeper understanding of MDK signaling, along with other similar pathways, could open new avenues for therapeutic interventions, particularly in the context of cellular therapy for SG diseases and injuries.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Rose-Anne Romano (rromano2@buffalo.edu).

Materials availability

This study did not generate new unique reagents.

Limitations of the study

In the present study, we utilized ex vivo organ explant models to discover a critical role for MDK in orchestrating branching morphogenesis programs in the murine SG. While our investigation employed a loss-of-function approach using the only chemical inhibitor currently available to assess the effects of MDK depletion, this method could be subject to potential off-target effects which cannot be entirely excluded. As such, the use of an MDK knockout animal model would offer a valuable complementary strategy. These studies would be particularly useful in evaluating a potential direct link, if any, between MDK and Rock1. Additionally, it is important to acknowledge the limitations of CellChat based analyses, particularly that the inferred cell-to-cell communication information from this bioinformatics tool is derived solely from transcriptomic output of ligands and receptors. Hence, as showcased in this study, experimental validation of cell communication mediators such as confirming cell type-specific protein expression with immunostaining and investigating their functional role through perturbation experiments will be a critical direction for future research.

Data and code availability

• •Data: Single cell RNA-sequencing and ChIP-sequencing data can be accessed from Gene Expression Omnibus (GEO) with the accession ID GEO:GSE150327 and GEO:GSE145264, respectively.
• •Code: All computational code used in this study is based on publicly available R packages, which are listed in the key resources table.
• •Other information: microscopy, western blot, and RT-qPCR data reported in this paper will be shared by the lead contact upon request.

Acknowledgments

This work was supported by National Institutes of Health/National Institute of Dental and Craniofacial Research (NIH/NIDCR) grants DE027660 to S.S. and R.-A.R. T.W. was supported by an F31 fellowship (NIH/NIDCR) DE032901 and by the State University of New York at Buffalo, School of Dental Medicine, Department of Oral Biology training grant (NIH/NIDCR) DE023526.

Author contributions

T.W., S.S., and R.-A.R. designed the experiments. T.W., J.O., and R.-A.R. performed experiments. T.W., J.O., S.S., and R.-A.R., analyzed the data. T.W., S.S., and R.-A.R., wrote the paper. All authors reviewed and edited the drafts and approved the final version.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-Actin, α-Smooth Muscle (Sma)	Sigma-Aldrich	A2547; RRID:AB_476701
Rabbit anti-Krt5	Gift from Julie Segre	–
Rabbit anti-Aquaporin 5	Alomone Labs	AQP-005; RRID:AB_2039736
Mouse anti-Cytokeratin 7	Abcam	ab9021; RRID:AB_306947
Guinea Pig anti-Krt14	(Rizzo et al., 2016)67	–
Rabbit anti-Mist1	Abcam	ab187978
Goat anti-Nkcc1	Santa Cruz Biotechnology	sc-21545; RRID:AB_2188633
Rabbit anti-ΔNp63	(Romano et al., 2006)68	–
Rabbit anti-p63α	Cell Signaling	13109; RRID:AB_2637091
Rat anti-Troma III	Development Studies Hybridoma Bank	http://dshb.biology.uiowa.edu/; RRID:AB_2133570
Rabbit anti-Rock1	Cell Signaling	4035; RRID:AB_2238679
Sheep anti-Midkine	R&D Systems	af7769; RRID:AB_2917965
Rabbit anti-Midkine	ProteinTech	11009-1-AP; RRID:AB_2250619
Rabbit anti-Sfrp1	Thermo Fisher Scientific	MA5-32675; RRID:AB_2809952
Rabbit anti-β-actin	Cell Signaling	4970; RRID:AB_2223172
Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 488	Invitrogen	A-21206; RRID:AB_2535792
Donkey anti-Goat IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488	Invitrogen	A-11055; RRID:AB_2534102
Goat anti-Rat IgG (H + L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 568	Invitrogen	A-11077; RRID:AB_2534121
Goat anti-Guinea Pig IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 568	Invitrogen	A-11075; RRID:AB_2534119
Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 568	Invitrogen	A-10042; RRID:AB_2534017
Donkey anti-Mouse IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 488	Invitrogen	A-21202; RRID:AB_141607
Donkey anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 647	Invitrogen	A-31573; RRID:AB_2536183
Donkey anti-Sheep IgG (H + L) Fluorescein (FITCH)-conjugated AffiniPure Secondary Antibody	Jackson ImmunoResearch Labs	713-095-003; RRID:AB_2340718
Chemicals, peptides, and recombinant proteins
(Z)-4-Hydroxytamoxifen	HelloBio	HB6040
Histo-Clear II	National Diagnostics	HS-202
TRIzol™ Reagent	Invitrogen	15596018
Bovine Serum Albumin Fraction V, heat shock	Roche	03116956001
DirectPCR Lysis Reagent (Tail)	Viagen Biotech	102-T
Restore™ Western Blot Stripping Buffer	Thermo Fisher Scientific	21063
DMSO (Dimethyl sulfoxide)	Sigma	D8418-100 ML
Mouse Midkine Recombinant Protein	Peprotech®	315-25-100UG
iMDK	Tocris Bioscience	5126
Critical commercial assays
Mouse on Mouse (M.O.M.) Basic Kit	Vector Laboratories	BMK-2202
Universal Green qPCR MasterMix 2×	Lamda Biotech	qMX-Green
cDNA synthesis: 5× All-in-One RT Plus MasterMix	Lamda Biotech	G209
Direct-zol RNA MiniPrep kit	Zymo Research	R2070
GoGreen Taq Plus Master Mix	Lamda Biotech	GD124
Chemiluminescent Substrate kit	Sera care	5430–0049
Deposited data
Single cell RNA-sequencing data	Hauser et al.22	GSE150327
ChIP-sequencing data	Sangwon Min et al.69	GSE145264
Experimental models: Organisms/strains
Mouse: B6.Cg-Ndor1Tg(UBC−cre/ERT2)1Ejb/1J	Jackson Laboratory	007001; RRID:IMSR_JAX:007001
Mouse: C57BL/6J	Jackson Laboratory	000664; RRID:IMSR_JAX:000664
Mouse: ΔNp63 fl/fl	Gifted (Chakravarti et al.50)	–
Oligonucleotides
Genotyping primers: Cre (F): 5′-GAG TGA TGA GGT TCG CAA GA-3′	Integrated DNA Technologies	–
Genotyping primers: Cre (R): 5′-CTC CAC CAG AGA CGG AAA TC-3′	Integrated DNA Technologies	–
Genotyping primers: ΔNp63 floxed (F): 5′-ACA GTC CTC TGC TTT CAG C-3′	Integrated DNA Technologies	–
Genotyping primers: ΔNp63 floxed (R): 5′-TTC ACA TTC ACA CAG ACA GCT CC-3′	Integrated DNA Technologies	–
Genotyping primers: ΔNp63 Knockout (F): 5′-TAC TTT CAA ACA GCT ATT CTC AGG-3′	Integrated DNA Technologies	–
Genotyping primers: ΔNp63 Knockout (R): 5′-CAC ACA GCA CTG GCC TTG C-3′	Integrated DNA Technologies	–
qRT-PCR primers: ΔNp63 (F): 5′-TGC CCA GAC TCA ATT TAG TGA GC-3′	Integrated DNA Technologies	–
qRT-PCR primers: ΔNp63 (R): 5′-GAC GAG GAG CCG TTC TGA ATC-3′	Integrated DNA Technologies	–
qRT-PCR primers: Krt14 (F): 5′-GTC TGC TGG AGG GAG AGG AC-3′	Integrated DNA Technologies	–
Software and algorithms
ZEN imaging software	Zeiss	https://www.zeiss.com/microscopy/us/products/microscope-software/zen.html
Adobe Illustrator	Adobe	www.adobe.com
CFX Manager	Bio-Rad	1845000
Image Lab	Bio-Rad	1709690
Minitab Statistical Software v21.2	Minitab®	https://www.minitab.com/en-us/
Fiji	ImageJ	https://imagej.net/software/fiji/downloads
CellRanger v7.1.0	Zheng et al.70	https://support.10xgenomics.com/single-cellgeneexpression/software/overview/welcome
Seurat v5.1.0	Bulter et al.71	https://github.com/satijalab/seurat
Rstudio v2024.04.2 + 764	The R Project for Statistical Computing	https://www.r-project.org
R v4.4.1	The R Project for Statistical Computing	https://www.r-project.org/
CellChat v2.1.2	Jin et al.24	https://github.com/jinworks/CellChat
Integrative Genomics Viewer	Robinson et al.72	https://igv.org/doc/desktop/
SRA toolkit v3.0.5	NIH	https://github.com/ncbi/sra-tools/wiki
Other
VECTASHIELD® Antifade Mounting Medium with DAPI	Vector Laboratories	H-1200
VECTASHIELD® Antifade Mounting Medium	Vector Laboratories	H-1000
SecureSeal™ imaging spacers	Grace Bio-Labs	654004
Nunc™ Lab-Tek™ II Chambered Coverglass	Thermo Fisher Scientific	155409
Penicillin-Streptomycin (10,000 U/mL)	Gibco	15140122
DMEM/F-12, HEPES	Gibco	11330032
Transferrin	Sigma	45-T8158-100 MG
Ascorbic acid	Sigma	45-A4403-100 MG
12 mm Transwell® with 3.0 μm Pore Polycarbonate Membrane Insert	Corning	3402
Biosmashers	TaKaRa	9790A
TOPRO iodide	Invitrogen	T3605

Experimental model and study participant details

Animal studies

All animal experiments and procedures were performed in accordance with Institutional Animal Care and Use Committee (IACUC) regulations of the State University of New York at Buffalo (Protocol number: ORB10074Y). Animal model design was adopted according to all the recommendations of the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines. Pregnant females were euthanized by CO₂ inhalation followed by cervical dislocation, which is the standard recommended method. Embryos were removed and euthanized via decapitation. Wild type C57BL/6J (Stock No. 000664) and UBC^CreERT2 (B6.Cg-Ndor1^{Tg(UBC−cre/ERT2)1Ejb}/1J; Stock No. 007001) mice were purchased from Jackson Laboratory (Bar Harbor, Maine). The ΔNp63-floxed (ΔNp63 ^fl/fl) mice were provided by Elsa Flores and have been characterized previously.⁵⁰ All mice were maintained on a C57BL/6J background. For timed pregnancies, mice were mated and noon of the day the vaginal plug was observed was considered E0.5. Embryos were harvested at either E13, E13.5, E14, or E16 as indicated. Sex was not considered as there is no sexual dimorphism at this stage in mice.⁷³

Method details

Ex vivo explant experiments

For ex vivo explant studies, the submandibular/sublingual glands were harvested under a dissecting microscope (Leica MZ6) from wild type C57BL/6J or ΔNp63 mice at embryonic day E13.5. Harvested glands were cultured on 12 mm Transwell with 3.0 μm Pore Polycarbonate Membrane Insert (Corning) at the air/liquid interface floating on ex vivo culture medium in 12 well plates as described by Steinberg et al.⁷⁴ Each embryo was transferred to one transwell ensuring no more than 2 glands on each filter. Culture medium contained DMEM/F12 supplemented with 100U/ml penicillin, 100 μg/ml streptomycin, 150 μg/ml ascorbic acid, and 50 μg/ml transferrin. Explants were incubated at 37°C and 5% CO₂ for 72 h. Culture media was replaced every 24 h. For inhibitory experiments of Midkine, control embryonic murine explants were treated daily with 0.1% v/v DMSO or 1 μM of iMDK (Tocris Bioscience) to inhibit midkine. For recombinant experiments, controls were treated with 0.1% v/v distilled water or 100 ng/ml of recombinant midkine (PeproTech) was added to cell culture medium to evaluate gain-of-function. For conditional deletion of ΔNp63 experiments, all explants were cultured with 2 μM of activated tamoxifen (4-OHT) supplemented to the culture medium. After 72 h in culture, glands were either fixed with 4% w/v PFA (paraformaldehyde), paraffin embedded and processed for immunofluorescence or frozen immediately on dry ice.

RNA isolation and real-time quantitative reverse transcription PCR (qRT-PCR)

Total RNA from frozen explants was extracted as previously described.⁶⁹ Frozen explants were homogenized using TRIzol (Invitrogen) with BioMashers (TaKaRa). RNA was further isolated and purified using the Direct-zol RNA Miniprep kit (Zymo Research). Isolated RNA was reverse transcribed using the cDNA Synthesis: 5× All-in-One RT Plus MasterMix kit (Lamba Biotech). qRT-PCR was performed on a CFX96 Touch Real-Time PCR Detection System (Bio-Rad) using Universal Green qPCR MasterMix 2× (Lamda Biotech). All qRT-PCR assays were performed in triplicates in at least three independent experiments. For each replicate (n) at least two control glands and 4 treated glands were pooled, respectively. Relative expression values of each target gene were normalized to hypoxanthine guanine phosphoribosyltransferase (Hprt) expression. Primer sequences can be found in the key resources table and Table S1.

Protein extraction and western blot analysis

Protein was extracted from at least three pooled control or iMDK treated explants originating from the same liter in RIPA buffer containing 2 μg/ml phosphatase and protease inhibitor cocktail (G-Biosciences) using BioMashers (TaKaRa). This process was repeated for each replicate. Protein concentration was determined by using the Bio-Rad Bradford protein assay. The protein samples were separated by SDS-PAGE and transferred to a PVDF membrane and blocked with 5% w/v non-fat dry milk in Tris-Buffered Saline with Tween 20 (TBST). Membranes were incubated with their Rock1 (1:2,500, Cell Signaling Technology), Sma (1:5,000, Sigma, 1A4), Sfrp1 (1:7,500, Thermo Fisher Scientific, MA5-32675), or ΔNp63 (1:10,000, Cell Signaling Technology, D2K8X) primary antibodies overnight at 4°C. KPL LumiGLO Reserve Chemiluminescent Substrate kit (Sera care) was applied to the membrane and the ChemiDoc MP Imaging System (Bio-Rad) was used for detection. Densitometry (measurement of band intensity) was performed by using ImageJ (NIH; Bethesda, Maryland). The blot was stripped using stripping buffer (Thermo Fisher) and was re-probed with β-actin (Cell Signaling Technology, 1:10,000 dilution) for normalization (n = 3).

Immunofluorescence and imaging

Paraffin embedded submandibular gland tissue were deparaffinized and sectioned to 5 μm thickness for immunofluorescence analysis. Antigen retrieval was performed with pH6 sodium citrate buffer and the sections were blocked with the M.O.M kit (Vector Laboratories). Primary antibodies used at the indicated dilutions include Midkine (R&D Systems, 1:100), Midkine (ProteinTech, 1:100), K14 (Rizzo et al., 2016,⁶⁷ 1:100), Sma (Sigma 1A4, 1:200), Nkcc1 (Santa Cruz Biotechnology, 1:100), ΔNp63 (Cell Signaling Technology, 1:50), ΔNp63 (⁶⁸,1:50) Mist1 (Abcam, 1:100), Aqp5 (Alomone Labs, 1:100), Tubb3 (R&D Systems, 1:100), K7 (Abcam, 1:50), K5 (gift from Dr. Julie Segre, 1:100), Troma-III/K19 (Development Studies Hybridoma Bank, 1:50), and Rock1 (Cell Signaling Technology, 1:25). Alexa 488 or Alexa 568 conjugated secondary antibodies were purchased from Life technologies. Sections were counterstained with TOPRO iodide (1:5000, Invitrogen) for nuclear staining and mounted using VECTASHIELD Antifade Mounting Medium (Vector Laboratories). Primary antibody specificity was assessed by immunofluorescence, with representative negative control images shown from tissue sections incubated with secondary antibodies alone, without primary antibodies, are shown in . Samples were imaged using an Andor Dragonfly Spinning Disk Confocal Microscope with Fiji.⁷⁵ Microscopy data in this study was acquired at the Optical Imaging and Analysis Facility, School of Dental Medicine, State University of New York at Buffalo.

Quantification of salivary gland differentiation markers

DMSO control and iMDK treated explants were paraffin embedded, sectioned and stained with antibodies to evaluate the alterations in the basal cell (ΔNp63⁺/K14⁺), MEC (ΔNp63⁺/Sma⁺), and acinar cell (Nkcc1⁺/Mist1⁺) populations. The percentage of the basal cell population was calculated by counting the total number of ΔNp63^+/K14⁺ double-positive cells and dividing by the total number of p63⁺ cells (n = 3). The percentage of the MECs was calculated by counting the total number of p63⁺/Sma⁺ double-positive cells and dividing by the total number of p63⁺ cells (n = 3). The percentage of acinar cell population was calculated by counting the total number of Nkcc1⁺/Mist1⁺ double-positive cells and dividing by the total number of Nkcc1⁺ cells (n = 3). Quantification analyses were performed using three to five fields of view (400× confocal images) using ImageJ (NIH; Bethesda, Maryland). Data are reported as mean ± standard deviation (S.D.).

Whole mount explant immunofluorescence staining

Fixed explants were permeabilized with 0.5% v/v Triton X-100 in 1× PBS and then blocked overnight using the M.O.M. kit (Vector Laboratories) or 5% w/v BSA (Bovine Serum Albumin). Explants were incubated with primary antibodies at dilutions as indicated in the Immunofluorescence and Imaging section for 48 h at 4°C and subsequently washed with 0.1% v/v Tween-20-PBS. Explants were incubated with Alexa 488, Alexa 647, or Alexa 568 conjugated secondary antibodies for 24 h and then subsequently washed. Explants were then counterstained either with VECTASHILED Antifade Mounting Medium with DAPI (Vector Laboratories) or TOPRO iodide (1:1000, Invitrogen) for a minimum of 1 h and mounted on a coverslip with secure-seal imaging spacers (Grace Biolabs) using VECTASHIELD Antifade Mounting Medium (Vector Laboratories). Andor Dragonfly Spinning Disk Confocal Microscope with Fiji was used to image the explants.⁷⁵ Fluorescence intensity quantifications were performed using ImageJ (NIH; Bethesda, Maryland). Videos of explants were generated using the “3D Project” function in ImageJ (NIH; Bethesda, Maryland).

Spooner’s ratio quantification

The number of end buds were counted at the time of dissection (E13.5) and at the end of the experiment (72hrs) using ImageJ (NIH; Bethesda, Maryland). Spooner’s ratio was calculated by dividing the final endbud number by the initial end bud number for each explant. For normalization, the control was set to 1 in each sample and the average was taken for DMSO control and iMDK treated groups. Paired t test was used to determine significance. Quantified values were reported as mean ± standard deviation (S.D.) of 10 or more independent experiments.

Single cell RNA-sequencing

Fastq files from scRNA-seq datasets at E12, E14, and E16 SMG generated by Hauser et al.²² were downloaded from Gene Expression Omnibus repository (GSE150327). Datasets were updated with 10× Genomics CellRanger version 7.1.0 software.⁷⁰ Reads were aligned to the mm10 reference genome using the pre-built annotation package provided on the 10× Genomics website. Only confidently mapped, non-duplicate with valid barcodes and UMIs were used to generate a gene-barcode matrix for further analysis. Computational analysis was performed using R & RStudio software and SEURAT package version 5.1.0.⁷¹ Basic filtering was conducted (nUMI >500, nGene >200, log10GenesPerUMI >0.80) to remove low quality cells. High mitochondrial transcript load (>20%) were filtered from the analysis. The data were normalized using SCTransform function in Seurat. Principle component analysis (PCA) and Uniform Manifold Approximation and Projection (UMAP) algorithm were used for dimensionality reduction and visualization, followed by the construction of a Shared Nearest Neighbor (SNN) graph and clustering analysis. Clustree version 0.5.1 package was used to identify the appropriate number of clusters per embryonic timepoint. ‘FindAllMarkers’ function from Seurat was used to find differentially expressed genes for each cluster to identify gene expression markers. After cell type annotations, data integration with SEURAT was performed for embryonic stages.

Subsetting of epithelial cells from integrative seurat object

Epithelium populations from the integrative Seurat object were separated from the previously annotated dataset utilizing the subset function from Seurat. Epithelial subset was re-normalized and scaled to generate a new Seurat object. Clusters were identified by known cellular markers (Aqp5, Sox10, and Bhlha15 for endbud; Trp63, Krt14, Myh11, and Acta2 for basal/MEC; and Krt7, Krt19, and Krt18 for Krt19⁺ duct). Data integration of the epithelium was used for correlation analysis.

Correlation analysis

The integrative epithelium Seurat object containing cells from E12-E16 end bud, Krt19⁺ duct, MEC and basal cell populations was used to obtain a gene expression matrix. The ‘corr.test’ function in R was used to identify genes that are correlated to Midkine gene expression. In order to narrow down transcription factors that were correlated to Midkine, the subsequent list was compared to a list of transcription factors.⁷² The top 15 positive and top 10 negative correlated transcription factors were plotted using the ‘corrplot’ version 0.9.5 package.

Systematic inference of cell-cell communication

The open-source R package “CellChat” version 2.1.2²⁴ was used to study the interactions between cells and identify communication patterns at a single cell resolution. This tool uses a Hill-function-based mass action model approach along with differential expression analysis to predict statistically significant cellular interactions. The tutorial in GitHub (https://github.com/sqjin/CellChat) was applied to convert the Seurat Objects into CellChat objects for subsequent analysis. Erythroid cells were excluded from analysis.

Chromatin immunoprecipitation-sequencing (ChIP-seq)

Previously published ChIP-seq datasets (GSE145264) were mapped to the Mus musculus genome (mm9 build) and ChIP-seq signals were visualized by using Integrative Genomics Viewer (IGV).^69,76

Quantification and statistical analysis

Statistical analysis

Statistical analysis was conducted with Minitab Statistical Software. Quantified results were reported as mean ± standard deviation (S.D.) of three or more independent experiments. Differences were considered statistically significant when P-values <0.05. Data comparison between control and iMDK or KO groups were performed with paired Student’s t test with false discovery rate of less than 5%. Specific n values are listed in the figure legends. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001.

Published: January 29, 2026