All aspects of mouse experiments were approved by the Finnish National Board of Animal Experimentation (ESAVI/1284/04.10.07/2016). The plug day was considered as embryonic day E0 and the date of birth as post-natal day P0. All embryos were staged according to limb morphological criteria.

Wild-type ICR mice and Rosa-R26R reporter transgenic mice [R26R-TdTomato (Madisen et al., 2010), R26R-Confetti (Snippert et al., 2010), and R26R-mT/mG (Muzumdar et al., 2007)] were purchased from Jackson Laboratory, USA. Fucci (Sakaue-Sawano et al., 2008) fluorescent cell cycle reporter mouse strain was used for cell cycle analyses. The reporter lines were crossed with the K14-Cre43 line (Andl et al., 2004), kindly provided by Dr. Sarah Millar, to induce genetic recombination. R26R-mT/mG mice were crossed with aSMA-Cre line (LeBleu et al., 2013), kindly provided by Prof. Kari Alitalo, for recombination tests.

All tissues were fixed with 4% paraformaldehyde (PFA). For E17 and E18 stages, tissues were decalcified with EDTA prior embedding. For histology and immunohistochemistry, tissues were embedded in paraffin and 5 μm thickness sections were used.

LGs were dissected out as previously described (Finley et al., 2014). Total RNA was extracted from whole LGs (mesenchyme and epithelium) of embryonic (E18) and adult (34 weeks of age) animals. E18 samples were composed of at least 10 LGs (5 individuals) pooled together. Biological triplicates for each sample were included in the analysis. The quality and concentration of the extracted RNA was monitored using a nanodrop spectrophotometer (ND-1000, Fisher Scientific). The transcriptomic analysis was performed by the Functional Genomics Unit (FuGU, Helsinki, Finland). Samples were hybridized on Affymetrix GeneChip® Mouse Transcriptome Assay 1.0 (Affymetrix, Santa Clara, CA, USA) microarrays, and data analysis was performed with the Affymetrix Transcriptome Analysis Console (TAC) Software.

LG cultures were established using the Saxén protocol (Munne et al., 2013) and the method described previously (Finley et al., 2014). Culture medium was composed of DMEM/F-12, GlutaMAX supplement (Thermo Fisher Scientific) complemented with 10% FBS, 0.1% Penicillin/Streptomycin and 0.1% Ascorbic Acid (Sigma Chemical Co.). The medium was changed every second day. Dissected LGs were cultured for a maximum of 5 days, at 37°C, in a controlled atmosphere (5% CO₂).

For Notch pathway inhibition experiments, DAPT 10 μM (Sigma-Aldrich) was added to the medium (Michon et al., 2007). Contralateral controls were supplemented with DMSO (Sigma-Aldrich). Medium and cultures were protected from light. Medium was changed every day. Each experiment was replicated at least five times with over 10 LGs each time.

Both whole mount and immunohistochemistry on slides were performed on PFA-fixed samples. For whole mounts, non-specific staining was blocked by incubating the organs over night at +4°C in a blocking solution (5% donkey/goat serum + 1% BSA in PBS-0.1% Triton). Primary antibodies (see Table below) were diluted in a fresh blocking solution and incubated o/n at +4°C. Subsequently, the glands were incubated with secondary antibodies (see below) o/n at +4°C in PBS-0.1% Triton +1% BSA.

For immunohistochemistry staining on slides, an antigen retrieval step was added to the protocol. Antigen retrieval was performed in 10 mM Na-citric acid (pH 6.0), using an antigen retrieval device (Aptum Biologics Ltd).

Primary antibodies used:

Anti-Notch2 (ImmunoWay, YC0069) was kindly provided by Irene Ylivinkka and Arvydas Dapkunas.

Secondary antibodies used included anti-rabbit AlexaFluor 488 (Life Technologies), anti-mouse AlexaFluor 568 (Life Technologies) and anti-rat AlexaFluor 647 (Invitrogen). Secondary antibodies were diluted at 1/500 for whole mounts and at 1/400 for immunohistochemistry on slides.

In both protocols, the samples were counterstained with Hoechst 33342 (1/2000, Life Technologies) for nuclei staining, and mounted in Vectashield (Vector Laboratories) prior to microscopy visualization.

RNeasy microkit (Qiagen) was used according to the manufacturer's instructions to extract total RNA from dissected LGs of animals ranging from E15 to adult (34w). cDNAs were generated from biological triplicates by using the SuperScript™ III Reverse Transcriptase kit (for RT PCRs, Invitrogen) or the QuantiTect Reverse Transcription Kit (for multiplex PCRs, Qiagen, 205310), according to the provider's recommendations.

Subsequently to cDNA synthesis, reverse transcription PCRs for Aquaporin 1, 5, and 8 were performed using an annealing temperature of 60°C for 40 cycles. One hundred fifty nanograms of total RNA was used for each reaction.

The primer sequences are given in the following table:

Multiplex qRT-PCRs (CFX96 Touch™ Real-Time PCR Detection System, Bio-Rad) were performed using iTaq universal probe super mix (Bio-Rad, 1725130). Ten nanograms of cDNA were used per reaction. Probe combinations (PrimePCR Probe Assay, Bio-Rad):

Combination 1: GAPDH-Cy5 (qMmuCEP0039581); Krt14-Hex (qMmuCEP0058885); Acta2-Cy5.5 (qMmuCIP0032840); Krt19-FAM (qMmuCIP0033699).

Gene expression levels were normalized using GAPDH expression levels.

Bright field organ morphology was imaged using a Zeiss Lumar stereomicroscope. Immunofluorescence confocal imaging was performed using a Leica TCS SP5 and SP8 confocal microscopes.

Images were analyzed and quantitative measurements performed with Imaris 8.4.1 (Bitplane) software. For the cell cycle analyses with the Fucci mouse line, only the cells that were distinctly identified as either G1 (red), or S/G2/M (green) were included. Proliferation ratios were obtained by dividing the number of green cells by the number of (green + red) cells.

Student t-test (paired, two-tailed) was used for statistical analysis. All statistical analyses were performed on biological triplicates. For cell cycle study, a total of 9 TEBs from 3 LGs were used for each developmental stage; for bifurcations analysis, at least 3 LGs of each stage were analyzed. For aSMA expression in Notch inhibition study, biological triplicates (at least 3 cultured LGs per condition and per time point) were used. Moreover, technical triplicates were used to validate the experiments.

Images were processed with Photoshop CC 2017 and Illustrator CC 2017 software (Adobe Systems).

Previous studies reported the expression of Krt14 and aSMA in the embryonic and adult LG (Dean et al., 2004; Hirayama et al., 2016). However, their embryonic expression pattern have not been described so far. Therefore, we focused on these two genes to characterize LG epithelium morphogenesis. By using qRT-PCR, we demonstrated that Krt14 and aSMA expression level constantly increased from E15 to 34 weeks of age (Figure S1A).

To further analyse the localization of Krt14+ and aSMA+ cell populations during LG morphogenesis, we studied Krt14 and aSMA patterns using immunofluorescence staining and confocal microscopy. At E15, Krt14 was broadly found in the most external cell layers of the bud epithelium, and not only localized in the basal cell layer (Figures 2A,B). Interestingly from E16 to E18, Krt14 pattern appeared more restricted to the basal cell layer of both TEBs and branches (Figure 2B). In addition, our observations of Krt14 pattern suggested that the luminal cells of the duct were Krt14-negative at E18 (Figure 2B).

In contrast, aSMA+ cells were scattered throughout the bud epithelium at E15 (Figures 3A,B). Later on, aSMA+ cells were progressively restricted to the basal cell layer of the TEBs from E16 to E18 (Figures 3A,B). A close-up visualization at E18 revealed a partial overlap of the Krt14+ and aSMA+ cell populations (Figure 3C). Due to this incomplete overlap, we found three cell populations in the E18 TEBs, namely the Krt14+ cells, the aSMA+ population and the double positive Krt14+;aSMA+ cells (Figure 3C).

To visualize the involvement of the aSMA+ cells and Krt14+ cells during LG morphogenesis, we crossed the aSMA-Cre (LeBleu et al., 2013) and K14-Cre43 (Andl et al., 2004) lines with reporter mouse strains.

Unfortunately, the aSMA-Cre transgenic line displayed a very low recombination level in the LG (Figure S2). As a result, it was impossible to use it for any genetic fate mapping experiment. Therefore, we took advantage of the K14-Cre43 mouse strain, reported to have a high recombination efficiency in various ectodermal organs (Jussila et al., 2015; Shirokova et al., 2016). It allowed the visualization of the LG epithelial morphological changes from E15 to E18 (Figure 4A). As expected from the Krt14+ cell localization (Figure 2), the Krt14+ cells and their progeny outlined perfectly the branching morphogenesis. Moreover, by comparing cultured LGs to the ones pictured in situ, we saw that LG development was similar in and ex vivo (Figure 4A), providing a great tool for microenvironment modulations.

By following LG postnatal expansion, we found that the epithelial compartment continued its growth until around P50 (Figure 4B). We investigated the aSMA and Krt14 localization in postnatal LG, and noticed that the MECs, recognizable by their long processes, expressed both aSMA (as previously reported, Makarenkova and Dartt, 2015) and Krt14 (Figure S3A). No other cells in the acini expressed any of these markers. Krt14 was also found in the basal cell layer of the ducts. The postnatal continuity of LG morphogenesis went along with a switch in Aquaporins expression, known to be responsible for the tear film composition (Schey et al., 2014). While Aquaporin1 expression decreased after the embryonic development, Aquaporin8 was greatly enriched in postnatal LG (Figure S3B).

Our results showed that aSMA and Krt14 can be used as embryonic markers for basal epithelial cell layer of the forming LG. Moreover, the presented data confirmed the continuity of LG formation in postnatal stages. However, the mechanisms involved in the LG growth remained elusive.

To comprehend the cellular mechanisms involved in LG enlargement, we focused on identifying possible proliferative zones. Therefore, we used the Fucci reporter mouse line (Sakaue-Sawano et al., 2008), in which cells express a green fluorescence in S/G2/M cell cycle phases, and red fluorescence in G0/G1 phases. The cells of the epithelial bud expressed green fluorescence, reflecting a global and unpatterned proliferation at E15 (Figure 5A). However, after the first branching event, the proliferation gradually decreased and seemed more localized in the acinar compartment (Figure 5A). We analyzed the cell cycle state of both the acinar and ductal compartment during branching morphogenesis, and detected proliferating cells in TEB basal and suprabasal cell layers (Figure 5B). Although some of the TEB cells remained in an active cell cycle throughout LG morphogenesis, the ductal cells progressively exited the cell cycle along with LG morphogenesis. By E18, almost no ductal cells were proliferating anymore (Figure 5B). As we noted a decrease of the green fluorescence in the TEBs, we quantified the proliferating cells in the TEBs from E16 to E18. While at E16, about 55% of the TEB cells were proliferating, by E18, only 27% of the cells were still in an active proliferative state (Figure 5C). This observation reflected the progressive decrease and restriction of cell proliferation during LG morphogenesis. We confirmed this gradual restriction of proliferation by quantifying the phosphorylated-Histone3+ cells (pH3), epigenetic modification found only in M phase cells (Hans and Dimitrov, 2001; Figures 5D–F). Nevertheless, we showed the continuation of LG growth after birth (Figure 4B). The general decrease of cell proliferation we observed seemed conflicting with the postnatal expansion of the LG tree. Therefore, we investigated other mechanisms that could support LG growth.

The expansion of a non-proliferative epithelium can result from (i) spreading of a cell layer by change of cell shape, (ii) cellular intercalation or (iii) convergent extension (for review, Keller et al., 2008). During convergent extension, cell rearrangement involves cell intercalation, and lessening of cell layers. This mechanism, in which epithelial cells change layers, can be an advantage for the fast expansion of an epithelial domain. Notably, epithelial rearrangement has been shown to be involved in branching morphogenesis (for review, Wang et al., 2017). To investigate the involvement of this process during LG expansion, we took advantage of the R26R-Confetti reporter mouse line (Snippert et al., 2010). Unlike for a clonal analysis, we used a constitutive Cre (K14-Cre43) to label as many epithelial cells as possible. Upon continuous recombinase activity, cells can be sorted into two populations, depending on the first recombination event (Figure S4A): GFP and/or YFP expressing cells, and RFP and/or CFP expressing cells (Figure S4B). Therefore, we separated epithelial cells into two domains: the orange cell domain (GFP+ and/or YFP+ cells), and the violet cell domain (RFP+ and/or CFP+; Figure S4C). Surface renderings of ex vivo cultures revealed a highly intermingled epithelial population, composed of mix of the two domains during LG morphogenesis. We detected single cells of one origin in the middle of a domain of the other origin. This observation would be incompatible with a simple proliferation-based territory expansion, and hinted toward an expansion of the epithelium via convergent extension (Figure 6A).

Interestingly, a few cells did not display any fluorescence with the Confetti reporter, already at E15 (data not shown). We made similar observations with K14-Cre43;R26R-TdTomato at E17 (Figure S5A) and with K14-Cre43;R26R-mT/mG at E16, E17 and E18 (Figure 6B). We hypothesized that these unlabeled cells could arise from Krt14 negative origin, or from cells escaping the recombination. To rule out a low efficiency of the K14-Cre43 recombinase, we localized Krt14+ cells in the K14-Cre43;R26R-mT/mG embryos (Figure S5B). Some of the mTomato+ cells (not recombined) were Krt14-negative, directing us toward a non-epithelial origin. Moreover, we detected clumps of mesenchymal cells in close contact to the acinus and ductal epithelium (Figure 6B). As cell morphology seemed to indicate a possible mesenchymal cell intercalation within the epithelial compartment (Figure S5B), we analyzed E-Cadherin expression, as E-Cadherin is known to act as a mesenchymal to epithelial (MET) effector (Vanderburg and Hay, 1996). Notably, we detected mesenchymal cells at close proximity to the LG epithelium that presented E-Cadherin at their membrane at E16 (Figure 6C). This observation could indicate the involvement of MET during LG morphogenesis, as previously suggested (Dean et al., 2004).

Collectively, we showed that LG epithelium growth is primarily fuelled by cell proliferation in the acini. However, epithelial cell rearrangement and MET might contribute to the ductal tree expansion. Nonetheless, the luminal domain morphogenesis process is still elusive.

Tubulogenesis is associated with lumen formation, which can be dependent (Wells and Patel, 2010) or independent (Nedvetsky et al., 2014) of cell death. Therefore, we investigated if apoptosis was involved in LG lumen formation. In the same way as in salivary gland duct development (Nedvetsky et al., 2014), we observed the formation of microlumens at E16, concomitantly to the first apoptotic events (Figure 7). This phenomenon peaked around E17, the microlumens joining to form a continuous lumen. By E18, apoptotic cells were close to absent and the lumen was already formed in the larger ducts.

Krt19 has been reported to be expressed in the salivary gland ductal cells (Ogawa et al., 2003), but has never been reported in LG development context. We performed qPCR analysis and found an increase in Krt19 expression during embryonic LG morphogenesis (Figure S1B). Therefore, we followed Krt19 pattern by immunofluorescence staining and confocal microscopy from E15 to E18 (Figures 8, 9).

Interestingly, we found Krt19 in the LG epithelium at E15, prior to any duct formation. We used short-term ex vivo cultures to study this phenomenon (Figure 8A). From the 30 glands dissected at E15 analyzed, around 17% (5 glands) presented a pre-bud phenotype, consisting in a very round LG. About 67% (20 glands) presented an elongated bud, and roughly 17% (5 glands) were already displaying the formation of the first branching event. This transition from pre-bud to the first branching event recapitulated what we observed during LG formation in vivo. While Krt19 was seen in a very small and restricted domain in the pre-bud stage, upon bud elongation, Krt19+ domain transformed into a 3D spring-like shape. This 3D conformation was transient and rapidly extended within the epithelium domain during the initiation of the branching morphogenesis. Moreover, the ex vivo results (Movie 1) were supported by similar observation in vivo (Movie 2), ruling out a possible artifact from the ex vivo cultures. Interestingly, E-Cadherin seemed to be more expressed around the spring-like shape (Figure 8B). This observation might reflect the importance of cell-cell contacts in the Krt19+ cells during the transition from E15 to E16.

Subsequently, we investigated the Krt19 localization during the rest of LG morphogenesis. From E16 to E18, Krt19 was detected in the forming ducts (Figure 9A), outlining the progressive formation of the LG tree. A closer observation at E18, when the ducts are already well-extended and the lumen formed, revealed Krt14+ cells in the ductal basal cell layer, and Krt19 expression in the luminal cells (Figure 9B). Notably, this ductal organization remained in all stages of the maturing LG (Figure S6).

We took advantage of this marker to analyse further the LG branching process. We quantified both the number of bifurcations (Figure 10A) and the number of TEBs (Figure 10B) between E16 and E18. Despite the decrease of cell proliferation, the formation of new ducts and TEBs seemed to follow a linear increment, reflecting LG morphogenesis regulated pace. Interestingly, by comparing four E18 LG trees, we realized that the junction pattern differed from one gland to another (Figure 10C). Two types of branch formation are described in a glandular tree, i.e. terminal bifurcation (a TEB splitting and forming two growing branches), or lateral branching (a new branch forming from an established one; for review, Affolter et al., 2009). At E18, both lateral branching and terminal bifurcations could be observed in all the studied glands. Notably, the positions of terminal bifurcations and lateral branches seemed to be stochastic, rather than predetermined, giving the tree final shape a large pattern variability.

The formation of lateral branches requires a degree of plasticity from the pre-established ductal cells in order to form a new sprouting TEB (Watanabe and Costantini, 2004). Therefore, we investigated the molecular regulation of the duct identity.

Notch1 has recently been associated with branching morphogenesis in the LG context (Dvoriantchikova et al., 2017). Moreover, Notch signaling has been linked to the maintenance of suprabasal cells in the salivary gland TEBs (Garcia-Gallastegui et al., 2014), and to cell differentiation in other branched organs, such as lung and liver (Tsao et al., 2008; Falix et al., 2014).

We used a transcriptomic analysis to investigate the expression levels of the Notch pathway elements at E18 and 34 weeks of age. This approach demonstrated an enrichment of various Notch pathway elements in embryonic LG. Interestingly, we found that Notch2, its ligand Jagged1 (Shimizu et al., 1999) and its target gene Hey1 (Rutenberg et al., 2006) were enriched during LG formation (Figures S7A,B). Quantitative PCR analysis confirmed the expression of Notch2 and Hey1 during LG morphogenesis from E15 to 34 weeks of age (Figure S7C), reflecting the involvement of Notch2 activity in forming LG, as well as in adult LG. We investigated Notch2 localization and activity by immunohistochemistry staining, targeting its intracellular domain (ICD), which is cleaved and translocated to the nucleus upon activation. We detected the cleaved Notch2 ICD in an increasing number of epithelial cells in both the TEB and the ductal territory from E16 to E18 (Figures 11A–C).

To study further the contribution of Notch pathway in LG morphogenesis, we used a γ-secretase inhibitor (DAPT) in ex vivo cultures (Figure 12). While having an effect on off-targets (Zhao et al., 2008; Yoo et al., 2012), DAPT is a small molecule widely used to impair NICD cleavage, and consequently Notch activation (Jiang et al., 2011; Dvoriantchikova et al., 2017; Jing et al., 2017). We validated the efficiency of the inhibition by assessing the decrease of Hey1 expression via qPCR (Figure S7D), and the drastic reduction of nuclear Notch2 ICD (Figure S7E). We observed that inhibition of the Notch pathway resulted in hollow TEBs already after 1 day of treatment (Figure 12). In line with previous report (Dvoriantchikova et al., 2017), we also observed extra-branching after 2 days of treatment. While the loss of TEB supra-basal cells, resulting from apoptosis as previously shown in salivary gland (Garcia-Gallastegui et al., 2014; Figure S8), was rescued promptly after DAPT treatment retrieval (Figure 12), the extra-branching phenotype remained.

To analyse the effect of Notch pathway on epithelium patterning, we investigated the localization of Krt19+ and aSMA+ territories in DAPT treated samples. Strikingly, the inhibition of Notch cleavage led to an arrest of LG tree formation (Figure 13A). We could not detect any Krt19+ cells in the branches, and the lumen did not form. Moreover, the aSMA domain expanded drastically, and was not anymore restricted to the TEBs. After inhibiting the Notch pathway for 5 days, aSMA was detected ectopically in the whole basal cell layer of the branches (Figure 13B). The increase of aSMA expression was confirmed by qPCR (Figure S9).

Our results suggested a crucial role of Notch pathway in enforcing the border between duct and TEB domain. Moreover, Notch activity seemed to be directly involved in the branch cell fate.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.