PIC polymers were synthesized through a nickel (II)-catalyzed copolymerization of a tri-ethylene glycol functionalized isocyano-(D)-alanyl-(L)-alanine monomer and the azide-appended monomer 2. Variation of the catalyst to monomer ratio (1:1000 and 1:5000) allowed us to synthesize polymers with different polymer lengths, respectively 1K and 5K. The molar ratio used between monomer M1 and M2 was 99:1. To functionalize the polymer, RGD peptide was reacted with DBCO to obtain the complex DBCO-RGD, which was reacted to M2 monomers present on the PIC polymer backbone. Subsequently, the polymers decorated with cell-adhering peptides were purified. For gel formation, 3 ml optimal SMG medium, consisting of Dulbecco’s modified Eagle’s medium/F12 (DMEM/F12) [Gibco, 11320-074] containing 1% Pen/Strep antibiotics [Gibco, 15140-122], 1% glutamax [ThermoFisher Scientific, 35050038], 1x N2 [Gibco, 17502-048], 20 ng/ml EGF [Sigma, E9644], 20 ng/ml FGF2 [Peprotech, 100-18B], 10 μg/ml insulin [Sigma, I6634], 1 µM dexamethasone [Sigma, d4902], 10 µM Y-27632 [Abcam, ab120129], 50 ng/ml noggin [Peprotech, 120-10C], 1 µM A8301 [Tocris Bioscience, 2939], and 10% R-spondin1 conditioned medium, was added to the freeze-dried polymers to prepare stock solutions of 5 mg/ml. After soaking on a shaking platform at 4°C for 4 h, a clear solution was obtained. PIC with two different molecular weights, 1 kDa (PIC-1K) and 5 kDa (PIC-5K) and various PIC concentration (0.5, 0.75, 1, 2, and 3 mg/ml), resulting in a higher stiffness and lower porosity with increasing concentrations, were used.

Human biopsies of non-malignant SMG were minced into small pieces using a sterile disposable scalpel followed by two rounds of mechanically dissociation using the gentleMACS dissociator [Milteny, 130-095-937] and enzymatic digestion in 5 ml HBSS [Gibco, 14175129] with 1% HSA [Sanquin, 15522636] buffer containing 125 µl Collagenase NB6/4 [Nordmark, N0002779], 62.5 µl Pulmozyme [KFF/Roche, RVG 16734] and 625 µl CaCl₂ [Pharmacy UMCG, G00115] (per 100 mg tissue). Isolated cells were collected by centrifugation, washed twice with HBSS/1% HSA buffer filtered through a 100 μm cell strainer [Greiner, 542000] and again collected by centrifugation. Pelleted cells were resuspended in Cryostor^® CS10 [StemCell Technologies, 07931] with 10 µM Y-27632 at a final concentration of 4–10 million cells per ml, cryopreserved using a Corning^® CoolCell [Corning, 432000] which provides freezing at the rate of −1°C/min and stored at −140°C (Figure 1A).

Cryopreserved cells were removed from the freezer, quickly thawed until a small clump of ice was left, washed twice with PBS/1% BSA and resuspended in DMEM/F12 containing 1% Pen/Strep antibiotics and 1% glutamax. 25 µl of cell suspension containing 400,000 cells, was mixed with 50 µl Matrigel^® [VWR, 35623] and seeded as a 75 µl drop in the middle of a well of a 12-well plate. Thirty minutes after seeding, culture medium was added. Culture media used in this study are described as follow: WRYTN, DMEM/F12 containing 1% Pen/Strep antibiotics, 1% glutamax, 1x N2, 20 ng/ml EGF, 20 ng/ml FGF2, 10 μg/ml insulin, 1 µM dexamethasone, 10 µM Y-27632, 50 ng/ml noggin, 1 μM A8301, 10% R-spondin1 conditioned medium and 50% Wnt3a conditioned medium, RYTN, DMEM/F12 containing 1% Pen/Strep antibiotics, 1% glutamax, 1x N2, 20 ng/ml EGF, 20 ng/ml FGF2, 10 μg/ml insulin, 1 µM dexamethasone, 10 µM Y-27632, 50 ng/ml noggin, 1 μM A8301, and 10% R-spondin1 conditioned medium. Where indicated, isoproterenol was added to the medium at a concentration of 1 µM. One week after seeding, salivary gland organoids (SGOs) were released from Matrigel^® by adding 1 mg/ml dispase [Gibco, 17105-041] for 1 h at 37°C, processed to single cells using 0.05% trypsin-EDTA [Invitrogen, 25300-096] and re-seeded in a 50 µl Matrigel^® or PIC hydrogel drop (15,000 cells per gel) to start a new passage. SGOs (14 days) were removed from PIC hydrogels by simply adding cold PBS (Figure 1A). At the end of each passage organoid and cell number were noted and used to calculate Organoid Forming Efficiency (OFE%) and Population Doubling (PD) as follows: O F E % = N u m b e r   o f   o r g a n o i d s   h a r v e s t e d   a t   t h e   e n d   o f   t h e   p a s s a g e N u m b e r   o f   s i n g l e   c e l l s   s e e d e d   a t   t h e   b e g i n n i n g   o f   t h e   p a s s a g e × 100 P D = ln h a r v e s t e d   c e l l s / s e e d e d   c e l l s ln ⁡ 2

Organoids recovered from Matrigel^® and PIC were quantified by counting to prepare a suspension of 30–40 organoids in 25 µl medium and 50 µl Matrigel^®, and seeded as 75 µl droplets in pre-coated wells (40 µl 1:1 diluted Matrigel^®). After solidification, 150 µl differentiation medium (DM) was added. DM was based on DMEM/F12 containing 1% Pen/Strep antibiotics and 1% glutamax, 10% FCS and supplemented with 1x N2, 20 ng/ml EGF, 20 ng/ml FGF2, 10 μg/ml insulin, 1 µM dexamethasone, 100 ng/ml FGF10 [Peprotech, 100-26], 50 ng/ml HGF [Peprotech, 100-39], 1 µm DAPT [Sigma, D5942], 200 nM Carbochol [Sigma, C4382] and 100 ng/ml Heparin sodium salt 0.2% [StemCell Technologies, 7980]. Medium was refreshed every 3–4 days for 3 weeks.

RNA was isolated from SGOs cultured in Matrigel^® (7 days) or PIC (14 days) using RNeasy Mini Kit [Qiagen, 74104] according to manufacturer’s instructions including DNase I [Qiagen, 79254] treatment. cDNA was synthesized from 500 ng mRNA using M-MLV Reverse Transcriptase [ThermoFisher Scientific, 28025013]. Quantitative PCR was performed in triplicate with the indicated primers (listed in Table 1), Bio-Rad iQ SYBR Green Supermix [Bio-Rad, 170880], and a Bio-Rad CFX Connect Real-Time PCR Detection System. Absolute data were first normalized to YWHAZ and subsequently to the Matrigel^® samples.

Human SGOs cultured in PIC or Matrigel^® were collected 7 and 14 days after passaging (at passage 2). RNA was isolated from the collected organoids and also from non-malignant submandibular gland tissues derived from three different human donors using the RNeasy Mini Kit according to manufacturer’s instructions including DNase I treatment. The quality and concentration of isolated RNA was determined with the Agilent TapeStation system. All samples were of high quality with an RNA integrity value >7.90 ng RNA was used as input from each sample for the QuantSeq 3′ mRNA-Seq Library Prep Kit [Lexogen, 015.96]. Briefly, RNA was incubated with oligodT primers containing the Illumina-specific Read 2 linker sequence to reverse transcribe RNA into cDNA. After RNA removal, second strand synthesis was performed using UMI-tagged random primers. Magnetic beads were used for size selection and to purify the double-stranded libraries. The libraries were amplified to add the complete adapter sequences. Libraries were equimolarly pooled and 1.8 p.m. of the pool with 15% PhiX were loaded on a NextSeq 500 [Illumina] for a 75 bp paired-end sequencing run.

All bioinformatic analyses were performed with available packages in RStudio (v4.0.2) and plots were generated with ggplot2 (v3.3.2) (Wickham, 2016). For PCA analysis logCPM counts (Supplementary Table S1) were used with the prcomp function of the stats packages (v4.1.2). The heatmap was created with pheatmap (v1.0.12) (Kolde, 2019), and row clustering was performed with the complete clustering method. For the differential gene expression analysis, lowly expressed genes (total count <1 in less than 2 samples) were excluded. The bioconductor package edgeR (v3.32.0) (Robinson et al., 2010)was used for normalization by using trimmed mean of M-values (TMM) and identification of differential expressed genes (logFC </>4, FDR<0.001) by fitting a generalized linear model. Gene ontology (GO) analysis for differentially expressed genes was performed with gProfiler with default settings (Raudvere et al., 2019). STRING analysis was performed with default settings (string-db.org) (Jensen et al., 2009).

To identify the enrichment of previously identified gene modules (unpublished data) in PIC- and Matrigel^® cultured SGOs, CPM values (Supplementary Table S1) of the present dataset of genes belonging to one gene module (Supplementary Table S2) were averaged per experimental group and presented as column row scores in a heatmap created with gplots (v3.1.1.) (Warnes et al., 2020).

All statistical analyses in this study were performed using GraphPad Prism8 software (GraphPad, La Jolla, CA, United States). The number of patients analysed (n), presented error bars (SEM), statistical analysis and p values are all stated in each figure legend.

Our previously reported 3D organoid culture system suffers from several shortcomings in its translation to the clinic for example by our optimized WRYTN medium. Hence, we previously developed a medium with a well-defined composition after risk assessment analysis had revealed that Wnt3a conditioned media, RSPO1 conditioned media, and Noggin did not meet the requirements for therapeutic use (unpublished data). In this new composition, Wnt3a is omitted as it is not yet available as a GMP-grade component. RSPO1, on the other hand, was replaced with human R-spondin-1 protein resulting in the so called GMP-grade expansion medium RYT, which performs similar to the “research” medium RYTN (unpublished data). Indeed, organoids in Matrigel showed a similar morphology and had the same size after omission of Wnt3a in the medium. Moreover, OFE was not affected. Interestingly, in PIC-RGD5K gels, the omission of Wnt3a in the medium led to an enlargement of the organoids while maintaining a round morphology and significantly improved the OFE (p = 0.0039) (Figures 1B, C). Based on these results further experiments were performed with RYTN medium.

Furthermore, a requirement for potential cell therapy is a 3D organoid culture system that allows expansion of stem cells using a GMP-compliant hydrogel. We assessed SGOs formation from Matrigel-thawed single human SG cells seeded in various PIC gels with RYTN medium. After 14 days in culture, we observed marginal organoid formation, with little to no proliferation when single cells were seeded in plain PIC resulting in an OFE below 50%, compared to Matrigel (Figures 1D, F). To stimulate organoid formation, we tested PIC hydrogels that were functionalized with RGD peptides. The incorporation of RGD motifs was not sufficient to increase organoid formation in the PIC-1K hydrogel, but it resulted in an average increase of 18% for PIC-5K hydrogels (Figures 1D, F). Although, the relative OFE in PIC-RGD5K was still lower compared to Matrigel after 7 days incubation, PIC-RGD5K consisting of 1.0 mg/ml PIC was identified to be most optimal for human SGOs (Figures 1E, F).

Next, we assessed the potential of PIC hydrogels to support long-term expansion of SG stem cells. Again, isolated single SG cells were thawed in Matrigel, recovered and embedded in PIC hydrogels at the same condition (1.0 mg/ml PIC-RGD5K). After 14 days, SGOs were harvested by simply adding cold PBS to break the gels. Subsequently, SGOs were enzymatically dispersed into single cells and reseeded into PIC-RGD5K. Repeated passaging in PIC-RGD5K resulted in SGOs formation which could be repeated up to 4 passages (Figure 1G).

Another characteristic of human SGOs is that they should be able to form all different cell types of the tissue of origin. Previously it has been shown that human SGOs can be differentiated toward mature SG cell lineages when cultured in differentiation medium, void of inducers of proliferation. The potential of human SGOs cultured in PIC to generate functionally mature SG cell lineages was investigated. Light microscopy demonstrated that PIC-derived organoids preserved a more spherical shape and increased in size while Matrigel-derived organoids developed structures that contained both branching and lobular structures indicative of differentiation into SG structures (Figure 2). This data suggests that the differentiation potential in PIC-derived organoids is affected.

To identify genes and associated pathways and functions that are differently regulated in human SGOs in PIC hydrogel and Matrigel, we performed bulk RNA-sequencing on fresh salivary gland tissue and SGOs cultured in Matrigel and PIC hydrogel. Principal component analysis (PCA) identifies a clear segregation of the tissue opposed to the organoids on PC1, identifying the difference of in vivo and in vitro conditions as the biggest driver of gene expression differences in this dataset. The second factor introducing variance in this dataset seems to be the difference in the culturing conditions, since PIC- and Matrigel-cultured organoids clearly separate on PC2 (Figure 3A).

1414 differentially expressed genes are identified between tissue and organoids (PIC and Matrigel together, Supplementary Table S3), confirming the difference between tissue and SGOs as seen in PC1 (Figure 3A). Clustering of these genes yields 3 clusters specific for tissue (cluster 2), PIC (cluster 1)- and Matrigel (cluster 3)-cultured SGOs (Supplementary Figure S1; Supplementary Table S4). Coherent with findings of the PCA analysis, clustering analysis indicates the expression of specific gene sets in PIC- and Matrigel-cultured SGOs.

When directly comparing the gene expression profiles of the two culturing conditions, 22 genes were depleted, and 188 genes were enriched in PIC-cultured SGOs compared to Matrigel-cultured SGOs (Figure 3B; Supplementary Table S5). To identify factors that are missing in PIC organoids compared to Matrigel organoids, we focused in the further analysis on the 22 genes found to be in PIC. Genes depleted in PIC-cultured SGOs were associated with GO-terms such as “response to oxygen-containing compound/lipid/hormone”, “skeletal muscle development”, “neurotrophic receptor tyrosine kinase 1 signalling,” whereas genes enriched in PIC were involved in amongst others “plasma membrane,” “extracellular space” and “immune response” (Supplementary Table S6).

To specify the biological processes in which the genes depleted in PIC-cultured organoids could be involved in, the present dataset was compared to a yet unpublished RNA-sequencing dataset of human salivary gland organoids. In the unpublished study, weighted gene co-expression analysis identified gene sets (modules) that are associated with different cellular states. Gene modules “Sienna4,” “Magenta3” and “IndianRed3” (Supplementary Table S2) positively correlated to a more primitive stem cell-like trait, whereas modules “Black” and “DarkOrange2” (Supplementary Table S2) positively correlated to a more differentiated trait. When compared to Matrigel-cultured SGOs, PIC-cultured SGOs were especially enriched in genes associated with gene modules correlated to the more differentiated cell state (Figure 3C). Indeed, expression of differentiation markers Aquaporin 5 (AQP5), α-Amylase (AMY), Mucin 7 (MUC7), and MIST1 was increased while the stem cell marker SOX2 and ductal markers KRT5 and KRT19 were decreased in PIC-cultured SGOs compared to Matrigel-cultured SGOs, as identified by qPCR (Figure 3D).

The genes that were found to be downregulated in PIC- compared to Matrigel-cultured SGOs mainly belong to the immediate early gene (IEG) family, such as FOS, EGR, and NR4A1 (Figure 3E; Supplementary Table S1) and these genes also seem to form a highly interacting network (Figure 3F).

In contrast to our results described in “Organoids in PIC cannot be differentiated into mature SG cell lineages” and as depicted in Figure 2, our RNA-sequence data suggests premature differentiation of PIC-derived organoids, possibly due to the lack of IEG expression.

To optimize a 3D organoid culture system that allows the production of a GMP-compliant cellular product, and due to the low OFE as a result of premature differentiation of SG stem cells, we aimed to further enrich the stem cell population in the organoid culture system. Therefore, we tested the addition of both the GSK3 inhibitor CHIR99021 (3 µM) and the histone deacetylase valproic acid (0.05 mM and 1 mM; VA) to the RYTN culture condition, a combination that previously improved the efficiency of organoid formation by stimulating Wnt and Notch signaling resulting in more Lgr5+ stem cells (Esfandiari et al., 2012; Yin et al., 2014; Pachenari et al., 2017). Unfortunately, we found that treating single cell seeded hSGSCs with CV did not enhance the OFE in both Matrigel and PIC (Supplementary Table S1).

To validate the importance of the expression of IEGs for the capacity of SG stem cells to form SGOs, PIC-cultured SGOs were exposed in vitro to the β-adrenergic agonist isoproterenol, known to activate multiple signalling pathways involved in IEG expression (Yeh et al., 2012; Bahrami and Drabløs, 2016). Indeed, SGOs cultures exposed to isoproterenol demonstrated a significant increase in OFE compared with untreated cultures. A maximum increase of 5% was observed following 1 µM isoproterenol exposure, compared with untreated control cultures without isoproterenol (p = 0.0156; Figures 4A, B). Higher concentrations of isoproterenol (10 µM) did not show an effect on OFE while concentrations of 100 µM isoproterenol appeared to be toxic (data not shown).