Protection against Glucolipotoxicity by High Density Lipoprotein in Human PANC-1 Hybrid 1.1B4 Pancreatic Beta Cells: The Role of microRNA

Simple Summary Loss of the cells which secrete insulin, the hormone which controls blood sugar levels, can occur due to the damaging effects of high levels of sugar and fat in the bloodstream, and is a key factor which triggers type 2 diabetes. Protection against this damage is provided by one of the fat and protein complexes which carry the ‘good’ cholesterol in the bloodstream, a complex which is called high-density lipoprotein. This study looks at the changes in the expression of small pieces of RNA, called microRNA, triggered in insulin-secreting cells by treatment with high density lipoprotein. These small RNA sequences can help to regulate the expression of genes which can contribute to the protective effect of high-density lipoproteins. Delivery of one of these sequences was able to modulate gene expression in cells, but could not provide enough protection against the damaging effects of high sugar and fat. We conclude that the complex changes in numerous microRNA sequences caused by high density lipoproteins cannot be replicated by a single sequence, suggesting that networks of regulatory microRNA sequences are more important than single sequences. The results will inform drug discovery strategies which focus on high density lipoprotein and microRNA. Abstract High-density lipoproteins provide protection against the damaging effects of glucolipotoxicity in beta cells, a factor which sustains insulin secretion and staves off onset of type 2 diabetes mellitus. This study examines epigenetic changes in small non-coding microRNA sequences induced by high density lipoproteins in a human hybrid beta cell model, and tests the impact of delivery of a single sequence in protecting against glucolipotoxicity. Human PANC-1.1B4 cells were used to establish Bmax and Kd for [3H]cholesterol efflux to high density lipoprotein, and minimum concentrations required to protect cell viability and reduce apoptosis to 30mM glucose and 0.25 mM palmitic acid. Microchip array identified the microRNA signature associated with high density lipoprotein treatment, and one sequence, hsa-miR-21-5p, modulated via delivery of a mimic and inhibitor. The results confirm that low concentrations of high-density lipoprotein can protect against glucolipotoxicity, and report the global microRNA profile associated with this lipoprotein; delivery of miR-21-5p mimic altered gene targets, similar to high density lipoprotein, but could not provide sufficient protection against glucolipotoxicity. We conclude that the complex profile of microRNA changes due to HDL treatment may be difficult to replicate using a single microRNA, findings which may inform current drug strategies focused on this approach.


Introduction
The 'pre-diabetic' condition is characterised by insulin resistance, which eventually triggers beta cell failure, loss of compensatory insulin secretion and the onset of frank type 2 diabetes. Beta cell damage can be induced by glucolipotoxicity (GLT) [1], which describes the deleterious effects of elevated levels of free fatty acids (FFA) on glucose Human PANC-1.1B4/human beta cell hybrid 1,1B4 cells (1.1B4 cells) were obtained from Public Health England Culture Collections (#10012801). This cell line was generated by electrofusion of human pancreatic beta cells and the immortal human PANC-1 epithelial cell line, characterised by McCluskey et al. (2011) [29], and recently employed by Nemecz et al. (2019) [30] to study glucolipotoxicity, and by Honkimaa et al. (2020) to study genetic adaptation of Coxsackievirus B1 [31]. This human cell line was selected for this study as human and rodent microRNA sequences and targets are not always conserved; however, please note that this cell line has recently been discontinued (ECACC) due to possible contamination with rat DNA (https://www.phe-culturecollections.org.uk/ products/celllines/discontinued-ecacc-cell-lines/11b4.aspx accessed on 20 January 2021).

Glucolipotoxicity Assays
A working solution of 5mM palmitic acid conjugated with 10% (w/v) fatty acid free BSA was diluted to the appropriate concentration in serum-free RPMI 1640, and glucose added from a 1 M stock to achieve a final concentration of 30mM. Cell viability (2.5 × 10 4 cells/well in a 96-well plate) was measured by conversion of thiazolyl blue tetrazolium bromide to formazan [33] and using the PrestoBlue cell viability assay (Thermo Fisher Scientific, Waltham, MA, USA) [34]. CellEvent Caspase-3/7 Green Detect Reagent (Thermo Fisher Scientific, Waltham, MA, USA) was used to monitor apoptosis [35].

Total RNA Isolation, Microchip Analysis and Q-PCR
Total RNA (5 × 10 5 cells/well in a 12-well plate) was collected using either a Directzol RNA MiniPrep Plus kit (Zymo Research, Cambridge, UK) or a Nucleospin RNA kit (Machery-Nagel, Düren, Germany). Complementary DNA (cDNA) used for measurement of gene expression (mRNA) was generated from 250 ng RNA using the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems, Thermo Fisher Scientific, Waltham, MA, USA) according to manufacturer's instructions; cDNA used to measure miRNA and/or mRNA was generated using the miScript II RT kit (Qiagen, Manchester, UK) and HiFlex buffer, as indicated by the manufacturer.
Microarray analysis of miRNA expression was performed commercially by LC Sciences (Houston, TX, USA); 2632 unique mature human miRNA sequences derived from version 22 of miRBase [36] were measured using µParaflo microfluidic chip technology (https: //www.lcsciences.com/discovery/applications/transcriptomics/mirna-profiling/mirna/ accessed on 3 January 2018), exemplified by Zhang et al. (2019) [37]. Relative quantitative PCR (Q-PCR) for miRNA and mRNA expression was performed using the primer sequences defined in Tables S1 and S2 (Supplementary Data) using HOT FIREPol EvaGreen qPCR Mix Plus (Solis BioDyne) on a C1000 thermal cycler with a CFX96 real-time system attached; melt curves were performed for all primers and none evidenced dimer formation. The expression of target miRNA and mRNA was determined using the 2 −∆∆ Ct method relative to an invariant control sequence/gene, as indicated in legends to figures; statistical analyses were performed using the ∆Ct values, compared with the relevant housekeeping sequence.

Transfection with miRNA Mimics and Inhibitors
Transfection of 1.1B4 cells (2 × 10 5 cells/well in a 12-well plate) with the hsa-miR-21-5p mimic, and scrambled mimic control (Qiagen), and the miR-21-5p inhibitor and scrambled inhibitor control (Qiagen) was achieved using HiPerFect Transfection Reagent (Qiagen), according to the manufacturer's instructions. In brief, transfection complexes Biology 2021, 10, 218 4 of 17 (100 µL) were added in serum-free RPMI 1640 and incubated for 4h prior to the addition of further serum-free RPMI 1640 (400 µL). The cells were incubated at 37 • C and 5% CO 2 with the concentrations of mimic and inhibitor, and for the periods of time, indicated in figure legends.

Bioinformatic Analysis
MicroRNA sequences differentially expressed after treatment with apoA-I and HDL, compared with the basal control, were analysed using DNA Intelligent Analysis (DIANA)-miRPath v3.0 [38,39] to identify pathways with p < 0.01 association with differentially expressed genes. Upregulated and downregulated miRNAs were analysed separately in miR-Path; this utilized experimentally validated miRNA interactions in DIANA-TarBase v7.0 database [35] incorporated into Kyoto Encyclopedia of Genes and Genomes (KEGG) [40] pathways to identify the most closely associated functions with each group. Diseasespecific pathways were excluded from the results. Interaction networks were created using Cytoscape v3.8.0 [41]. The graphical abstract was created using Biorender.com.

Statistical Analysis
All datasets and methodologies were deposited in Mendeley Data (doi:10.17632/ s3bznmgtpg.1). Results are expressed as means ± SD or SEM of the number of experiments indicated in legends to figures. Statistical analysis was performed by one-way, two-way, or three-way ANOVA and Dunnett's post-test, repeated measures ANOVA and Dunnett's post-test, or Student's t test, as indicated in legends to figures. All statistical testing was performed using Graph Pad Prism software; * p < 0.05; ** p < 0.01 and *** p < 0.001.

Cholesterol Efflux from 1.1B4 Cells
Functional responses to apoA-I and HDL in 1.1B4 cells were established by measurement of [ 3 H]cholesterol efflux to these acceptors, shown in Figure 1. Efflux to apoA-I ( Figure 1A) was relatively low (estimated Bmax = 0.99 ± 0.15%) and plateaued (Kd = 3.70 ±1.84 µg mL −1 ), with significant increases in efflux above the control condition noted at 10 and 20 µg mL −1 . By contrast, efflux to HDL ( Figure 1B) exhibited a curvilinear relationship, with a substantially higher estimated Bmax (37.75 ± 3.69%) and Kd (77.66 ±12.78 µg mL −1 ); concentrations of HDL (5-80 µg mL −1 ) significantly increased [ 3 H]cholesterol efflux above control. The levels of mRNA for ABCA1, which effluxes cholesterol to apoA-I, and of ABCG1, which effluxes to HDL, were measured: gene expression (2 −∆Ct ; mean ±SEM) for ABCA1, relative to PPIA, was 2.01 × 10 −5 ± 1.25 × 10 −5 (n = 3), while levels of ABCG1 relative to PPIA were 4.77 × 10 −5 ± 3.6 × 10 −5 (n = 3); equivalent primer efficiencies were noted here (102% and 97%, respectively). These values are not significantly different, suggesting that the activity and stability of these proteins contribute to the efflux outcomes observed. Over time, [ 3 H]cholesterol efflux to apoA-I (20 µg mL −1 ) and HDL (20 µg mL −1 ) adhered strictly to linearity ( Figure 1C) with highly significant (p < 0.001) differences noted between the slope of each curve.  (A,B), respectively. The data were analysed using repeated measures ANOVA with post hoc test (Dunnett's) compared to the 0 μg mL −1 measurement; *p < 0.05, **p < 0.01 and ***p < 0.001. Values are the mean ± SD of three independent experiments (each performed in triplicate). The time dependency of [ 3 H]cholesterol efflux under basal conditions, and to apoA-I (20 μg mL −1 ) and HDL (20 μg mL −1 ) is shown in (C). The data was analysed using two-way ANOVA with post hoc test (Dunnett's) compared to the 0h timepoint; * p < 0.05, *** p < 0.001. Values are the mean ± SD of between three and four independent experiments (each performed in triplicate). Standard deviations, instead of SEM, have been shown, as the SEM for (B,C) are too small to be viewed.  (A,B), respectively. The data were analysed using repeated measures ANOVA with post hoc test (Dunnett's) compared to the 0 µg mL −1 measurement; * p < 0.05, ** p < 0.01 and *** p < 0.001. Values are the mean ± SD of three independent experiments (each performed in triplicate). The time dependency of [ 3 H]cholesterol efflux under basal conditions, and to apoA-I (20 µg mL −1 ) and HDL (20 µg mL −1 ) is shown in (C). The data was analysed using two-way ANOVA with post hoc test (Dunnett's) compared to the 0h timepoint; * p < 0.05, *** p < 0.001. Values are the mean ± SD of between three and four independent experiments (each performed in triplicate). Standard deviations, instead of SEM, have been shown, as the SEM for (B,C) are too small to be viewed.

Glucolipotoxicity in 1.1B4 Cells
The impact of treatment with high glucose (30 mM) combined with a palmitic acid (PA)/BSA complex (0.1 mM to 0.5mM PA) on the viability of 1.1B4 cells over 24 h is shown in Figure 2A. No significant changes in viability due to any of these treatments were noted at 0 h, as judged by conversion of MTT to formazan (data not shown); however, significant reductions in viability were noted after 24 h at concentrations of 0.25 mM PA (43.4%; p < 0.05) and 0.5 mM PA (72.6%; p < 0.01), compared with the control incubation. The effect of high glucose in combination with 0.25 mM PA on 1.1B4 cell viability was also confirmed via cellular reduction of resazurin to fluorescent resorufin (47%; p < 0.001) ( Figure 2B). Addition of apoA-I at concentrations of 40 and 80 µg mL −1 (8 µM to 16 µM) significantly increased cell viability, as judged by conversion of MTT to formazan, by 1.15-fold (p < 0.05) and 1.19-fold (p < 0.01), respectively. However, delivery of ApoA-I at 20-80 µg mL −1 did not protect against glucotoxicity enhanced by 0.25 mM PA ( Figure 2C). By contrast, the addition of HDL at 40 and 80 µg mL −1 (~0.05 mM to 0.1 mM HDL-cholesterol) increased conversion of MTT to formazan by 1.68-fold (p < 0.05) and 2.33 fold (p < 0.001), compared with the GLT challenge alone ( Figure 2D).

Glucolipotoxicity in 1.1B4 Cells
The impact of treatment with high glucose (30 mM) combined with a palmitic acid (PA)/BSA complex (0.1 mM to 0.5mM PA) on the viability of 1.1B4 cells over 24 h is shown in Figure 2A. No significant changes in viability due to any of these treatments were noted at 0 h, as judged by conversion of MTT to formazan (data not shown); however, significant reductions in viability were noted after 24 h at concentrations of 0.25 mM PA (43.4%; p < 0.05) and 0.5 mM PA (72.6%; p < 0.01), compared with the control incubation. The effect of high glucose in combination with 0.25 mM PA on 1.1B4 cell viability was also confirmed via cellular reduction of resazurin to fluorescent resorufin (47%; p < 0.001) ( Figure 2B). Addition of apoA-I at concentrations of 40 and 80 μg mL −1 (8 μM to 16 μM) significantly increased cell viability, as judged by conversion of MTT to formazan, by 1.15-fold (p < 0.05) and 1.19-fold (p < 0.01), respectively. However, delivery of ApoA-I at 20-80 μg mL −1 did not protect against glucotoxicity enhanced by 0.25 mM PA ( Figure 2C). By contrast, the addition of HDL at 40 and 80 μg mL −1 (~0.05 mM to 0.1 mM HDL-cholesterol) increased conversion of MTT to formazan by 1.68-fold (p < 0.05) and 2.33 fold (p < 0.001), compared with the GLT challenge alone ( Figure 2D). ; values are the mean ± S.D. of eight independent replicates in a single experiment; the data were analysed by Students' t-test; *** p < 0.001. The impact of apoA-I (0-80 μg mL −1 ) on the viability of 1.1B4 cells under GLT challenge is shown in (C) (mean ± SEM; n = 3); the data were analysed by two-way ANOVA with post hoc test (Dunnett's) compared to the basal group in the absence of apoA-I * p < 0.05, *** p < 0.001. The effect of HDL (0-80 μg mL −1 ) under the same conditions is shown in (D) (mean ± SEM; n = 3). The data were analysed by two-way ANOVA with post hoc test (Dunnett's) compared to the basal group in the absence of HDL. ### p < 0.001 compared with the GLT challenge alone, or ***p < 0.001 compared with the GLT challenge alone. ; values are the mean ± S.D. of eight independent replicates in a single experiment; the data were analysed by Students' t-test; *** p < 0.001. The impact of apoA-I (0-80 µg mL −1 ) on the viability of 1.1B4 cells under GLT challenge is shown in (C) (mean ± SEM; n = 3); the data were analysed by two-way ANOVA with post hoc test (Dunnett's) compared to the basal group in the absence of apoA-I * p < 0.05, *** p < 0.001. The effect of HDL (0-80 µg mL −1 ) under the same conditions is shown in (D) (mean ± SEM; n = 3). The data were analysed by two-way ANOVA with post hoc test (Dunnett's) compared to the basal group in the absence of HDL. ### p < 0.001 compared with the GLT challenge alone, or *** p < 0.001 compared with the GLT challenge alone.
Further, in the more sensitive resorufin cell viability assay, 20 µg protein mL −1 HDL (~0.025 mM HDL-cholesterol) significantly improved cell viability by 1.38-fold (p < 0.01) under the same conditions ( Figure 3A). Accordingly, this lowest concentration of HDL was tested for its impact on cellular apoptosis ( Figure 3B): analysis by two-way ANOVA indicate a significant effect of time in all four conditions tested (p < 0.001) and no significant overall effect of HDL. There was a significant interaction between time and HDL treatment (p < 0.001), with HDL significantly reducing apoptosis due to GLT challenge, under serum-free conditions.
ANOVA with post hoc test (Dunnett's) compared to the basal group in the absence of HDL. ### p < 0.001 compared with the GLT challenge alone, or *** p < 0.001 compared with the GLT challenge alone.

Figure 5.
Levels of miR-21-5p relative to RPL13A in 1.1B4 cells, after treatment (24 h) with apoA-I (20 µ g mL −1 ) and HDL (20 µ g mL −1 ) are shown in (A); values are mean ± SEM of 5 independent experiments, analysed by one-way ANOVA with post hoc (Dunnett's) test; * p < 0.05. Gene expressions of the miR-21-5p targets indicated (relative to PPIA), after 24 h treatment with HDL (20 µ g mL −1 ) are shown in (B); values are mean ± SEM of five to eight independent experiments, analysed by Student's t-test for the comparators indicated; * p < 0.05. Expression of hsa-miR-21-5p relative to RPL13A in 1.1B4 cells after transfection with mimic and inhibitor, together with the indicated gene targets, relative to PPIA, are shown in (C,D), respectively; values are the mean ± SEM of five to seven independent experiments. Statistical analysis was carried out by Students' t-test for the comparators indicated; * p < 0.05 and ** p < 0.01.

Hsa-miR-21-5p and Glucolipotoxicity in 1.1B4 Cells
The effect of the miR-21-5p mimic and inhibitor on the viability of 1.1B4 cells, in the presence and absence of HDL, is shown in Figure 6. Transfection with the control, mimic or inhibitor did not alter cell viability compared with the wild type cells; neither the mimic ( Figure 6A) or the inhibitor ( Figure 6B) could alter the impact of the GLT challenge on the viability of 1.14 cells, compared with the relevant control. The effect of the miR-21-5p inhibitor, in the presence of HDL in the GLT challenge, is shown in Figure 6C; the protective effect of HDL was sustained under transfection conditions, but was not altered by the presence of the miR-21-5p inhibitor, despite the established anti-apoptotic effects of this miRNA sequence [28]. ; values are mean ± SEM of five to eight independent experiments, analysed by Student's t-test for the comparators indicated; * p < 0.05. Expression of hsa-miR-21-5p relative to RPL13A in 1.1B4 cells after transfection with mimic and inhibitor, together with the indicated gene targets, relative to PPIA, are shown in (C,D), respectively; values are the mean ± SEM of five to seven independent experiments. Statistical analysis was carried out by Students' t-test for the comparators indicated; * p < 0.05 and ** p < 0.01.

Hsa-miR-21-5p and Glucolipotoxicity in 1.1B4 Cells
The effect of the miR-21-5p mimic and inhibitor on the viability of 1.1B4 cells, in the presence and absence of HDL, is shown in Figure 6. Transfection with the control, mimic or inhibitor did not alter cell viability compared with the wild type cells; neither the mimic ( Figure 6A) or the inhibitor ( Figure 6B) could alter the impact of the GLT challenge on the viability of 1.14 cells, compared with the relevant control. The effect of the miR-21-5p inhibitor, in the presence of HDL in the GLT challenge, is shown in Figure 6C; the protective effect of HDL was sustained under transfection conditions, but was not altered by the presence of the miR-21-5p inhibitor, despite the established anti-apoptotic effects of this miRNA sequence [28].  (A,B), respectively. This showed that while there was a significant effect of GLT challenge on cell viability in both figures, neither the miR-21-5p mimic or inhibitor modulated the impact of GLT change on the 1.1B4 cells; values are mean ± SEM of three independent experiments, analysed by two-way ANOVA and post hoc (Dunnett's) test *** p < 0.001. The effect of the miR-21-5p inhibitor in the presence of HDL (20 µg mL −1 ) is shown in (C). The GLT challenge adversely affected cell viability and there was a significant protective effect of HDL; mean ± SEM of four independent experiments, analysed by three-way ANOVA and post hoc (Dunnett's) test; * p < 0.05 and *** p < 0.001.

Discussion
This study demonstrates the protective effects of HDL, at concentrations below those required for transporter saturation but which sustain linear cholesterol efflux, in 1.1B4 cells subject to a GLT challenge. These well-characterised conditions induce a network of changes in miRNA sequences, as determined by microchip array, including increased expression of miR-21-5p which mediates anti-apoptotic effects [28]. Delivery of a miR-21-5p mimic replicated the impact of HDL on gene expression of STAT3 (signal transducer and activator of transcription 3) and SMAD7 (decapentaplegic protein 3), two established targets of this sequence; no changes in expression of STAT3 and SMAD7 were induced by delivery of miR-21-5p inhibitor. However, neither the mimic nor the inhibitor, the latter in the presence or absence of HDL, could significantly change the response of 1.1B4 cells to a GLT challenge. The data suggest that the complex epigenetic changes induced by HDL in human beta cells is unlikely to be replicated by a single miRNA sequence, supporting the emerging concept that global changes in miRNA networks may be more biologically relevant than changes in a single sequence [55].
Exposure to HDL has previously been linked with epigenetic changes in microRNA in a number of differing cell types [56], but this is the first report of the impact of these (apo)lipoproteins in an in vitro human beta cell model; equally, a number of microRNA sequences have been linked with (dys)function of beta cells [28]. Examination of these literature, compared with the microchip array reported here (Figure 4) reveal some common sequences [28,47,48,[55][56][57][58]. For example, miR-21 has been shown to target Snc2, decreasing Vamp2 and GSIS, but also to significantly modulate apoptosis in a range of rodent and human islet cell lines and tissues [28,48,49]. Mir-29 has been linked with increases in insulin exocytosis, but also with increased apoptosis via reductions in Mcl1 [28]; miR-130 and miR-335 exert opposing actions on GSIS, while miR-146a is thought to increase apoptosis by increased expression of c-Jun [28,[47][48][49]. However, treatment with HDL also alters expression of other sequences, not previously linked to maintenance of beta cell function [55,59]. Some have no function ascribed to them at present (hsa-miR-15b-5p, hsa-miR-181a-5p, hsa-miR-426-5p) while hsa-miR-16-5p was found to be down-regulated after bariatric surgery in obese patients [49], and hsa-miR-191-5p emerged as differentially regulated in Alzheimer's disease and major depressive disorder [58].
Delivery of the miR-21-5p mimic significantly altered the level of this sequence in 1.1B4 cells, to a similar extent as HDL treatment, exerting comparable reductions in gene expression of its established target genes STAT3 and SMAD7, indicating that the mimic was effective in this cell type. High density lipoprotein has previously been linked with activation of the pro-survival STAT3 signalling pathway [59,60], which protects against inflammation and apoptosis in many cell types [59][60][61][62][63]. However, Zeng et al. (2017) found that inhibition of expression and phosphorylation of STAT3 in hepatic cells restored insulin sensitivity in mice fed a high fat diet [64], while inhibition of the STAT3 pathway has also been linked with reductions in murine arterial atherosclerosis [65]. Repression of SMAD7 has not previously been reported after HDL treatment, but this protein is a wellestablished inhibitor of the TGFβ signalling pathway which regulates fibrosis, inflammation and wound healing and cancer, and loss of Smad7 in mice is associated with enhanced cell proliferation and reduced apoptosis [66]. Thus, it is clear that multiple molecular mechanisms may underlie the overall impact of HDL on apoptosis [28].
Indeed, the function of miR-21-5p in beta cells also needs careful consideration. Many studies indicate an anti-apoptotic role for this sequence in differing cell types and via a variety of mechanisms [42][43][44][45][46]. In particular, knockdown of miR-21 increases translation of proinflammatory tumour suppressor programmed cell death protein 4 (PDCD4), restricts cell proliferation and promotes apoptosis, while high levels of miR-21 are associated with cancer [42]. It is therefore possible that the elevation in miR-21-5p due to HDL may contribute to survival of 1.1B4 cells, via a mechanism which is not affected by gene repression of STAT3 but could be aided by reduction in expression of SMAD7 ( Figure 5). However, it is clear that the transient (and modest) increase in miR-21-5p levels due to the presence of the mimic alone was not sufficient to prevent the loss of viability due to GLT challenge, and nor could the inhibitor block the protective impact of HDL in this cell line under the same conditions ( Figure 6). Ruan et al. (2011) reported that elevated levels in miR-21, which induces a deficiency in PDCD4, can inhibit nuclear factor-kB (NF-κB) and protect beta-cells against streptozotocin-induced cell death, and spontaneous autoimmune diabetes in non-obese diabetic mice and C57BL7 mice [57,58]. Pro-inflammatory cytokines induce the expression of miR-21 [48], and Backe et al. (2014) suggested this is a protective response to preserve beta cell viability [47]. The feasibility of miR-21-5p as a pro-survival therapeutic agent, however, has been cast into doubt by the finding that stable (long-term) expression of miR-21-5p not only increased the proliferation of rat INS-1 insulinoma cells, but also stimulated nitric oxide synthesis and apoptosis [28,64].
Analysis by miRPath established numerous KEGG pathways associated with miRNAs differentially expressed in 1.1 beta cells following exposure to HDL (Figure 7). Three pathways that were linked to mRNAs down-or upregulated by HDL were fatty acid biosynthesis, fatty acid metabolism and lysine degradation, previously linked to the role of HDL in reverse cholesterol transport [67]. Additional pathways, including the Hippo signalling pathway, extracellular matrix (ECM)-receptor interactions and thyroid hormone signalling suggest a role for HDL in regulating developmental processes in beta cells. It is understood that miRNAs are involved in maturation of beta cells and maintenance of beta cell identity with miR-7, miR-9, miR-375, miR-376, miR-17-5p and miR-29b-3p examined in this context [68,69]. In conjunction with identification of adherens junction, cell cycle and p53 signalling pathways, this suggests that HDL modulate pathways in beta cells that control responses to stress, cell proliferation and apoptosis, supporting our observations that HDL protect 1.1B4 beta cells from damage due to GLT challenge. Two KEGG pathways uniquely linked to miRNAs upregulated by HDL were also associated with miR-21-5p ( Figure 8): fatty acid elongation and the FoxO signalling pathway. Notably, the FoxO signalling pathway integrates a number of signalling pathways such as TGF-beta, PI3K-Akt and Jak-STAT, to regulate functions including cell metabolism, cell cycle regulation and apoptosis [70,71]. Interaction networks between miRNAs differentially expressed following 20 µg mL −1 HDL treatment and their associated pathways were produced in Cytoscape v3.8.0. Ten downregulated miRNAs and 4 upregulated miRNAs were used for analysis. The interaction networks depicting miRNA-pathway connections for downregulated and upregulated miRNAs are presented in (A,B), respectively. The networks are annotated so that the size of the pathway labels are proportional to the number of genes in the network associated with the relevant miRNAs and the thickness of the edges corresponds to the proportion of the total miRNA:gene interactions in each pathway.
If we consider the limitations of this study, there are several points to note. One key point is that the study is restricted to use of the 1.1B4 cell line (with the associated drawbacks discussed above): replication in primary human islet beta cells would provide greater impact to the findings. Further, the use of metabolic activity assays (MTT and resorufin conversion) as measures of cell viability, and of caspase 3/7 activity as a measure of apoptosis, do not provide insight into the mechanism(s) by which HDL protect the 1.1B4 cells from a GLT challenge. Finally, the analysis of microRNA sequences altered by HDL treatment, compared with the control, derive from single samples using three technical replicates on each microchip, albeit validated in three independent experiments by Q-PCR.

Conclusions
This study observed that low concentrations of HDL protect 1.1B4 beta cells from the reduction in cell viability, and increased apoptosis, induced by challenge with 30 mM glucose and 0.25 mM palmitic acid (GLT). Analysis by microchip array following exposure to HDL demonstrated differentially expressed miRNAs which were associated with KEGG pathways linked with beta cell development, cell proliferation and apoptosis. Hsa-miR-21-5p upregulation by HDL was validated by qPCR and shown to lead to downregulation of SMAD7 and STAT3. Modulation of miR-21-5p by mimics and inhibitors also reduced SMAD7 and STAT3 expression, however, this did not affect the deleterious effects of GLT challenge on the 1.1B4 beta cells. We suggest that examination of the networks of miRNA pathways regulated by HDL is necessary to understand the epigenetic regulation required to simulate HDL protection of beta cells in developing future therapeutics.