mRNA analysis of intracytoplasmically-stained, FACS-purified pancreatic islet cells using the quantitative nuclease protection assay

Tissues are complex cellular aggregates, whose functional properties are dictated by the unique set of genes they express. Islets of Langerhans, for instance, the glucose-controlling mini-organs dispersed throughout the pancreas, harbor phenotypically different cells of distinct endocrine, neuronal, mesenchymal, and hematopoetic lineages¹. Pancreatic islets regulate carbohydrate metabolism by secreting insulin (from β-cells) and glucagon (from α-cells), while other islet cell subsets provide vascularization, innervation, and structural support. Islets can be isolated for transplantation or to examine their functional properties ex vivo or in vitro², but dissecting islets into functional cell subsets for analytical purposes has been difficult³. Since surrogate endocrine cell surface markers have not been identified, hormone-secreting islet cells are defined and characterized by their cytoplasmic hormone content. Using flow cytometry (FCM), we have recently dissected islet cell suspensions into endocrine lineage subsets by employing a multiparameter staining approach involving intracytoplasmic hormone-staining using paraformaldehyde (PFA)-fixed, membrane-permeabilized islet cells^4,5.

To date quantitative information of individual islet cell subsets has been vague, presumed from whole islet cell preparations, or estimated from protein expression patterns or fluorescence in situ hybridizations⁶. Others have obtained intact full-length RNA required for conventional gene expression analysis by Northern blot analysis or RT-PCR from FACS-purified, genetically-marked, viable islet cells^7,8. Whereas paraformaldehyde (PFA)-fixation and membrane-permeabilization, necessary for intracellular staining, prevents conventional full-length RNA isolation due to chemical cross-linking of proteins and nucleic acids⁹, such paraformaldehyde-induced cross-linking does not interfere with short cDNA-probe RNA hybridizations used by nuclease protection analysis, an established technique for quantifying specific cellular mRNA transcripts^10,11. We therefore reasoned that the technique may be applicable to fixed cells, and sought to combine our ability to purify the multiple islet cell subsets by intracytoplasmic flow cytometry⁴ with gene expression analysis using a nuclease protection technique, optimized to meet high-throughput formats with minimal cell input¹². Fig. 1a illustrates the qNPA^™ detection strategy. Multiple RNA sequence-protecting cDNA probes are captured to defined locations on a microplate support, then quantified by measuring chemiluminescence. Customized arrays using complementary cDNA-probes targeting gene-specific 50-mer RNA-motifs were designed for mouse islet cell studies (Supplementary Tables 1 and 2). Fig. 1b exhibits a charge-coupled device (CCD) camera-captured luminescence profile of purified, non-dissociated islets (top), and the multiwell-plate localization map of individual genes (bottom).

Using mouse islet cell suspensions (Supplementary Fig. 1a), and a mouse insulinoma cell line (Min6, Supplementary Fig. 1b) we tested the effect of paraformaldehyde (PFA) fixation on mRNA-detection. Split-sample aliquots of fresh and PFA-fixed/permeabilized aliquots were analyzed in parallel. qNPA^™ -detected gene-specific mRNA levels were indistinguishable between fresh and PFA-fixed/permeabilized samples, with no discernible impact on transcript abundance. Next we examined if mRNA transcripts are lost if such PFA-fixed cells are saponin-permeabilized and subjected to staining procedures in the presence of detergents, and then high-speed FACS sorting. Fig. 1c demonstrates that little or no loss of cellular mRNA occurred, while a tight regression (r²=0.97) indicates that all gene-specific transcripts were affected similarly. Similar results were also obtained with Min6 insulinoma cells (Supplementary Fig. 2). Finally, graded dilutions of sample input confirmed that qNPA^™ is quantitative over a gene-transcript range of approximately 3 orders of magnitude (Fig. 1d). Excellent quantitative dilution curves were also obtained from tissues containing lower islet cell gene transcripts (i.e by manually depleting the visible islets from viable pancreatic digests), Min6 cells, or purified whole islet RNA (Supplementary Fig. 3a–c).

Having established that qNPA^™ is compatible with PFA-fixation and saponin-permeabilization and that it is quantitative over a broad dynamic range, we tested the technique using FACS-purified major endocrine islet cell subsets from healthy C57BL/6 mice. We applied combination gates throughout to define single-cell suspensions of the three major endocrine cell subsets β-, α-, and δ-cells, respectively secreting insulin, glucagon, and somatostatin. The gates consisted sequentially of forward (FSC)/sideward scatter (SSC), doublet exclusion¹³, and the separation of all three populations from each other and from non-specifically stained debris introduced by the ribonuclease (RNase)-inhibitor vanadyl ribonucleoside complexes (VRC). VRC was present throughout all staining steps to prevent RNA degradation. Optically, VRC caused the appearance of primarily very small, multi-fluorescent ”particles” (Fig. 2a, grey-colored population in dot plots), which when purified by FACS and analyzed by qNPA^™ did not contain gene-specific mRNA, resembling apoptotic cells containing only degraded RNA (data not shown). Fig. 2a illustrates the purity of FACS-prepared endocrine subsets, as demonstrated by re-analysis of small aliquots of the sorted cells (middle panel). Supplementary Table 3 displays the endocrine cell frequencies of islet cell suspensions before and after FACS. Purified cells were then submitted to gene expression analysis by qNPA^™. CCD-camera images exhibit the expected and quite distinct gene-expression profile among the cell populations (Fig. 2a, compare to unseparated islet cells in Fig. 1b). These results were then quantified, normalized to chemiluminescence per 1,000 cells, and displayed in Fig 2a, right panel. As expected, all endocrine subsets expressed those genes whose protein product was used for sorting, and the rigorous purification regimen resulted in negligible cross-contaminating gene expression. More importantly, we note that the three major endocrine subsets exhibit different steady-state hormone-encoding mRNA transcript levels; that is, normal adult mouse α- and β-cells produced on average substantially more hormone transcripts than δ-cells (Fig. 2a, right panel, top 2 vs. 3rd chart). Interestingly, while pancreatic polypeptide (Ppy) transcripts were found as expected in the unstained islet cell population (Fig. 2a, lowest chart), we also observed that δ-cells appeared to express Ppy mRNA. This suggests that some δ-cells may co-express Ppy and somatostatin (Sst), a possibility that to our knowledge has not been adequately addressed in the literature. An alternative explanation is that Sst(+) Ppy(+) doublets form unusually strong heterologous interactions that resist dissociation, and are small enough to escape our forward light scatter-based doublet exclusion strategy.

Our qNPA^™ panels included quantification of four ubiquitously-expressed so-called “housekeeping” genes (Ppia, Actb, Hprt1, and Rsp29, Supplementary Tables 1 and 2), used to normalize and compare gene-transcription between tissues^14,15. We were surprised to find that none of them were equivalently expressed among the three major endocrine islet cell subsets or non-endocrine islet cells (Fig. 2b), consistent with general concerns about using “housekeeping” genes to normalize gene-expression between different tissues¹⁶. Normalization to any one or all four tested housekeeping genes would have distorted comparisons between phenotypically (Fig. 2b) or developmentally distinct islet cell populations (Supplementary Table 4), or from diseased or genetically-mutated mice. We elected therefore to normalize gene-expression to the precise number of cells dispensed and recorded by the cell sorter’s software. Table 1 shows the gene-expression pattern characterizing the four sorted adult islet cell subsets: β-, α-, δ-cells, plus all remaining cells, termed non-endocrine cells (3-neg). The results confirm several previous proposals on the distribution of islet subset-enriched gene products, based largely on multicolor immunofluorescence, gene mutational analysis, or transformed cell lines¹⁷. For instance, our data supports studies demonstrating β-cell expression of both prohormone convertases 1/3 (Pcsk1) and 2 (Pcsk2), whereas α-cells express only Pcsk2^18,19. For most genes tested, clear lineage segregation was apparent in healthy adult islets (Table 1) supporting the current consensus of islet biology and function, as discussed in more detail in recent reviews^20,21. Specifically, β-cells expressed Nkx2-2, Nkx6-1, Mafa, Pdx-1, and Glp1r, whereas Mafb mRNA expression was detected only in α-cells²², and interestingly also in the non-endocrine compartment. In agreement with published reports^23,24 Neurod1 and the zinc transporter Slc30a8 were detected in both α- and β-cells. On the other hand, we found Pax6 expression in adult β-cells, somewhat less in δ-cells, and none in adult α-cells. The latter observations differs from the prevailing view of endocrine cell lineage patterns where Pax6 is thought to be a panendocrine marker²⁵.

Satisfied that the technique accurately measures the expected transcript levels in isolated cell subsets, we further examined gene expression in β- and α-cells at times of physiologically-elevated generative or functional activity, such as embryonic development, adolescence, and metabolic adjustments during pregnancy. These studies revealed a profoundly altered picture (Table 1, cursive font) with several notable observations. For example, relative to adult β-cells, embryonic ins(+) β-cells express approximately 10 fold less insulin mRNA, similarly reduced islet amyloid polypeptide (Iapp) and diminished Pax6, Mafa, Ccnd2, and Glp1r mRNA (Table 1, light grey cells), consistent with their functional immaturity in E15.5–E18.5 embryonic mice. Unlike β-cell insulin transcripts, α-cell glucagon mRNA at the same gestational age did not differ much from adult α-cells. Also, the expression of Gcg and Mafb was substantially higher in α-cells during phases of increased metabolic demand (pregnancy, adolescence) relative to normal adult mouse α-cells (Table 1, dark grey cells). Such profoundly enhanced function was not observed in β-cells of the same mice. Finally, the absolute lineage segregation observed with regard to several key transcription factors in normal adult islet cells (Nkx2-2, Nkx6-1, Pax6), was lost as reflected by the detectable to substantial transcript levels found in embryonic hormone-secreting cells, or during pregnancy (Table 1, dark grey cells). These data suggest that α-cells can turn on some (but not all) typical β-cell lineage genes under certain conditions or at different developmental stages. Our findings are consistent with a greater than expected plasticity in islet cell gene expression patterns during physiological challenges in adult mice.

We wish to draw particular attention to Mafa and Mafb, members of the basic leucine zipper v-maf transcription factor family, which are involved in the development and/or function of various organ systems²⁶. In the pancreas, Mafa facilitates insulin gene expression, and islet β-cell function^27,28, but is not involved in embryonic β-cell differentiation prior to the second transition phase²⁸. Unlike Mafa, Mafb-expression has been associated with islet α- and β-lineage decisions during embryonic development^22,29, and is important for glucagon gene expression in α-cells²⁹. Specifically, we and others have shown that embryonic, immature β-cells temporarily express Mafb prior to the second transition^22,29, suggesting it plays a role in embryonic β-cell neogenesis, differentiation, and/or maturation. While we could not detect Mafb mRNA in normal adult islet β-cells, qNPA^™ revealed Mafb mRNA in purified β-cells isolated from pregnant mice, and to a somewhat lesser extent in adolescent mice (Fig. 3a). We next histologically examined pancreas tissue sections to test whether MafB protein was expressed and to assess cellular heterogeneity of β-cell Mafb expression. As shown in Fig. 3b, pancreata from pregnant mice revealed a small but distinct population of Ins(+)Mafb(+) double stained cells amidst the majority of Ins(+)Mafb(−) β-cells. Since others have demonstrated that pregnant mouse β-cells undergo cell replication during the gestational period³⁰, we tested if β-cell Mafb is preferentially expressed in those cells caught in cell cycle progression. We administered BrdU for 6h to pregnant mice on gestational day 15.5, then examined the pancreata by immunohistology (Fig. 3c). We found the expected small subset of pancreatic Ins(+) β-cells with nuclear BrdU indicating recent cell division, but we found no Ins(+)BrdU(+) β-cells that co-stained for Mafb protein. These results are consistent with the possibility that during pregnancy the increased metabolic demand is met by both enhanced β-cell proliferation and facilitated β-cell differentiation.

We have recently argued that multicolor flow cytometry represents a powerful analytical tool to study islet cell biology^4,5, but suffered, like immunofluorescence microscopy, from the inability to study islet cell-subset gene-expression profiles. Combining FACS-sorting with qNPA^™ analysis overcomes this limitation and is ideally suited to profile endocrine or non-endocrine islet cell populations in health and disease. We demonstrate for the first time that multiple different islet cell subsets can be purified simultaneously, allowing each subsets’ transcriptome to be analyzed. Moreover, the exceedingly efficient and sensitive qNPA^™ technology can be applied using as few as 10³–10⁴ cells per array of currently 15 genes, an easily achievable cell-yield even for rare cells using standard high-speed cell sorters. Further, we envision many other applications, not limited to islet cell biology, for the technique as it can be applied to cells identified by cytoplasmic or nuclear antigens previously very difficult to characterize since sorting such cells also requires fixation. qNPA^™ of FACS-sorted cells will permit much more specific insight into the regulation of cellular function in tissues, including human specimens, whose viable cell subsets can not be genetically marked nor purified using cell surface markers.