To the Editor:

Sjögren disease (SjD) is widely recognized by the presence of B cell–dominated lymphocytic infiltration in the salivary glands (SGs). Contrary to what was originally presumed, however, SG hypofunction in SjD is not strongly correlated with the degree of SG lymphocytic infiltration in the SGs. The ductal epithelium has been implicated in SG pathogenesis in SjD, following demonstration of the ductal cell ability to express Toll-like receptors and receptors for interferon signaling to produce interferon themselves, to express SjD-associated autoantigens SSA/Ro60, SSA/Ro52, and SSB/La following stimulation, and to express proinflammatory cytokines including interleukin (IL)-1, IL-6, IL-7, IL-18, tumor necrosis factor (TNF), B cell–activating factor (BAFF), CXC motif chemokine 10 (CXCL10), CXCL12, and CXCL13 (reviewed in Verstappen et al1). Many of these critical pieces of work probing the involvement of the SG epithelium in SjD pathology employ SG epithelial cell (SGEC) cultures. SGEC cultures are derived using the explant culture technique, whereby a small piece of SG tissue is plated into a flask, and the outgrowing cell is presumed to represent epithelial cells.2

Organoids are defined as self-organizing 3-D structures, derived from the proliferation and differentiation of organ-based stem and progenitor cells.3 Organoids recapitulate the spatial organization of their organ of interest, and have been optimized from the murine and human submandibular and parotid SGs.4-8 Organoids have been employed to understand disease pathology, design patient-specific drug screening platforms, and predict disease prognosis, and as such are an invaluable tool for the SjD research community.

Institutional review board approval was granted by the Medical Ethical Testing committee (METc) for the use of healthy SG patient material (project no. METc 2008.078) and material from patients with SjD (METc2016.010).

We leveraged deconvolution analysis to directly compare predicted cell types present in SG organoid (SGO) and SGEC culture derived from matched donors (Figures 1A,B), with a view to understanding if significant differences between the model systems exist, and to facilitating a choice of platform for future studies.

SGEC cultures express pathways associated with the extracellular matrix, and more fibroblast-associated genes, than matched SGO cultures. (A,B) Representative phase-contrast microscopy of SGEC and SGO cultures. Cultures were harvested simultaneously to avoid any effects of culture artefacts. Scale bar applies to both microscopy panels and represents 100 μM. (C) Heat map analysis comparing bulk RNA expression in SGECs to SGO cultures. Red indicates higher gene-level expression and green lower, compared to the mean. (D,E) Top 15 enriched biological pathway processes in each cluster in panel B. Box color relates to cluster color in panel B. (F) Expression levels of acinar, myoepithelial, and ductal cell genes in SGEC and SGO cultures, normalized to housekeeping genes. (G) Expression levels of SG progenitor cell–associated genes in SGEC and SGO cultures, normalized to housekeeping genes. (H) Expression levels of fibroblast-associated genes in SGEC and SGO cultures, normalized to housekeeping genes. In (F-H), n = 4 for each group. Bars represent SEM. * P < 0.05. ** P < 0.01. AdjPval: adjusted P value; Myo: myoepithelial; SEM: standard error of the mean; SG: salivary gland; SGEC: salivary gland epithelial cell; SGO: salivary gland organoid.

Bulk RNA sequencing (RNA-Seq) analysis of 4 healthy SGEC cultures and matched SGO cultures from the same biopsies was performed, resulting in detection of expression of 57,965 genes in total, with approximately 12 million total read counts per sample. Of these 57,965 genes, 18,130 passed the filter for a minimum of 0.5 counts per million in at least 1 library. There were 16,096 genes converted into Ensembl gene IDs. When the top differentially expressed genes between SGOs and SGECs were plotted in heatmap format, 2 distinct clusters could be defined, namely those with genes highly expressed in SGECs (384 genes) and those in SGOs (644 genes; Figure 1C). Pathway analysis demonstrated that extracellular matrix organization pathways were highly expressed in SGEC cultures, and epithelial development pathways in SGOs (Figures 1D,E). No striking difference in SG ductal, acinar, myoepithelial, SG stem cell–associated, or acinar progenitor cell–associated marker gene expression between SGECs and SGOs was observed (Figures 1F,G). In contrast, 8/23 mesenchymal marker genes, including CD90, COL15A1, COL1A2, DCN, and PDGFRA, were significantly higher expressed in SGEC cultures (Figure 1H). Deconvolution suggested that SGECs and SGOs are heterogeneous cultures, with both containing basal, striated, Krt19+, Ascl3+, intercalated, and Gstt1+ male and female and granulated convoluted tubule ductal cells (Figure 2A). SGO cultures contain more total ductal cells (46%) compared to SGECs (39%; Figure 2B). SGECs and SGOs both contain seromucous, mucous, and Smgc+ acinar cells, with SGOs also containing more Bpifa2+ acinar cells (Figure 2A). According to our analysis, mesenchymal cells are likely to comprise 8% of SGEC cultures, compared to only 4% in SGOs (Figure 2C). SGEC cultures are also predicted to contain 36% acinar cell types, suggesting that, in total, 44% of cells in a SGEC culture are not ductal in nature. This value is markedly smaller in SGOs, with 23% acinar cells and 4% mesenchymal cells, for a total of 27% nonductal (Figure 2C).

Deconvolution of bulk RNA-Seq data suggests heterogeneous ductal and acinar cell content of SGOs and SGECs, with more variation in epithelial cell content in SGOs, presence of more mesenchymal cells in SGEC cultures compared to SGOs, and skewed SG progenitor cell dynamics in SjD. (A) Deconvolution of bulk RNA-Seq using the Cibersortx platform. Analyzed samples are named on the x-axis, the relative percentage of each determined cell type on the y-axis, given in the indicated color from the key. (B) Mean percentages of each cell type, represented as pie charts of cell types detected in the 4 matched SGEC and SGO cultures. (C) Combined acinar and ductal cell types, mesenchymal cells, and mitotic cells, as a percentage of total cells. Values represent mean. Key represents percentage scale. (D) Statistical analysis of individual cell types in SGEC vs SGO cultures. Bars represent SD. * P < 0.05. ** P < 0.01. (E) Deconvolution of 6 HC SGO samples, 3 incomplete SjD SGO samples (Inc SjD) and 3 SjD SGO samples (SjD). (F) Mean proportion of epithelial and mesenchymal cell types in incomplete SjD and SjD SGO samples compared to mean content in HC samples. Error bars represent SD. * P < 0.05. ** P < 0.01. *** P < 0.001. #P < 0.05 when comparing Inc SjD to SjD. ###P < 0.001 when comparing Inc SjD to SjD. HC: healthy control; Inc SjD: incomplete Sjögren disease; NK: natural killer; RNA-Seq: RNA sequencing; SG: salivary glad; SjD: Sjögren disease; SGEC: salivary gland epithelial cell; SGO: salivary gland organoid.

We have previously shown that SG dysfunction in SjD may be at least partly due to senescence of SG progenitor cells, suggested to reside in the basal layer of the striated duct.6 Deconvolution analysis of existing SGO bulk RNA-Seq data in the current study suggests a significant decrease in the percentage of basal duct cells contained in SGOs isolated from patients with SjD compared to healthy controls (Figures 2E,F). Deconvolution analysis also suggested a shift in SG progenitor cell dynamics in SjD, with predicted proportions of Gstt1+ female ductal cells, Bpifa2+ proacinar cells, and serous acinar cells significantly decreasing in SjD SGOs compared to healthy SGOs, and Ascl3+ cells increasing (Fig. 2E,F). In our original paper,6 we also analyzed a group of patients with “incomplete SjD,” namely, those with some characteristics of SjD but not fulfilling classification criteria. Significantly more basal ductal cells were observed in the incomplete SjD group compared to the SjD group, as were more Gstt1+ female ductal cells, confirming our previous report6 (Figure 2F). All 3 patients with incomplete SjD and 3 patients with SjD had biopsies analyzed and did not have a positive focus score, which is a proxy of the degree of lymphocytic infiltration; this finding confirms previous studies suggesting that loss of SG function, or in this case, change in cell composition, in SjD is not likely to be linked with lymphocytic infiltration.9,10

To summarize, we advise caution in the interpretation of data derived from SGECs, considering that any effects observed are more likely to be attributable to the immunocomplicity of mesenchymal cells present in SGEC cultures, compared to those from SGO cultures. We also show that SGOs contain a more diverse repertoire of ductal cell types and fewer acinar cell types than SGECs, a finding with potential implications for future studies. In terms of choosing the most appropriate model for new studies, we suggest that SGOs represent a platform most suitable for studying the interaction between epithelial cells, immune cells, and neuronal cells, among others, owing to their ability to faithfully recapitulate the spatial organization of the SG. SGECs, due to their ease of generation and costly nature, may potentially lend themselves more in the future to pilot studies, prior to development in a 3-D context.

All data are freely available on request.