Insulin secretion deficits in a Prader-Willi syndrome β-cell model are associated with a concerted downregulation of multiple endoplasmic reticulum chaperones

Prader-Willi syndrome (PWS) is a multisystem disorder with neurobehavioral, metabolic, and hormonal phenotypes, caused by loss of expression of a paternally-expressed imprinted gene cluster. Prior evidence from a PWS mouse model identified abnormal pancreatic islet development with retention of aged insulin and deficient insulin secretion. To determine the collective roles of PWS genes in β-cell biology, we used genome-editing to generate isogenic, clonal INS-1 insulinoma lines having 3.16 Mb deletions of the silent, maternal- (control) and active, paternal-allele (PWS). PWS β-cells demonstrated a significant cell autonomous reduction in basal and glucose-stimulated insulin secretion. Further, proteomic analyses revealed reduced levels of cellular and secreted hormones, including all insulin peptides and amylin, concomitant with reduction of at least ten endoplasmic reticulum (ER) chaperones, including GRP78 and GRP94. Critically, differentially expressed genes identified by whole transcriptome studies included reductions in levels of mRNAs encoding these secreted peptides and the group of ER chaperones. In contrast to the dosage compensation previously seen for ER chaperones in Grp78 or Grp94 gene knockouts or knockdown, compensation is precluded by the stress-independent deficiency of ER chaperones in PWS β-cells. Consistent with reduced ER chaperones levels, PWS INS-1 β-cells are more sensitive to ER stress, leading to earlier activation of all three arms of the unfolded protein response. Combined, the findings suggest that a chronic shortage of ER chaperones in PWS β-cells leads to a deficiency of protein folding and/or delay in ER transit of insulin and other cargo. In summary, our results illuminate the pathophysiological basis of pancreatic β-cell hormone deficits in PWS, with evolutionary implications for the multigenic PWS-domain, and indicate that PWS-imprinted genes coordinate concerted regulation of ER chaperone biosynthesis and β-cell secretory pathway function.

Omnibus (GEO) under SuperSeries GSE190337 including GSE190334 (total RNA-seq) and GSE190336 (small RNA-seq), assigned to BioProject PRJNA786769. The mass spectrometry (MS) proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifiers PXD034471 (secreted peptides), PXD037874 (quantitative TMT MS; insoluble cellular proteins), and PXD037875 (soluble cellular proteins). A full set of bash scripts for analysis of both total and small RNA-seq data sets including custom annotations are provided at the github repository (https://github.com/KoppesEA/INS-1_PWS_RNA-Seq). All other numerical data and statistical analyses for graphical results reported in this manuscript are presented in S1 Data.
Our earlier studies of a transgenic-PWS (TgPWS) mouse model harboring a deletion of the orthologous PWS-imprinted domain demonstrated severe failure to thrive, abnormalities in fetal pancreatic islet development and architecture, reduced α-and β-cell mass, increased apoptosis, and postnatal onset of progressive hypoglycemia that led to lethality within the first postnatal week [16,25]. Plasma insulin and glucagon levels were low during fetal and neonatal life of TgPWS mice [16,25], and, at postnatal day 1 prior to onset of hypoglycemia, there was significantly reduced basal and glucose-stimulated insulin secretion (GSIS) from cultured TgPWS islets [25]. Furthermore, using an insulin-Timer fluorescent protein biomarker to image postnatal day 1 β-cells in vivo, TgPWS mice showed a striking accumulation of aged insulin whereas wildtype control littermates only displayed newly synthesized insulin [25]. These results suggested that PWS-imprinted genes are required for the development and secretory function of pancreatic endocrine cells.
Based on the salient insulin secretion dysfunction in the PWS-mouse model [25], we sought to assess the role of PWS genes in pancreatic β-cell secretory pathway function in a cellular model system. Herein, we used CRISPR/Cas9 genome editing within the INS-1(832/13) insulinoma (β)-cell line [34] to generate deletions of the complete PWS-domain. Following validation of a cell autonomous insulin secretion deficit in the PWS INS-1 model, we performed molecular profiling by proteomic and transcriptomic approaches. The data revealed deficits in multiple secreted hormones as well as endoplasmic reticulum (ER) chaperones that are components of secretory protein folding and trafficking pathways [35][36][37][38][39][40], providing mechanistic insight into PWS-gene function in pancreatic β-cells.

Generation of INS-1 cell lines with silent, maternal allele PWS-gene deletions
Various insulinoma cell lines, predominantly of rodent origin, are widely utilized to investigate β-cell mechanisms as a tractable in vitro model system [41]. To investigate the functions of PWS-genes in β-cells, we used CRISPR/Cas9 genome-editing to target 3. 16 Mb deletions encompassing the PWS-genes (Figs 1A and S1A) in rat INS-1(832/13)::mCherry insulinoma cells (hereafter termed INS-1) that secrete rat and human insulin [34] and express a mouse Ins2 C-peptide-mCherry biosensor in insulin secretory granules [42]. The large deletions were visualized in about 9% of unselected, transfected cells as determined by fluorescence in situ hybridization (FISH) (S1B Fig). INS-1 lines with PWS-domain deletions were derived through sequential targeting and clonal isolation of cell lines initially harboring deletions of the silent maternal allele, followed by targeting of the remaining paternal loci, culminating in the creation of homozygous PWS-deletion lines (Fig 1B).
This process first led to isolation of two lines with a PWS-region deletion (5)(6)(7)(8)(9) identified by deletion-breakpoint PCR (Fig 1C) and DNA sequencing (S2A and S2B Fig), and confirmed by FISH using BAC probes from within and outside the PWS-domain (S1C-S1F Fig).
One PWS allele was deleted in all interphase and metaphase cells of line 5-5 (S1D Fig), indicating clonal isolation of a cell line with a PWS-domain deletion. Initially, line 5-9 was more complex, as FISH showed two cell types with either no deletion (most cells) or a PWS-domain deletion (S1E Fig); additionally, similar proportions of cells lacked mCherry fluorescence, likely due to epigenetic silencing (as DNA analysis indicated the transgene was present) or were mCherry-positive, respectively. This allowed separation by fluorescence-activated cell sorting (FACS), with establishment of a clonal mCherry-positive cell line (5-9) harboring a PWS-domain deletion (S1F Fig). Clonal engineered cell lines 5-5 and 5-9 were inferred to have maternal-deletions of the rat PWS-domain as evidenced by 1) detection of only an unmethylated allele at the PWS-imprinting center (PWS-IC) by bisulfite PCR whereas parental INS-1 and control lines had both an unmethylated and a methylated allele (Fig 1D) indicative of biparental inheritance of the PWS-IC imprint; and 2) expression of PWS-imprinted loci by reverse-transcription-PCR (RT-PCR) (S1G Fig), indicating the presence of an active, paternalallele (S1H Fig). Finally, the presence of mutations at sgRNA target sites on the intact, paternal chromosome, termed scarred mutation alleles, identified a single nucleotide insertion at the sgRNA1 site (S1I and S2C-S2E Figs) but different nucleotide insertions at the sgRNA3 site (S1I and S2F-S2H Figs), the latter indicating an independent origin for the cell lines, 5-5 and 5-9. In contrast, no sequence changes occurred at top-ranked off-target sites for sgRNA1 or sgRNA3 (S3   (3, lack mRNA expression of the PWS-imprinted genes, Snrpn and Snord116, as detected by RT-PCR, whereas control lines (5-9, 2, 16) express PWS-imprinted genes. All cell lines express the control genes (Gapdh, Ins2). The full panel of PWS-imprinted genes is shown in S10A and S10B Fig (H) Use of whole cell extracts for a western blot shows that PWS-deletion lines (3, lack expression of the spliceosomal SmN polypeptide (24.6 kDa) encoded by Snrpn, but retain expression of the paralogous SmB polypeptide (23.6 kDa) encoded by the unlinked Snrpb gene. The control is α-Tubulin. https://doi.org/10.1371/journal.pgen.1010710.g001

Generation of PWS INS-1 cell lines with deletions on the active paternal allele
To specifically target the paternal allele of lines 5-5 and 5-9, we used rat-specific sgRNA sets internal to the first targeting sites to generate slightly smaller 3.143 Mb PWS-deletions (S4A- S4C Fig). The use of FISH demonstrated that about 5% of unselected, transfected cells had a PWS-deletion (S4D Fig). We chose maternal deletion (control) line 5-9 for a second round of transfections and clonal isolation, generating control lines 2 and 16 (Figs 1B, 1E, S4E and S4F) as well as PWS lines 19 and 25, with an expected paternal deletion-PCR product and PWS line 3 with a larger than expected deletion-PCR product (Figs 1B, 1E and S4G). Genomic copy number was confirmed by ddPCR (Figs 1F, S6, and S9), establishing a set of three control lines (5-9, 2, 16) with hemizygosity for maternal-allele deletions (S1H An absence of PWS-gene expression in all PWS lines was shown by RT-PCR for each PWS-imprinted gene (Figs 1G, S10A and S10B) and by western blot analysis for the SmN polypeptide encoded by Snrpn (Fig 1H). Intriguingly, PWS INS-1 lines showed a very low level of apparent expression of the PWS-imprinted gene Snurf by RT-PCR (S10A, S11A, and S11B Figs). Sequencing of the RT-PCR products identified an expressed ψSnurf sequence within a recently evolved ψSnurf-ψSnrpn locus located in an intron of the Mon2 gene (S11B and S11C

PWS INS-1 cell lines show deficits in insulin secretion and ER chaperones
To determine whether PWS INS-1 cell lines have secretory deficits, we carried out insulin secretion assays under low (2.2 mM) and high (22 mM) glucose conditions. The three PWS cell lines displayed a striking deficit in both basal and glucose-induced insulin secretion as compared to the isogenic control INS-1 lines (Fig 2). However, both the PWS and control cell lines had a similar increase from each basal level for GSIS (Fig 2), indicating that PWS β-cells were not deficient in glucose responsiveness.
Next, mass spectrometry was used to assess changes in the cellular proteome and all secreted peptides in PWS vs. control INS-1 cell lines, under insulin secretion conditions of 22 mM glucose, separating cellular proteins by methanol-acetic acid extraction into soluble (mostly small) and insoluble (mostly large) protein fractions with the latter analyzed by quantitative proteomics (Fig 3A). Unexpectedly, in PWS INS-1 cells there was a striking deficiency of multiple ER chaperone proteins, including GRP78/BiP (HSPA5), GRP94/endoplasmin (HSP90B1), PDIA4, HYOU1, CRELD2, and DNAJB11, with lesser reductions in SDF2L1, DNAJC3, PDIA6, and PDIA3, and a modest decrease in PPIB (Fig 3B). Similar deficits of residual amounts of many of these ER chaperones as well as MANF were detected in the soluble protein fraction (Fig 3C). Furthermore, in PWS β-cell lines there was reduced abundance of numerous hormones co-secreted from INS-1 cells, including rat insulin-1 and insulin-2, mouse insulin-2, and human insulin (pre-pro-, pro-, C-peptide, and mature versions of each), as well as reductions in processed and precursor forms of IAPP and NPY, detected in both the cellular small protein fraction (Fig 3C) and residual amounts in the large protein fraction by quantitative proteomics (Fig 3B). In contrast, the full-length secretory granule protein, chromogranin B (CHGB), and the C-terminal CHGB (CCB)-peptide were increased 2-fold (Fig 3B  and 3C). Extending both the insulin secretion data (Fig 2) and the cellular protein deficiencies (Fig 3B and 3C), mass spectrometry analysis of peptides secreted into the culture media demonstrated reduced levels of all forms of insulins and IAPP in PWS β-cell media compared to control (Fig 3D). There was also a reduction in secreted levels of the 14 amino acid WE14 peptide processed from chromogranin A (CHGA), but no changes in the CHGA precursor and Insulin secretory rates are presented as boxplots displaying the interquartile range (IQR) with notches marking the 95% confidence interval of the median and whiskers extending 1.5 x IQR for control (grey) and PWS (red) INS-1 lines at low (2.8 mM) and high (22 mM) glucose, with underlying data points indicated by unique symbols for each cell line (inset). Although GSIS increased 1.75-fold for control and 1.85-fold for PWS, loss of the PWS genes decreased the secretory rate by 36% in basal 2.8 mM glucose conditions and by 32% in stimulatory 22 mM glucose conditions. Statistical comparison (n = 36 biological replicates per group, with n = 12 per cell line) by ANOVA with Tukey's post hoc test (***, P < 0.0001).
https://doi.org/10.1371/journal.pgen.1010710.g002 processed forms in either secreted or cellular fractions. Secreted levels of chromogranin B were increased in PWS (Fig 3D), further illustrating the concordance between cellular and secreted peptide levels in PWS vs. control INS-1 lines. These results indicate that deletion of PWS-genes sharply lowers secretion of insulin and other peptide hormones (IAPP, NPY, CHGA-WE14) and this reduction is associated with deficiencies in many ER chaperones.

Transcriptome studies of control and PWS INS-1 cell lines
Many of the PWS-imprinted genes (Fig 1A) are suggested to function in gene expression [18,19,21,22,43,44]. To gain insight into the molecular basis for insulin secretion and ER chaperone deficits in PWS INS-1 lines, high-throughput total RNA sequencing (RNA-seq) was used to identify differentially expressed genes (DEGs) under standard cell culture conditions (7.5 mM glucose) (Figs 4A, 4B, and S12A and S1 and S2 Tables). Visualization by a heatmap clustergram (S12A Fig) indicates that PWS clonal lines and control clonal lines each grouped together with similar expression profiles. A volcano plot depicting the magnitude, direction, and statistical significance of DEGs in the 3 PWS vs. 3 control lines illustrates no appreciable expression (Log2 Fold Change < -5) of PWS-imprinted genes with remaining DEGs symmetrically orientated around the ordinate axis with a near equivalent number of significantly upregulated (105) and downregulated (123) genes (Fig 4A). The PWS transcripts with complete loss of expression specifically in PWS INS-1 lines include Snurf, Snrpn, Ipw, Mkrn3, all four snoRNAs (Snord116, Snord115, Snord107 and Snord64), miRNA (Mir344 isoforms), and duplicated U1-Snurf sequences (Fig 4A), as also seen by RT-PCR (Figs 1G, S10A, S10B, S11A and S11B). Three PWS-imprinted genes, Ndn, Magel2, and Frat3 (Peg12), were not detected as DEGs as these are not expressed by RT-PCR or RNA-sequencing in any of parental, control, or PWS INS-1 lines. These loci are present by genomic PCR, suggesting an epigenetic inactivation in the INS-1 founder cell line, although silencing is not widespread, as Mkrn3 and Mir-344 are interspersed with the silenced loci (see Fig 1A) and are well-expressed in control INS-1 lines (S10B and S13 Figs).
To ensure complete coverage of the INS-1 transcriptome, RNA-seq of small stable RNAs in the PWS vs. control cell lines was carried out, culminating in the identification of 58 significant differentially expressed miRNAs and snoRNAs (S13A and S13B Fig and S3 Table). Congruent with the RNA-seq data, all but three correspond to PWS-imprinted small RNAs, including duplicated MiR-344, Snord116 and Snord115 loci, as well as single copy Snord107 and Snord64 loci (S13A and S13B Fig Table), including Mir375, Mir148a,Mir183,Mir30d,Mir27b,Mir25,Mir26a2,Mir26a,Mir192,Mir125a, and Mir141 are also in the top 25 expressed miRNAs for a human β-cell model [45]. Within INS-1 cells, expression of PWS-region Snord116 copies make up 13.8% of the top 188 highly expressed snoRNAs and cumulatively would be the thirteenth highest expressed, with Snord64 also in and Snord107 just outside the top 100 (S5 Table), although snoRNAs are poorly studied as a small RNA class.
Several other observations merit note. There was a discrepancy for rat Ins1 and Ins2 mRNA levels between the RNA-seq (Fig 4A) and quantitative RT-ddPCR (Fig 4E) expression profiling, with only the latter showing statistically significant differences between PWS and control lines; despite a trend for reduced Ins1 and Ins2 in the PWS lines observed in the normalized RNA-seq data (S16M Fig). Only a minority of DEGs identified by RNA-seq did not validate by RT-ddPCR, including several encoding neuronal active zone proteins (S12D Fig). These minor discrepancies may be due to differences in cell culture environment or stochastic variability between the biological replicates. Additionally, an apparent increase in expression of the human INS-neoR transgene in PWS lines (S12C Fig) was an artifactual consequence of epigenetic silencing of the transgene, specifically in control line 16 (S16A, S16I-S16K and S16M Fig). Interestingly, although both mRNA (Pcsk1) and protein levels of prohormone convertase PC1 were reduced in iPSC-derived neurons from PWS patients and in whole islets of inbred Snord116-deficient mice [23,60], neither PC1 protein nor Pcsk1 or Pcsk2 mRNA levels were changed in PWS INS-1 cell lines (S12E Fig). Finally, no markers of activation of apoptotic or other cell death pathways were observed by RNA-seq or proteomics, consistent with the absence of any increase in cell death observed in cultured PWS or control INS-1 cells. Indeed, electron microscopy showed normal mitochondria, rough ER, and other identifiable subcellular organelles in PWS and control INS-1 lines (S17 Fig). The difference in observations in vivo where increased apoptosis was observed in a subset of α-and β-cells in TgPWS fetal islets [25] and in vitro (this study) may reflect the use of enriched culture medium with reduced cellular stress under cell culture conditions. In summary, the transcriptome studies show that loss of PWS-gene expression in PWS β-cell lines is accompanied by alterations in levels of mRNAs encoding numerous secreted peptides and ER chaperones.

Confirmation of insulin and ER chaperone protein deficits in PWS INS-1 lines
To further address the predicted disruptions of the ER and secretory pathway, we used western blot analyses to measure cellular levels of insulin and ER chaperones in control and PWS INS-1 lines. Levels of insulin polypeptides detected by an anti-insulin antibody were each significantly decreased in PWS INS-1 lines compared to controls, including diminished amounts of pro-insulin, pre-pro-insulin, and pro-mInsulin-2::mCherry bands (Fig 5A and 5B). Additionally, using an anti-mCherry antibody, we observed that PWS cells have significantly lower levels of the proinsulin form of the mouse insulin2-mCherry transgene but no decrease in the processed C-peptide (CP) form (Fig 5C and 5D). Importantly, use of a KDEL antibody to identify proteins with the ER retention motif confirmed significant deficiencies in levels of the two major ER chaperone proteins, GRP78/BiP and GRP94, in PWS INS-1 cell lines (Fig 5E  and 5F). These results indicate that there are major disruptions of the protein folding machinery and attendant reductions in insulin processing and secretion in the PWS INS-1 cells.

PWS INS-1 cells are more sensitive to activation by ER stressors
Chaperones, such as GRP78 and GRP94, not only have essential roles in protein folding and trafficking in the ER but disruptions of chaperone functions sensitize cells to agents that stress this organelle [61][62][63][64][65][66]. Therefore, we assessed whether the PWS INS-1 lines and their reduced levels of ER chaperones (Figs 3-5) would accentuate activation of the three sensory proteins of the UPR. First, ER stress activates IRE1α to catalyze mRNA processing of the Xbp1 mRNA to generate the functional XBP1 transcription factor [66]; this non-canonical "splicing" of Xbp1 transcripts occurs at low and equal levels in PWS and control INS-1 cells under DMSO control growth conditions but is enhanced by treatment with thapsigargin to initiate ER stress (Figs 6A, 6B, S18A and S18B). Importantly, production of "spliced" Xbp1 mRNA in thapsigargintreated cells occurs significantly more robustly at earlier 2 h and 3 h timepoints for PWS than for control INS-1 lines (Figs 6A, 6B, S18A and S18B Fig) but normalizes by 4-5 h (S18A and S18B Fig). These results reveal that while the initial magnitude of IRE1 activation is greater in the PWS INS-1 cells, the duration and cumulative response is comparable. Second, phosphorylation of eIF2α by PERK [65] was assessed, with PWS INS-1 lines showing significantly higher levels of eIF2α pSer51 phosphorylation than control lines when ER stress was induced by 5 h of thapsigargin treatment (Fig 6C and 6D). Phosphorylated eIF2α inhibits general translation but preferentially translates certain stress adaptive genes, including ATF4 and CHOP; transcripts of these genes were unaffected in unstressed PWS INS-1 cells as measured by RNA-seq.
Third, dissociation of ER-retained ATF6 from GRP78 by ER stressors such as de-glycosylation agents [62,63] enable it to move to the Golgi, where it is processed to an activated nuclear (N) form, ATF6-N, which regulates expression of genes encoding many ER chaperones, including GRP78 and others. In PWS and control INS-1 cells, ER stress induced by tunicamycin, as expected, de-glycosylates full-length glycosylated (FL-G) ATF6 of~102-kD to an unglycosylated 90-kD (FL-UG) form and the processed ATF6-N form of 55-kD (Figs 6E and S19A). The loss of the ATF6 FL-G form by deglycosylation occurs at a similar rate for PWS and ; n = 9 each genotype) and thapsigargin (3 h; n = 6 each) conditions. PWS β-cells are more sensitive to thapsigargin-induced ER stress with earlier and more robust activation of Xbp1 mRNA "splicing" than control cell lines. (C) The PERK pathway was assessed by comparison of PERK phosphorylated eIF2α-P (pSer51) to total eIF2α, using whole cell lysates from PWS and control INS-1 cell lines treated with DMSO or 0.1 μM thapsigargin for 5 h. Anti-αTUB was used as a control for protein-loading levels. At bottom, the eIF2α P/T (phosphorylated/ total) ratios for each cell line are shown under the untreated (DMSO) and ER stress (thapsigargin) conditions. (D) Quantitation of the relative (rel.) level of phosphorylated (pSer51) to total eIF2α in the PWS and control INS-1 lines (n = 3 each) under untreated (DMSO) and thapsigargin conditions. PWS cells are more sensitive to thapsigargin-induced ER stress with more robust activation of eIF2α phosphorylation. (E) For the ATF6 pathway, we used whole cell lysates from PWS and control INS-1 cell lines treated with DMSO or 10 μg/ml tunicamycin for 5 h. Under control DMSO conditions, full-length (FL) and glycosylated (G) ATF6 of~100-kD is ER-localized. Tunicamycin inhibits biosynthesis of N-linked glycans, resulting in unglycosylated (UG) ATF6-FL of 90-kD in the ER, with proteolytic processing in the Golgi to produce nuclear (N) ATF6-N of~55-kD. *, non-specific band detected by the anti-ATF6 antibody. Anti-GAPDH was used as a control for protein-loading levels. At bottom, the ATF6 N/Total (nuclear/Total) ratios for each cell line are shown under untreated (DMSO) and ER stress (tunicamycin) conditions. (F) Quantitation of the ratio of ATF6 nuclear/Total (N/Total) in the PWS and control INS-1 lines under untreated (DMSO) and ER stress (tunicamycin) conditions (n = 6 each, from two biological replicates). PWS cells are more sensitive to tunicamycin-induced ER stress with more robust activation of ATF6-N. For (B,D,F), statistical comparison by ANOVA with Tukey's HSD post-hoc test: *, P < 0.05; **, P < 0.005; ***, P < 0.0005; ns, not significant. control cell lines (S19B Fig). However, the ratio of the processed ATF6-N (due to higher production in PWS than control INS-1 cells) to total ATF6 (Fig 6F) or of the ratio of ATF6-N to full-length unglycosylated 90-kD (with a higher level in control than PWS INS-1 cells) is significantly greater for PWS compared to control INS-1 cells at 2-5 h timepoints (Figs 6E, S19A and S19B). Thus, ATF6 is being activated to ATF6-N earlier and more robustly in PWS than in control INS-1 β-cell lines. Combined, these results on ER stress activation of the IRE1α/ XBP1, PERK/eIF2α, and ATF6-N pathways indicate that PWS INS-1 β-cells are more sensitive than control INS-1 β-cells to ER stressors.

Discussion
Although PWS has long been assumed to be a neuroendocrine disease of the hypothalamus [3,23], studies using mouse models have implicated a role for the pancreatic endocrine system [25,60]. Here, we generated a novel model through deletion of the~3.16 Mb PWS-orthologous imprinted domain in INS-1 cells, as a proxy for β-cells, in which to investigate PWS βcell biology. The PWS deletion results in a β-cell autonomous defect in production of peptide hormones and in deficiencies of β-cell ER chaperones required for folding, exit and trafficking of secretory peptides. Furthermore, the PWS deletion results in β-cell autonomous deficits in both basal and regulated insulin secretion, although the latter is at least in part consequent to the effect on basal insulin secretion, and a deficit in the secretion of other peptide hormones. Presciently, studies reported 25 years ago demonstrated that PWS adults compared to matched obese controls had significantly reduced first-and second-phase insulin secretion during intravenous glucose tolerance tests [11], attesting to the translational impact of the PWS INS-1 model. These findings strongly support a hypothesis that PWS-genes are required for fundamental β-cell mechanisms affecting production of peptide hormones.

PWS β-cells have concerted deficits in ER chaperone expression and insulin production/secretion
Basal insulin secretion and GSIS are dramatically reduced both in vivo in the TgPWS-deletion mouse model [25] and the new in vitro PWS INS-1 β-cell model. As PWS β-cells in both models are glucose-responsive this indicates the defect is in a fundamental component of the secretory apparatus. In TgPWS mice, perinatal pancreatic insulin secretion dysfunction was associated with intracellular retention of aged insulin and reduced islet insulin and C-peptide content as well as reduced islet glucagon content and plasma glucagon levels suggestive of broader endocrine cell dysfunction [25]. Extending these observations, in our current study we have found that all insulin isoforms that are expressed in and secreted from the PWS INS-1 β-cell are decreased, as measured by proteomics and western blot analyses. Remarkably, these equivalent insulin secretion deficits between PWS mice [25] and the cell autonomous INS-1 model of PWS (this work) were associated with distinct transcriptional responses. In vivo, an apparent physiological compensatory mechanism increased levels of mRNAs encoding all major pancreatic hormones, including insulin, amylin, glucagon, somatostatin, and pancreatic polypeptide, and several secretory polypeptides although this failed to correct the secretory deficit [25]. In contrast, in PWS INS-1 cell lines with no neural or humoral input there was diminished mRNA expression of the endogenous rat Ins1 and Ins2 genes, a mIns2::mCherry transgene, and those encoding two other secreted peptides (Iapp and Npy). Taken together, the physiological regulation of insulins is dysregulated at multiple levels in PWS β-cells.
A further critical finding in the PWS INS-1 model is a concurrent down-regulation at the transcriptional and polypeptide levels of numerous ER chaperones, including for GRP78 and GRP94 that are major facilitators of insulin folding and trafficking in the ER. GRP94 directly binds pro-insulin, and chemical inhibition or shRNA knockdown or genetic ablation of the Hsp90b1 gene results in diminished insulin processing and secretion, larger but immature insulin granules, and activation of the PERK ER stress pathway [67]. Similarly, knockdown of Hspa5 in INS-1(832/13) cells reduced insulin biosynthesis and secretion [68], a finding validated in human EndoC-βH1 cells [69]. The acute reduction of single ER chaperones resulting in reduced insulin production and secretion indicates a direct impact at the level of ER insulin protein folding and trafficking that is recapitulated in the PWS INS-1 model where we describe a chronic shortage of multiple ER-chaperones. Therefore, we interpret the broad ER-chaperone deficiency in the PWS INS-1 model as a primary molecular abnormality that contributes to diminished insulin secretion.

Chronic stress-independent ER chaperone deficits in PWS β-cells prevents ER chaperone dosage compensation
Prior evidence from mouse knockout models of key ER chaperone genes has shown that they are subject to dosage compensation to maintain homeostasis. Although knockouts for either Hspa5 (GRP78) or Hsp90b1 (GRP94) are lethal early in mouse embryogenesis, heterozygous mice are viable [35, [70][71][72]. Intriguingly, upregulation of GRP94 and other ER chaperones occurs in Hspa5 +/-mice [70] and similarly for GRP78 in Hsp90b1 -/-ES cells [71], indicative of dosage compensation among ER chaperones. A similar mechanism operates in Hspa5 +/mice to attenuate diet-induced obesity and insulin resistance [72]. Compensatory changes with upregulation of GRP94 or GRP78 and other ER chaperones including PDIA6 also occurred with shRNA knockdown of Hspa5 or Hsp90b1, respectively, in a mouse cell line [73]. In contrast, the widespread deficiency of ER chaperones, including GRP78 and GRP94, that we describe for PWS INS-1 β-cells would preclude compensatory increases. Because of an inability to compensate, we hypothesize that PWS β-cells have a chronic deficit in ER chaperone production which would interfere with folding of proinsulin and/or delay ER transit of cargo (e.g., hormones) along the secretory pathway. Consistent with this hypothesis, our previous in vivo results using a Timer fluorescent protein demonstrated an accumulation of aged insulin in islets of PWS mice [25].
ER chaperones also function as sensors of ER stress, a physiological quality control mechanism responding to accumulation or failure in degradation of misfolded secretory proteins [3,61,64,74,75]. ER stress responses, governed by the IRE1α/XBP1, PERK/eIF2α/Ddit3 (CHOP), and ATF6 regulatory pathways, aim to recover ER homeostasis by upregulating ER chaperone gene expression as part of the UPR [39, 61,64]. In unstressed cells, GRP78 forms stable complexes with each of IRE1α, PERK, and ATF6 in the ER lumen [61,62], whereas ER stress releases GRP78 and induces each pathway. Due to the demand to fold high levels of insulin, a protein notoriously difficult to fold [75], β-cells have a basal level of ATF6 activation to maintain higher levels of GRP78 than for most cell types [76]. In contrast, unresolved ER stress within β-cells leads to diabetes in mouse models and human [39,77]. For example, insulin gene mutations lead to insulin misfolding with induction of ER-stress and broad upregulation of the UPR [78], a converse mechanism to the reduced ER chaperone levels we demonstrated within PWS INS-1 cells under both basal conditions and during GSIS. Indeed, lower GRP78 levels in PWS β-cells likely accounts for the earlier and more robust ER stress induction by thapsigargin or tunicamycin by lowering the threshold required to dissociate ER stress activators resulting in cells that are poised to a greater degree for response by all three master regulatory pathways. Although normal β-cells in vivo cycle through states of elevated and low UPR coordinate with insulin gene expression and ER protein folding load [79][80][81], our omics data on steady-state mRNA and protein levels of bulk PWS INS-1 cells under basal unstressed or GSIS conditions suggests a stress independent defect in ER chaperone gene expression.

Mechanisms for dysregulated peptide hormone and ER chaperone expression in PWS β-cells
The coordinate downregulation of at least ten ER chaperone genes, including Hspa5 and Hsp90b1, in PWS β-cells, but not the whole suite of UPR genes, supports the hypothesis that a PWS-imprinted gene or genes governs a pathway that coordinately regulates ER chaperones in a unique gene regulatory network (GRN). Similarly, the GRN may include genes encoding hormones given the corresponding coordinate reductions in expression. It is unknown whether the putative GRN is controlled via transcriptional, mRNA stability, and/or mRNA processing mechanisms. Although many of the DEGs that encode ER chaperones in PWS INS-1 β-cells are known ATF6 and/or XBP1 targets, numerous ATF6, IRE1α/XBP1, and ATF4/CHOP (PERK pathway) target genes [82][83][84][85][86][87][88][89][90][91] are not dysregulated in PWS INS-1 lines, suggesting a novel GRN in PWS β-cells. Candidate transcription factors (TFs) within the PWS-ER chaperone GRN are those with binding site motifs enriched in the promoters of DEGs in PWS β-cells, including, but not limited to the known ATF6 co-factor NFYA, which has roles in insulin secretion and glucose homeostasis [92], as well as PPARB/D also with known roles in β-cell mass and insulin secretion [93]. Intriguingly, a recently published study found that TFs of the nuclear receptor 4A (NR4A) family that participate in long-term memory in mice act through coordinate regulation of numerous ER chaperones [94], with a high degree of concordance with the dysregulated ER chaperones that we identified in PWS INS-1 cell lines; nevertheless, NR4A pathway factors are not DEGs in PWS β-cells.
An alternative hypothesis is that a reduction in secretory cargo (i.e., insulin) results in a lowered cellular requirement for ER chaperones. Insulin production has been successfully silenced at the post-transcriptional level using RNA interference in the mouse pancreas [95] or human insulinoma cells [69], but neither study assessed downstream cellular or metabolic consequences. In contrast, transcriptional and proteomic changes were assessed from islets of an insulin-deficient mouse model with tamoxifen-inducible deletion of Ins2 alleles on an Ins1 global knockout background (Ins1 -/-) [96]. At the single time point analyzed in this mouse model with~50% reduced insulin production and prior to onset of diabetes and hyperglycemia, Ins2 mRNA levels were 1.4-fold reduced (with no Ins1 transcripts) and was associated with transcriptional and functional reductions in regulators of ER stress, including Trib3, Atf5, Ddit3, and Xbp1 splicing, leading also to an increase in β-cell proliferation [96]. These reductions in insulin production are similar in level to the TgPWS mouse model [25] and to the PWS INS-1 model described here, nevertheless, in the tamoxifen-induced Ins2 (Ins1 -/-) knockout mouse pancreas levels of mRNAs encoding ER chaperones were mostly unchanged (Creld2, Pdia6, Dnajb11, Dnac3, Hyou1, Ppib), to slightly increased (Hsp90b1, Hspa5) or decreased (Sdf2l1, Pdia4). These considerations therefore provide strong support for a GRN hypothesis for the coordinate reduction in ER chaperones in PWS β-cell models.
It remains to be determined which PWS gene or genes regulates the putative GRN. Neither Magel2 nor Ndn have a role, as they are not expressed in INS-1 cells. The top candidate PWS genes that are highly expressed in INS-1 cells are Snrpn, encoding the SmN spliceosomal protein [43], Snurf, encoding a small arginine-rich nuclear protein [18], and Snord116, a tandemly duplicated C/D-box snoRNA that may function as a sno-lncRNA that sequestrates FOX2 and other RNA binding proteins in alternative splicing pathways [22,97], or through direct binding and regulation of as yet undefined RNA targets. Further studies are necessary to distinguish amongst the candidate PWS genes and to identify downstream GRN steps leading to coordinated ER chaperone expression.

Evolution of PWS gene functions in secretory endocrine cells
Recent work has shown hypothalamic deficiencies in Magel2-null mice as well as for iPSCand dental-pulp stem cell neuronal models, specifically in secretory granule components including PCSK1, PCSK2, CPE, granins (e.g., CHGB and others), and numerous neuropeptides [23]. These studies established that MAGEL2 plays a role in hypothalamic neuroendocrine cells by regulating neuropeptide production and secretion via endosome recycling to prevent lysosomal degradation of secretory granule proteins [23], a process downstream of the ER in protein trafficking and secretory pathways. As discussed above, our study shows that a different PWS-gene (e.g., Snrpn, Snurf, and/or Snord116) is putatively responsible for regulating ER chaperone and hormone (e.g., insulin) biosynthesis and secretion from pancreatic βcells. In contrast to the reported reduction of prohormone convertase PC1 in PWS iPSCderived neurons and in Snord116-null mice [60], no deficiency in Pcsk1 mRNA or PC1/3 levels or in Pcsk2 or PC2 was observed in PWS β-cell lines, which may reflect differences in cell type, genetics, or experimental conditions. Combined, these observations in hypothalamic neuroendocrine cells and pancreatic β-cells indicate that at least two PWS-imprinted genes regulate related neuropeptide and peptide hormone secretory pathways, suggesting that the PWS domain may function as a "mammalian operon" or synexpression group [98,99]. As the PWSdomain arose evolutionarily in a eutherian mammalian ancestor [100], this suggests emergence of functions acting as molecular rheostats to regulate secretory pathways in endocrine cells that control growth, metabolism, and neural pathways.

PWS genes regulate glucose and hormone homeostasis: implications for metabolic disease
Maintaining glucose homeostasis relies on exquisite coordination between secretion of the opposing hormones insulin and glucagon from pancreatic β-and α-cells, respectively, in response to changes in blood glucose, and disruption of these processes contributes to the pathogenesis of metabolic disorders, including type 2 diabetes [39, [101][102][103][104]. Although insulin and glucagon play opposite regulatory roles both in normal metabolic homeostasis and dysfunction in type 2 diabetes, PWS mice are remarkable in having low blood levels of both insulin and glucagon [16,25]. As recent studies indicate that GRP78 interacts with glucagon in αcells [105], it will be important to further assess dysregulation of ER chaperones and glucagon secretion in PWS α-cells both in cell culture and within PWS mouse models. Intriguingly, quantitative trait locus (QTL) studies in mice link genetic variation in blood insulin and glucose levels in or near to the PWS-orthologous domain [106]. Further studies are warranted to examine the role of PWS-imprinted genes in these metabolic traits in humans, and whether βcell or other endocrine cell deficits in ER chaperones and hormone secretion contribute to clinical phenotypes, including episodes of hypoglycemia in PWS subjects [13,14]. Indeed, recent studies in mice show that hypoglycemia induces hunger and food intake acting through brainstem neurons projecting to the hypothalamus [107], and it will be important to assess the potential contribution of such mechanisms to hyperphagia and obesity in PWS. As PWS genes regulate islet development, β-and α-cell secretion [16,25], and a GRN affecting ER chaperone and insulin secretion (this study), identifying mechanisms by which PWS-genes carry out these critical β-cell functions will illuminate the pathogenesis and may reveal effective treatments for not only PWS, but for common disorders with deficits in glycemic homeostasis and islet hormone secretory pathways.

Cell culture
INS-1 lines were cultured using "RPMI 1640 media without glucose" (Life Technologies)

Genome editing
Pairs of sgRNAs (S6 Table) were designed (http://crispr.mit.edu/) to target sites flanking the rodent PWS-orthologous domain to delete all paternally-expressed imprinted genes (Fig 1A), and cloned into the pX330 CRISPR/Cas9 vector [108] (Addgene). INS-1 parental cells were transfected using lipofectamine 3000 with pX330-vectors encoding sgRNAs targeting proximal of Frat3 and distal (between Snord115 and Ube3a) of the PWS-domain and/or the control pEGFP-N3 vector (using 3 μg or 500 ng of DNA per vector per T75 flask or 6-well plate, respectively). After 4 days culture at 30˚C or 37˚C, cells were harvested for DNA isolation or flow cytometry (FACS ARIAII in the Rangos Flow Cytometry Core Laboratory) with~300 GFP-positive cells plated in five 96-well plates and clonally expanded to 12-well plates for DNA isolation. Clonal lines were screened by deletion-PCR and positive lines further expanded for DNA, RNA, protein, and molecular cytogenetic analyses. PCR primers for deletion-PCR, inversion-PCR, and scarred-or intact-allele PCRs are in S7 Table. For DNA sequencing (Genewiz) of deletion breakpoints and scarred or intact sgRNA sites, we Sanger sequenced PCR products directly or from pJET (Thermo Fisher) cloned PCR products. Potential CRISPR/Cas9 off-target sites were predicted in silico using CRISPOR (http://crispor.tefor. net/); top predicted off-targets were PCR amplified (S7 Table) and directly Sanger sequenced.

Molecular cytogenetics
Cytogenetic studies were carried out in the University of Pittsburgh Cell Culture and Cytogenetics Facility. Briefly, cultured INS-1 cell lines were treated with 0.1 μg/ml Colcemid 1 h, harvested, fixed, and slides processed for metaphase FISH by standard methods [109]. Fluorescent probes were prepared by labeling BAC (BACPAC Genomics) DNA using nick translation (Enzo Life Sciences, Inc.), including CH230-114P11 from the central PWS-domain (encodes U1A-Snurf-Snrpn-Snord107-Snord64) with Orange-dUTP and control CH230-2B12 mapping several Mb distal of the PWS domain (encodes Cyfip1-Nipa2-Nipa1-Herc2) with Green-dUTP. Probe and slide preparation, hybridization, and DAPI staining were by standard cytogenetics methods, with FISH analyses on an Olympus BX61 epifluorescence microscope (Olympus Microscopes) and image capture and analysis using the Genus software platform on the Cytovision System (Leica Microsystems) [109].

Droplet digital PCR (ddPCR) genomic copy number assays
For genomic copy number 5 ng of EcoRI digested genomic DNA was used as input with amplification of Snord107, PWS-IC, and Mirh1 as targets and Ube3a as a reference; during clonal derivatization EvaGreen chemistry was used whereas Fam/Hex TaqMan probes were used in a final analysis. Absolute copy number of Snord107 and Poisson confidence intervals were calculated by numerical approximation [110]. Primers and probes for copy number ddPCR are listed in S7 Table. DNA methylation Genomic DNA was bisulfite converted using the EZ DNA methylation Gold kit (Zymo Research). Outer first round genomic PCR was performed using primers to amplify the Snurf-Snrpn promoter region (annealing temperature or Ta of 64˚C; S7 Table), with subsequent second round genomic PCR performed using PCR primers specific for the maternal, methylated or paternal, unmethylated alleles (Ta of 60˚C; S7 Table). HotStart Taq polymerase (Qiagen) was used for all methylation PCR.

Insulin secretion assays
Cells were plated at 1.0 x 10 6 cells/well, then 24 h later at~80% confluency were washed with PBS containing Mg ++ and Ca ++ (Gibco), pre-incubated 1.0 h in KRBH (129 mM NaCl, 5 mM NaHCO 3 , 4.8 mM KCl, 1.2 mM KH2PO 4 , 1.2 mM MgSO 4 , 2.5 mM CaCl 2 , 10 mM HEPES, 0.1% BSA) with low glucose (LG; 2.8 mM) and washed in KRBH (without glucose). Following a 30 min incubation in KRBH-LG, 1.0 ml low glucose secreted fractions were removed and centrifuged 13,000 rpm for 5 min., aliquoted and stored at -80˚C. The cells were then incubated 30 min in KRBH-high glucose (HG; 22 mM), and high glucose secreted fractions similarly centrifuged and stored at -80˚C. Finally, cells were washed in PBS containing Mg ++ and Ca ++ , harvested in RIPA lysis buffer and protease inhibitor, centrifuged at 13,000 rpm for 15 min at 4˚C and the supernatant (protein) stored at -80˚C. Protein was determined using Pierce BCA protein assay kit (Thermo Fisher). Insulin was measured by the RIA and Biomarkers Core of the Penn Diabetes Research Center, University of Pennsylvania, using an ultrasensitive rat insulin ELISA kit (Alpco Diagnostics). ANOVA followed by the Tukey's HSD post-hoc test was used to assess differences between conditions and genotypes using Prism 8 (GraphPad Software, Inc).

Reverse transcription-PCR
Total RNA was isolated from PWS and control INS-1 lines grown as above (7.5 mM glucose) by Trizol harvest and miRNeasy (Qiagen) column purification. RNA quality was assessed by RNA TapeStation (Agilent) and quantified by broad range Qubit (Thermo Fisher) fluorometric analysis. First strand cDNA synthesis from 1 μg RNA was carried out using random hexamer primed RT by using Super Script IV (Thermo Fisher). Primers for gene specific RT-PCR amplification are in S8 Table. Quantification of the ratio of "spliced" Xbp1 isoform to total Xbp1 was determined by ethidium bromide gel densitometry with Image Lab (Bio-Rad).

RT-Droplet digital-PCR (RT-ddPCR) assays
Template cDNA pools were generated from 1 μg RNA using cells grown under standard glucose conditions (as above) by using iScript (Bio-Rad) RT with first-strand synthesis priming from a mixture of random hexamers and oligo-dT. Droplet generation, PCR and reading of singleplex EvaGreen-based ddPCR reactions was carried out using an automated QX200 ddPCR system (Bio-Rad). A total template input volume of 2 μl of cDNA diluted in a range from undiluted to 1:400 (40 ng to 0.10 ng RNA equivalents per reaction) determined empirically based on the absolute expression level of target genes was used per 20 μl reaction. Primer sequences are listed in S8 Table. Expression of target genes was normalized to the stable and modestly expressed Gpi (1:5 diluted, 8.0 ng RNA equivalents). Relative expression of each target gene was calculated using multiple independent RNA biological samples and technical replicates, with removal of rare statistical outliers greater than two standard deviations, by normalizing to both Gpi as a reference gene, and the average expression level in control lines. Comparison of relative target gene expression in PWS and control lines was evaluated using the Welch's t-test for unequal variance with a significant threshold of P < 0.05.

Proteomics
Each of the 3 control cell lines (5-9, 2, 16) and the 3 PWS cell lines (3, were grown in 6-well plates, harvested and then washed with PBS containing Mg ++ and Ca ++ , incubated 30 min in KRH-high glucose (22 mM, glucose-induced insulin secretion condition, with BSA excluded from the KRBH buffer) after which the media was collected, centrifuged 10 min 1300 rpm, and supernatant (secretory fractions 1-6; Fig 3A) collected and stored at -80˚C. Cellular proteins were harvested by direct addition of 1.0 ml of methanol-acetic acid (90% methanol, 9% water, 1% acetic acid) to each well, transferred to a microcentrifuge tube and centrifuged at 1300 rpm 15 min. The supernatant (soluble protein fractions 7-12; Fig 3A) and pellets (insoluble protein fractions 13-18; Fig 3A) were collected and stored at -80˚C. The secretory and cellular fractions were dried in vacuo and stored at -80˚C. In preparation for mass spectrometry, proteins from each fraction were dissolved in 8 M urea, 100 μM Triethylammonium bicarbonate (TEAB) pH 8.5, reduced with tris(2-carboxyethyl)phosphine (TCEP) and alkylated with chloroacetamide, then diluted to 2 M urea with 100 mM TEAB, addition of 0.5 μg trypsin (Promega) and placed in a 37˚C shaker for 16 h. Secretory protein fractions were analyzed on an Orbitrap Elite mass spectrometer (Thermo Fisher) while cellular protein fractions were analyzed on an Orbitrap Fusion mass spectrometer (Thermo Fisher). The 6-plex Tandem Mass Tag (TMT) system for quantitative proteomics (Thermo Fisher) was used to compare methanol-acetic acid insoluble cellular proteins (fractions 13-18) for control and PWS cell lines. The TMT labeled samples were analyzed on an Orbitrap Fusion Lumos mass spectrometer (Thermo Fisher).
Peptide/protein identification and quantification were determined using IP2 (Integrated Proteomics Applications). The MS raw data files were converted using RawConverter [111] (version 1.1.0.23) with monoisotopic option. For peptide identification, tandem mass spectra were searched against a database including the UniProt Rat database one entry per gene (21589 entries released 5/30/2021), these entries scrambled between K and R to make a decoy database, common contaminants, peptides and custom proteins using ProLuCID [112], and data filtered using DTASelect [113]. Quantitation of TMT samples was calculated with Census version 2.51 [114] and filtered with an intensity value of 5000 and isobaric purity value of 0.6 [115]. The Quantitative@COMPARE feature of IP2 was used to determine statistical significance.

Immunoblot analyses
Cells were grown under standard glucose conditions with or without the addition of DMSO, thapsigargin or tunicamycin (as above). Whole cell lysates were harvested by direct lysis in cold radioimmunoprecipitation buffer (RIPA) with the addition of EDTA and combined protease and phosphatase inhibitors (Thermo Fisher) followed by clearing of insoluble material by centrifugation at 13,000 RPM for 10 minutes. Proteins were separated by SDS-PAGE with 2-mercaptoethanol as a reducing agent using criterion sized 4-15% gradient tris-glycine for broad range of molecular weights greater than 30 kDa or 16.5% tris-tricine for high resolution of smaller proteins (Bio-Rad). Proteins were transferred to nitrocellulose membranes and probed with antibodies [commercially obtained except for ATF6 [65], with dilutions as in S9 Table]. Chemiluminescence detection of HRP-conjugated secondary antibodies was performed with either Clarity or Clarity Max Western ECL detection kits on a Bio-Rad Chemi-Doc XRS+ imager. Densitometry measurements were made using Image Lab (Bio-Rad) software and exported for analysis in R. Protein normalization to a reference protein (i.e., GPI, GAPDH, or α-Tubulin) was made by dividing the band intensity for the target protein by the reference intensity and then subsequently by the average expression of the target protein in control cell lines. Statistical comparison of normalized protein levels between PWS and control samples was made by either Welch's unequal variance t-test or by ANOVA with Tukey's HSD post-hoc test for multiple comparison where appropriate, with a significance threshold of P < 0.05.

Transmission electron microscopy
Monolayers of INS-1 cells were fixed in 2.5% glutaraldehyde in 100 mM PBS, post-fixed in aqueous 1% osmium tetroxide, 1% Fe 6 CN 3 for 1 h, and dehydrated prior to embedding in Polybed 812 resin (Polysciences). Ultrathin cross sections (60 nm) of the cells were obtained on a Riechart Ultracut E microtome, post-stained in 4% uranyl acetate for 10 min and 1% lead citrate for 7 min. Sections were viewed on a JEOL JEM 1400 FLASH transmission electron microscope (JEOL) at 80 KV. Images were taken using a bottom mount AMT digital camera (Advanced Microscopy Techniques).

RNA-seq and bioinformatics
Total RNA (from cells grown under 7.5 mM glucose) for each of the 3 control cell lines (5-9, 2, 16) and the 3 PWS cell lines (3, was used to prepare rRNA depleted Illumina TruSeq 75-bp paired-end stranded (fr-firststrand) sequencing libraries and sequenced to a depth of~40 million reads on a NextSeq500 instrument at the UPMC Children's Hospital of Pittsburgh Genomics Core. RNA-seq data analysis was performed using computational resources available from the University of Pittsburgh Center for Research Computing. Quality of reads was assessed by FastQC and sequencing adapters trimmed by cutadapt using the combined TrimGalore tool [116]. The sequencing reads were aligned to the Ensembl v98 rat genome to which the sequence and annotation of the human INS-neo R and mouse Ins2-C-mCherry transgenes had been added along with annotation of previously undefined genomic features (e.g., Ipw, ψSnurf, and upstream U1 Snurf-Snrpn exons) using the STAR splice-aware aligner [117]. Gene feature counts were tabulated with HTSeq using either the default (-nonunique none) or customized (-nonunique all) options to enable the inclusion of multi-copy (i.e., Snord116) and overlapping (bicistronic Snurf-Snrpn) genes which otherwise were excluded as ambiguous reads under the default options [118]. Differential expression analysis was performed with DESeq2 using a cutoff of Padj < 0.1 as calculated by the Benjamini-Hochberg multiple comparison procedure [119]. Similar overall transcriptomic results were obtained with RSEM based analysis [120] but had lower counts for both multi-mapped reads for Snord116 and for ambiguous bicistronic Snurf-Snrpn reads. Gene ontology enrichment analysis of up-and down-regulated gene sets was done using DAVID [121]. Additional gene ontology and upstream analysis utilized Enrichr [122].
Sequencing libraries of small RNAs including snoRNAs were made from the same total RNA isolation but using a miRNA-Seq (Qiagen) library prep kit with a modified size selection for up to 200-bp fragments. Single end reads were generated and sequenced. UMI tools [123] and Cutadapt [124] were used to deduplicate and remove adapters in preprocessing of the fastq reads for alignment with Bowtie2 [125]. Similar to the total RNA-seq, HTSeq required the-nonunique all option to accurately quantitate the multicopy snoRNA and miRNA genes, and DESeq2 was used for differential expression analysis to compare genotypes. A full set of bash scripts for analysis of both total and small RNA-seq data sets including custom annotations are provided at the github repository (https://github.com/KoppesEA/INS-1_PWS_ RNA-Seq). Sequenced PCR fragments spanning sgRNA1 are from the gel in S1I Fig (F-H) Sequence of intact sgRNA3 site in parental INS-1 (F), and scarred alleles with a G/C insertion or a C/G insertion at the PAM-3 site (pink highlight) in line 5-9 (G) or line 5-5 (H), respectively. Sequenced PCR fragments spanning sgRNA3 are from the gel in S1I Fig It may be noted that as the deletions for 5-9 and 5-5 are on the maternal allele, the sgRNA1 and sgRNA3 scarred alleles can be inferred to occur on the paternal allele for each cell line. Further, as each of lines 5-5 and 5-9 have different sgRNA3 scarred alleles (despite sharing deletion breakpoints and sgRNA1 scarred allele mutations) then these two cell lines clearly arose as independent genome editing events. (JPG) S3 Fig. Sanger sequencing of sgRNA1 and sgRNA3 off-target sites in maternal deletion INS-1 clonal lines. Sanger sequencing chromatographs of direct sequenced off-target genomic PCRs are shown for the top three ranked predicted off-target sites for sgRNA1 highlighted by the sgRNA seed (yellow) and for sgRNA3 (blue), with SpCas9 NGG PAMs (grey), and deviation in off-target from sgRNA sequence (brown).  S1 Fig), providing the PAT-allele specificity. Further proximal, the alt-sgRNA70-3 site differs at two 5 0 nucleotides of the sgRNA and is part of a segmental-duplication (to be described elsewhere; manuscript in preparation). All symbols are as for S1A Fig  (B) Deletion-PCR assays for a CRISPR/Cas9 screen of four pairs of sgRNAs flanking the 3.143 Mb PWS-domain, using control INS-1 cells. EGFP represents control transfections. Of the four pairs of sgRNAs, the highest efficiency is obtained for sgRNA70-3 + sgRNA79-1. (C) Deletion-PCR assay following transfection of MAT-del lines 5-5 and 5-9 using the CRISPR (sgRNA70-3 + sgRNA79-1)/Cas9-expressing vector shows specific targeting of the paternal allele of the PWS-domain. The positive (+) control is from S4B Fig (D-F, H-L) FISH studies using rat FISH probes, as for S1B-S1F

FISH for control line 16 with a MAT-deletion in most cells. (G)
Deletion-PCR for PAT-del line 3 using alt-sgRNA70-3 specific F and sgRNA79-1 R primers. As compared to the faint deletion breakpoint band using a sgRNA70-3 F PCR primer (Fig 1E), this alternate (alt) assay provides a high degree of specificity, with fainter deletion-PCR bands for lines   Figs 1A and S4A). The alt-sgRNA70-3 site differs only at 2 of the most 5 0 nucleotides of the sgRNA (purple highlight) and is part of an uncharacterized segmental-duplication. The line 3 genome-editing event with a 3.268 Mb deletion has at the breakpoint a 233-or 234-bp insertion of sequence with 97% identity to E. coli (purple sequence, with gsiD amber stop codon indicated in red), followed by a distal deletion breakpoint with a single G/C nucleotide insertion or polymorphism in E. coli sequence (orange highlight) that occurred at a DSB at the PAM-2 position of the distal sgRNA79-1 site (pink highlight). The deletion-PCR breakpoint fragment for the PAT-del line 3 is from the gel in S4G Fig (E-G) Sequence of intact sgRNA70-3 site in parental INS-1 (E) and maternal (MAT)-deletion (del) control lines 2 (F) and 16 (G). Sequenced PCR fragments spanning sgRNA70-3 are from the gel in S4N Fig (H-J The upper 6 rows of gels are from standard RT-PCR assays that use random hexamer primers for the RT primer and typical gene-specific RT-PCR primers, while the lower 3 rows of gels are from RT-PCR assays that use a strand-specific primer for RT followed by RT-PCR with either a Ube3a (ex12 F + ex13 R) or Ube3a-ATS (ex13 R + int12 F) primer set. Combined, the data show that transcripts from the Ube3a-ATS-region that are detected when using random primers for RT derive from Ube3a primary (1˚) transcripts since there is no expression from the Ube3a-ATS strand. (G) Potential ψSnurf amino acid sequence encoded by the ψSnurf gene. Green or blue highlight represents potential amino acid changes or in-frame deletion, respectively, and red asterisks the stop codons. (H) Potential truncated ψSmN amino acid sequence encoded by the ψSnrpn gene. Symbols as for (G), except blue highlight, amino acids resulting from a frame-shift mutation. Assuming no evolutionary selection for function, we can date the age of rat ψSnurf-ψSnrpn based on a total of 26 mutations/substitutions in 1,288-nt of homologous sequence to rat Snurf-Snrpn. The 2.02% divergence (97.98% similarity) corresponds using a mutation rate formula for silent site substitutions and intronic sequences of 5 × 10 −9 to 7 × 10 −9 mutations/site/year [126] to a pseudogene origin~2.88-4.0 million years ago (mya), which correlates well with prior molecular dates for divergence of black and brown rats about 2.0-2.9 mya [127,128].  Cacna1a, Tmem176a, Tmem176b), with apparent upregulation for an exogenous human insulin (Hs, Homo sapiens)-neomycin resistance (NeoR) transgene in the INS-1 lines, although the latter is artifactual due to transgene silencing in line 16 (see S16A, S16I-S16K and S16M Fig). (D,E) However, the RT-ddPCR analyses failed to validate (D) several candidate upregulated DEGs from the RNA-seq data (Fig 4A), or (E) candidates for downregulated DEGs based on data from PWS iPSC-derived neuronal and Snord116-deficient mouse models [60,129]; likewise, another mouse study [130] did not confirm Pcsk1. Statistical comparison by Welch's t-test: *, P < 0.05; **, P < 0.005; ***, P < 0.0005. . Clustergrams generated based on enrichment P-value significance which shows as the red highlight of TF names. Notably in the downregulated genes in PWS INS-1 cells a prominent cluster of genes including many ER chaperones were enriched for ATF6-cofactor NFYA and NFYB binding sites and other potential regulators including CPEB1, RFX5, IRF3, CREB1 and SREBF1 (see S15A-S15D Fig). Similarities to other model systems revealed analogous gene-expression changes including for Xbp1 perturbations in an adipose cell line [131] and the Pparb/d KO mouse pancreas [93] (see S15E-S15F Fig).  (Figs 2-5) and other secretory peptides such as IAPP (Fig 3C and 3D) do not demonstrate clearly visible insulin secretory vesicles, likely because of the relative rarity of these compared to a number in the tens of vesicles detected in the INS-1E cell line [132] that produces much more abundant insulin (see Fig 5A). Relative ratio of each ATF6 isoform as a fraction of the total ATF6 during the timecourse of tunicamycin treatment. Line graphs represent mean of three measurements at each time point, except n = 6 using technical replicates for 4 hr and n = 6 using two biological replicates for 5 hr timepoints, for full-length glycosylated (FL-G) over total (Red); FL-UG over total (Blue) and N over total (Green) with control (Con) as solid lines and PWS cell lines as dotted lines. Data indicates that there is a clear trend to more robust processing of FL-UG to N at early tunicamycin timepoints in PWS cell lines. (JPG) S1 Table. Differentially expressed genes (DEGs) in PWS vs. control INS-1 lines from RNAseq with HTSeq feature counts. Data presented from RNA-seq expression bioinformatics pipeline using a custom rat annotation build with STAR aligner, gene level counts with HTSeq (with options-mode union and-nonunique all) and differential expression with DESeq2.