Ex Vivo Expansion of Murine MSC Impairs Transcription Factor-Induced Differentiation into Pancreatic β-Cells

Combinatorial gene and cell therapy as a means of generating surrogate β-cells has been investigated for the treatment of type 1 diabetes (T1D) for a number of years with varying success. One of the limitations of current cell therapies for T1D is the inability to generate sufficient quantities of functional transplantable insulin-producing cells. Due to their impressive immunomodulatory properties, in addition to their ease of expansion and genetic modification ex vivo, mesenchymal stem cells (MSCs) are an attractive alternative source of adult stem cells for regenerative medicine. To overcome the aforementioned limitation of current therapies, we assessed the utility of ex vivo expanded bone marrow-derived murine MSCs for their persistence in immune-competent and immune-deficient animal models and their ability to differentiate into surrogate β-cells. CD45−/Ly6+ murine MSCs were isolated from the bone marrow of nonobese diabetic (NOD) mice and nucleofected to express the bioluminescent protein, Firefly luciferase (Luc2). The persistence of a subcutaneous (s.c.) transplant of Luc2-expressing MSCs was assessed in immune-competent (NOD) (n = 4) and immune-deficient (NOD/Scid) (n = 4) animal models of diabetes. Luc2-expressing MSCs persisted for 2 and 12 weeks, respectively, in NOD and NOD/Scid mice. Ex vivo expanded MSCs were transduced with the HMD lentiviral vector (MOI = 10) to express furin-cleavable human insulin (INS-FUR) and murine NeuroD1 and Pdx1. This was followed by the characterization of pancreatic transdifferentiation via reverse transcriptase polymerase chain reaction (RT-PCR) and static and glucose-stimulated insulin secretion (GSIS). INS-FUR-expressing MSCs were assessed for their ability to reverse diabetes after transplantation into streptozotocin- (STZ-) diabetic NOD/Scid mice (n = 5). Transduced MSCs did not undergo pancreatic transdifferentiation, as determined by RT-PCR analyses, lacked glucose responsiveness, and upon transplantation did not reverse diabetes. The data suggest that ex vivo expanded MSCs lose their multipotent differentiation potential and may be more useful as gene therapy targets prior to expansion.


Introduction
T1D results from the autoimmune destruction of the pancreatic insulin-producing β-cells, which leads to hyperglycaemia and the lifelong dependence on exogenous insulin therapy [1]. Currently, the only cures for T1D are pancreas or islet transplantation; however, these interventions are limited by a shortage of donor organs and the requirement for lifelong immunosuppression [2]. To overcome the limitations of current therapies, a promising alternative strategy is the ex vivo generation of surrogate β-cells through the directed-differentiation of nonpancreatic target cells [3][4][5][6][7][8][9][10].
Pancreatic transcription factors play an important role both in islet cell differentiation and specialization and mature β-cell function during embryonic and neonatal development and adult life, respectively [11,12]. Our laboratory and others have investigated the direct transfer of β-cell transcription factors and insulin as mediators of pancreatic transdifferentiation in nonpancreatic cells/tissues with varying success [3][4][5][13][14][15][16][17]. We have previously shown that the endocrine-specifying transcription factor, NeuroD1, which lies downstream of Pdx1 in the transcription factor hierarchy of pancreatic development, was capable of inducing pancreatic transdifferentiation of a rat hepatocyte cell line (H4IIE). Transdifferentiation was characterized by the upregulation of both upper and lower hierarchy pancreatic transcription factors, without the development of exocrine differentiation [12,18]. In addition, due to the overexpression of the furin-cleavable human insulin (INS-FUR) gene (a modified form of human proinsulin, which permits cleavage into mature insulin via furin enzymes in nonpancreatic cells), these cells were capable of synthesizing, storing, and secreting mature human insulin in a glucose-responsive manner and reversed diabetes upon transplantation in STZ-diabetic NOD/Scid mice [18]. However, one of the current challenges of clinical translation of combinatorial gene and cell therapies for T1D is upscaling the production of functional surrogate β-cells [19].
Due to their high plasticity, immunomodulatory properties, fewer ethical concerns, and ease of ex vivo expansion and gene modification [20,21], MSCs are an attractive alternative target cell for the autologous and allogeneic treatment of T1D. Several studies have investigated the ex vivo targeting of MSCs for transdifferentiation into islet progenitor cells (IPCs) via viral-mediated transfer of pancreatic transcription factors [14][15][16][17]. Previously, the transfer of the "master regulator" of pancreatic differentiation, Pdx1, to MSCs resulted in their differentiation into glucose-responsive IPCs, which reversed diabetes, for a period of 6-8 weeks (experimental endpoint), upon transplantation into STZ-diabetic NOD/Scid mice [14]. However, Pdx1 transfer has also been associated with exocrine differentiation and concomitant tissue damage, which is undesirable for a T1D cell replacement therapy [22]. Therefore, in this study, we assessed the pancreatic differentiation potential of ex vivo expanded murine bone marrow-derived MSCs as a preclinical model to overcome the shortage limitations of current therapies, via the overexpression of murine NeuroD1 and INS-FUR using a lentiviral vector. We found that due to a loss of the intrinsic multipotent differentiation potential of MSCs with increasing culture, transcription factor-mediated β-cell differentiation, via the forced expression of NeuroD1 and INS-FUR, failed to occur. This was confirmed via the overexpression of murine Pdx1, which is known to induce β-cell differentiation of MSCs at early passage numbers. The data highlight the limited timeframe for MSCs to function as effective gene therapy targets and suggest that MSCs do not represent a suitable alternative source of cells to overcome the shortage limitations of current β-cell replacement therapies.

Methods
2.1. Sourcing of Animals. NOD and NOD/Scid mice were sourced from the Animal Resources Centre (WA, Australia). All animal work was approved by the UTS Animal Care and Ethics Committee (ACEC 2011-447A; ACEC 2009-244A) and complied with the Australian code for the care and use of animals for scientific purposes [23].

MSC Proliferation and Clonogenicity.
For proliferation assays, MSCs at early (P3-15), mid (P15-30), and late (P30-60) passage numbers were seeded in 24-well plates (2.5 × 10 3 cells/well) (Falcon ® BD Biosciences, San Jose, USA) in triplicate and maintained in standard MSC medium for 15 days, with the medium replenished weekly. Cell viability was assessed by Trypan blue (0.4% v/v; Gibco ® , Thermo Fisher) exclusion. Total cell and viable cell numbers were determined and represented as mean ± standard deviation (SD) for each time point (n = 3).
For clonogenicity assays, MSCs at early, mid, and late passage numbers were seeded in 10 cm 2 tissue culture-treated plates (5 × 10 2 cells/plate) (Falcon ® BD Biosciences) and maintained in standard MSC medium for 10 days. Colonies were stained with 0.4% v/v methylene blue in methanol and counted by microscopy. Data were represented as mean colony count per 5000 cells ± SD (n = 3). Standard MSC medium was replenished weekly.

Morphological Analysis.
Images of four fields of view at 10x or 20x magnification were acquired at early, mid, and late passage numbers using a Leica © DM microscope (Leica Microsystems © , Wetzlar, Germany) and processed using the image processing software, Leica Application Suite (V4.4.0) (Leica Microsystems © ). Scale bars on figures are equivalent to 100 μm.
2.6. Differentiation Assays 2.6.1. Adipogenesis. Early, mid, and late passage number cells were seeded in standard MSC medium in 24-well plates (2.5 × 10 4 cells/well) in triplicate and grown to 80-90% confluence. The medium was subsequently replenished with either adipogenic control or differentiation medium, as previously described [24]. The cells were stained with 0.2% w/v Oil Red O in methanol (Fronine ® , Sydney, Australia) and semiquantitatively scored as previously described [24]. Values were expressed as count per cm 2 and were represented as means ± SDs (n = 3).
2.6.2. Osteogenesis. Early, mid, and late passage cells were seeded in standard MSC medium in 24-well plates (1.25 × 10 4 cells/well) in triplicate and grown to 90-95% confluence. The medium was subsequently replenished with either osteogenic control or differentiation medium, as previously described [24]. The cells were stained with 2% w/v Alizarin Red S (pH 4.1) (Fronine ® ) and semiquantitatively scored, as previously described [24]. Values were expressed as count per cm 2 and were represented as means ± SDs (n = 3).

Induction of Diabetes in NOD/Scid
Mice. NOD/Scid mice received 170 mg/kg of STZ in 0.1 M sodium citrate buffer (pH 4.0) via i.p. injection. All animals, including nondiabetic controls, had their body weights and blood glucose concentrations measured daily using an Accu-Chek ® Performa glucometer (Accu-Chek ® , Roche, Castle Hill, Australia). Animals that did not develop hyperglycaemia (blood glucose concentration >8 mmol/l) within 1 week post-STZ injection received a second low-dose (40 mg/kg) STZ injection. Animals that displayed hyperglycaemia for four consecutive days were considered diabetic and were used for in vivo experiments.
2.18. Statistical Analysis. All statistical analysis was performed using GraphPad Prism 7 ® software. Values were presented as means ± SDs or SEMs. One-way or two-way ANOVA, with the appropriate posttests, was performed, with p < 0 05 indicating significance.

In Vitro Characteristics of NOD MSCs.
MSCs identified by FACS were characterized by the surface marker profile CD45 -/Ly6 + and constituted~33% of the parental stromal cell population (Figure 1(a)). MSCs displayed the characteristic fibroblast-like morphology from early to late passage numbers (Figure 1(b)). However, a decrease in MSC diameter was observed with increasing passage, from~100 μm (early passage) to 50 μm (late passage). Although MSCs underwent a period of early passage replicative crisis during P5-8 (data not shown), MSCs continued to self-renew up to 60 passages (maximum culture period). An intrapopulation analysis of MSC proliferation and clonogenicity showed no significant difference in proliferation (Figure 1(c)) and conservation of clonogenicity potency (Figure 1(d)) from early to late passage numbers. To demonstrate that NOD MSCs underwent trilineage differentiation, as defined by the International Society Cell and Gene Therapy (ISCT), trilineage differentiation assays were performed at early, mid, and late passage numbers. NOD MSC demonstrated trilineage differentiation into adipocytes (Supplementary Figure 2), osteocytes (Supplementary Figure 3), and chondrocytes (Supplementary Figure 4). Semiquantitative analysis of adipogenesis and osteogenesis was assessed by scoring the degree of differentiation, as previously described [24]. MSCs displayed reduced adipogenesis (Figure 1(e)) and osteogenesis (Figure 1(f)) with increasing passage number.
To confirm that MSCs did not intrinsically express any pancreatic transcription factors or hormones, RT-PCR analyses were performed and showed a lack of expression of pancreatic transcription factors, hormones, and proteins in all adherent bone marrow cell populations (Figure 1(g)).
As expected, all genes were expressed in normal mouse pancreas (positive control), and FoxA2, Scl2a2, and Gck were expressed in normal mouse liver (negative control).

Syngeneic MSCs Are Cleared in an Immune-Competent
Animal Model. Noninvasive BLI is an established and sensitive tool for assessing cell replacement therapy safety and efficacy in living preclinical small animal models. Furthermore, preclinical BLI results often serve as the decision point of the suitability of a cell replacement therapy for clinical trial testing in humans. This study utilized the Firefly luciferase reporter gene, Luc2, a Luc2-specific light-producing substrate D-luciferin, and an IVIS Lumina II imaging system (Perkin Elmer © ). Prior to performing in vitro and in vivo BLI, a clonal population of MSCs expressing Luc2 (MSC-Luc2) was obtained by selection with Hygromycin B. MSC-Luc2 retained a fibroblast-like morphology similar to that observed for parental MSCs (Figure 2(a)); however, these cells exhibited a reduced cell diameter (~100 μm nucleofected versus 150 μm parental). In vitro analyses of BLI at multiple time points over a 3-hour period (Figure 2(b)), in combination with linear regression analysis (Figure 2(c)), confirmed that a clonal population of MSC-Luc2 had been selected and that bioluminescence was stable for up to 3 h in vitro.  We chose to assess the persistence of MSCs transplanted subcutaneously in immune-competent NOD and immune-deficient NOD/Scid animal models as this most closely reflects the route of administration of a cell replacement therapy for individuals with T1D. Thus, MSC-Luc2 were transplanted subcutaneously at multiple cell concentrations to determine the lowest cell concentration and the length of time for which bioluminescence could be detected (Figure 3(a)). Quantitative analysis of BLI data showed that in both NOD/Scid and NOD mice there was an increase in bioluminescence with increasing cell dose (Figure 3(b)), which resulted in a dose-dependent increase in persistence of bioluminescence in both animal models (Supplementary Tables 2A and 2B). Bioluminescence, albeit diminished, could be detected in NOD/Scid mice for up to 12 weeks, suggesting poor survival of MSCs at the s.c. transplant site. By comparison, bioluminescence persisted for 2 weeks in NOD mice, after which signal could no longer be detected, suggesting an immune-mediated clearing of the MSC graft. In fact, upon challenge with a follow-up injection of 1 × 10 6 MSCs, clearing of the MSC graft occurred within 1 week postinjection (data not shown). These kinetics are consistent with the generation of memory T cell populations stimulated after the initial exposure to MSCs.  Figure 3: Persistence of syngeneic MSCs in immune-competent and immune-deficient animal models. (a) NOD (n = 4) and NOD/Scid (n = 4) mice (6-10 weeks of age) received a total of six subcutaneous (s.c.) injections of 1 × 10 4 (n = 2), 1 × 10 5 (n = 2), and 1 × 10 6 (n = 2) midpassage MSC-Luc2 cells/mouse. Untreated age-matched NOD (n = 2) and NOD/Scid (n = 2) mice were utilized as negative controls. BLI images were acquired following i.p. administration of D-luciferin (15 mg/ml) at 150 mg/kg or 10 μl/g. Images are representative of a single experimental NOD/Scid and NOD animal. (b) Analysis of MSC BLI in NOD and NOD/Scid mice. Regions of interest were established surrounding the areas corresponding to the sites of cell transplantation using the Living Image 3.1 (PerkinElmer™) software. Quantitative data were subsequently analyzed using GraphPad Prism 7 ® . Data were presented as mean radiance ± SEM over time (weeks). A two-way ANOVA and Tukey's posttests were performed, * p < 0 05.

Neurod1 and Pdx1
Fail to Induce β-Cell Differentiation of Ex Vivo Expanded NOD MSCs. The pancreatic transcription factors, Neurod1 and Pdx1, were overexpressed in MSC-Luc2 to function as mediators of pancreatic transdifferentiation, whilst the INS-FUR gene was overexpressed to allow for mature human insulin production. Transduced MSCs were analysed via FACS and sorted into individual populations, as outlined in Supplementary Tables 3A and 3B. The sorted MSCs were returned to culture and imaged for eGFP and mCherry expression 7 days posttransduction (Figures 4(a)  and 4(b)). As can be seen, eGFP expression in MSCs transduced with Neurod1 and INS-FUR/Neurod1 was lower than that in cells transduced with the existing HMD and HMD-INS-FUR lentiviral vectors, a consequence of lower viral titers.
Transduced MSCs were subsequently cultured for a period of 28 days, at which point morphological analysis of the differentiation process was performed (Figure 4(c)). Transduced MSCs expressing both murine Neurod1 and Pdx1 alone, or in combination with INS-FUR, retained a fibroblast-like morphology in comparison to their untransduced counterparts, demonstrating that no change in morphology was attributable to transgene overexpression.
Gene expression profiling was performed to determine if ectopic expression of Neurod1, Pdx1, and INS-FUR in ex vivo expanded MSCs resulted in pancreatic differentiation. As can be seen in Figure 4 Figure 4(e), a similar pattern of gene expression was observed following the ectopic expression of murine Pdx1, as was observed with ectopic expression of Neurod1, with the exception that ectopic Pdx1 did not result in endogenous Pdx1 expression. Together, these data confirm the lack of pancreatic differentiation in ex vivo expanded MSC expressing Neurod1 and Pdx1.

Gene-Modified Ex Vivo Expanded NOD MSCs Demonstrate
Abnormal Glucose-Stimulated Insulin Secretion. The ability of transduced MSCs to secrete mature insulin in vitro in response to glucose stimulation was determined. A significant quantity of mature insulin (~4.5 pmol/ml/1 × 10 5 cells) was detected in the medium of MSCs expressing INS-FUR alone and in combination with Pdx1. By comparison, untransduced MSCs and MSCs transduced with empty vector or Neurod1 alone did not secrete mature insulin (Figure 4(f)). There was also a significant difference (p < 0 005) in the quantity of insulin secreted between MSC-INS-FUR (~4.5 pmol/ml/1 × 10 5 cells) and MSC-INS-FUR/Neurod1 (~0.3 pmol/ml/1 × 10 5 cells), which directly correlated with differences in the expression of the fluorescent reporter. Acute glucose stimulation assays showed that following stimulation with 20 mM D-Glucose, glucose-stimulated insulin secretion (GSIS) was not present (Figure 4(g)). The absence of GSIS was expected due to the lack of Slc2a2 expression, as detected by RT-PCR.

INS-FUR-Expressing MSCs Fail to Restore
Normoglycaemia in STZ-Diabetic NOD/Scid Mice. To determine if INS-FUR-expressing MSCs could restore normoglycaemia in STZ-diabetic NOD/Scid mice, animals received a s.c. transplant of either 1 × 10 7 or 5 × 10 7 cells. In animals treated with 5 × 10 7 cells, within 24 h following transplantation, there was a decrease in blood glucose concentrations, and a significant (p < 0 05) decrease was observed at days 8 and 12 posttransplantation (Figure 5(a)). In addition, there was a significant decrease in the blood glucose levels of the animals treated with 5 × 10 7 cells, from before to after transplantation, for a period of~15 days (pretransplant vs. posttransplant, p < 0 05) (Supplementary Tables 4A and  4B). By comparison, blood glucose concentrations in diabetic animals and animals treated with 1 × 10 7 cells remained significantly higher (p < 0 0001) than those in the normal controls for the duration of the experiment. Most importantly, animals treated with either 1 × 10 7 or 5 × 10 7 cells did not normalize blood glucose concentration at any time for the duration of the experiment. In addition, there was a significant difference in the body weights of treated animals in comparison to normal controls both before and after transplantation, despite random allocation of animals to the different groups ( Figure 5(b)). However, at no time point was there a significant decrease in body weight observed over the time course of the experiment.
Prior to euthanasia, animals treated with 5 × 10 7 cells (n = 3) were assessed for glucose tolerance via an IPGTT. Treated animals displayed an abnormal glucose tolerance in comparison to normal controls ( Figure 5(c)). The high degree of variation in the glucose tolerance observed for the treatment group was due to one treated animal beginning the IPGTT at a significantly lower blood glucose concentration than the other treated animals, a consequence of the cell transplant having successfully reduced blood glucose levels in this animal.

Discussion
Stem cells are characterized by two defining long-term characteristics: (i) the ability for renewal without differentiation into other cell types when cultured under standard conditions and (ii) the continued potential to develop into specialized cell types when cultured under defined experimental conditions. The early, mid, and late passage number NOD-derived MSCs utilized in this study fulfilled the ISCT criteria of plastic adherence, self-renewal, and colony formation without differentiation into other cell types under standard MSC culture conditions and trilineage mesenchymal differentiation into bone, fat, and cartilage under defined cell culture conditions [20,21].
In this study, NOD-derived MSCs maintained a fibroblast-like morphology, potent self-renewal, and clonogenicity throughout early, mid, and late passage numbers. This finding is in contrast to previous studies, which suggest that MSCs undergo age-related changes with continued cell passage [26][27][28]. The maintenance of a fibroblast-like morphology can be attributed, in part, to medium supplementation with bFGF. These results are supported by another study, where bFGF inhibited apoptosis and promoted proliferation of MSCs through a reduction in cellular oxidative stress [29]. Although MSC characteristics of self-renewal and colony-forming units (CFU-F) were conserved following cell culture expansion, MSCs showed some reduction in trilineage differentiation with increasing passage, which is in concordance with other studies [30][31][32]. Furthermore, several studies have demonstrated the detrimental effect of aging on multipotential differentiation, proliferation, and senescence [31,33,34]. Collectively, these findings highlight the importance of ongoing surveillance of stem cell-like properties, amongst other defining characteristics, when engineering replacement cell therapies.
To clarify whether NOD-derived MSCs were immune-evasive, we subcutaneously transplanted syngeneic MSCs into immune-competent NOD mice, which were detectable by BLI up to 14 days posttransplantation. In a similar study, MSCs that were transduced to coexpress luciferase and GFP, subsequently transplanted via the kidney artery in BALB/c mice, were detected via BLI up to 14 days posttransplantation [35]. This demonstrates that the site of cell administration in immune-competent models does not affect cell survival and that syngeneic MSCs do not appear to be immune-privileged [36]. This may be due to several factors, including the overexpression of the nonmammalian Luc2 transgene and contaminating FBS, factors that may override the inherent immune privilege characteristics of MSCs. In fact, previous studies have demonstrated that sustained high levels of luciferase expression induce luciferase-specific immune responses in immune-competent animal models, thereby limiting the utility of luciferase as an in vivo reporter in transplantation studies [37,38].
However, in immune-deficient NOD/Scid mice, syngeneic MSCs were detected for a significantly longer period (up to 12 weeks posttransplantation), albeit with diminishing persistence. These results are supported by similar studies, which show loss of transplanted MSCs in immune-deficient animal models [39][40][41]. The current study highlights the importance of reporter gene selection and the limited timeframe in which MSC therapeutic effects can be evaluated, in both immune-competent and immune-deficient animal models. Assessment of MSCs in short-term studies has demonstrated their protective benefits [42], whereas long-term studies show little or no protection [43], which may be attributable to their lack of persistence in immune-competent models.
We also assessed the ability to differentiate ex vivo expanded MSCs into surrogate β-cells. Considering the success of Neurod1 as a mediator of β-cell differentiation in nonpancreatic tissues, we sought to determine the potential of Neurod1 to induce β-cell differentiation of MSCs. In  [44]. Surprisingly, in this study, the lack of β-cell differentiation following overexpression of Pdx1 in culture-expanded MSCs indicated that ex vivo expansion results in defects in the pancreatic differentiation potential of MSCs.
A previous study showed that in vivo transplantation is required for functional β-cell maturation [45]; therefore, INS-FUR-expressing MSCs were assessed for their ability to reverse diabetes following transplantation into STZ-diabetic NOD/Scid mice. Upon transplantation of 5 × 10 7 INS-FUR-expressing MSCs, blood glucose concentrations decreased from 25-30 to 15-20 mmol/l. However, at no time point did blood glucose levels fall within the normal physiological range (5-7 mmol/l). This is likely due to the severe hyperglycaemia induced in these animals, which likely requires higher cell numbers to restore blood glucose levels to within the normal physiological range. Despite this, there was a significant . Data were presented as means ± SEMs (n = 6 mice per group). Two-way ANOVA with Sidak's posttests were performed, * p < 0 05. (b) Body weights of treated STZ-NOD/Scid mice. Body weight measurements were recorded daily post-STZ injection for the duration of the experiment (36 days). Data were presented as means ± SEMs (n = 6 mice per group). Two-way ANOVA was performed with Sidak's posttests, * p < 0 05. (c) IPGTT in treated STZ-diabetic NOD/Scid mice. Blood glucose measurements were obtained at 0, 5, 15, 30, 60, and 90 min post-D-Glucose injection. Data were presented as means ± SEMs (n = 5, normal; n = 3, treated). Two-way ANOVA with Sidak's posttests were performed, * p < 0 05. decrease in the blood glucose levels of treated vs. diabetic animals (p < 0 05) that was maintained for~2 weeks; after which, blood glucose concentrations began to increase to pretransplant values. The subsequent increase in blood glucose concentrations of treated animals correlated with the results of the MSC persistence studies in NOD/Scid mice. In addition, IPGTTs of transplanted animals showed that they exhibited abnormal glucose tolerance, indicating a lack of in vivo GSIS. This finding corroborated the in vitro characterization studies.
In conclusion, the results of this study highlight several caveats to MSC-based gene therapy for T1D, which warrant careful consideration prior to the formulation of clinical trials. Considering that overexpression of Pdx1 in early-passage MSCs results in pancreatic differentiation [14], these data show that ex vivo expansion impairs pancreatic differentiation of NOD-derived MSCs through the age-related loss of multipotency. Therefore, gene modification should be performed as soon as practicable after the isolation of MSCs [46]. In addition, given that ex vivo expansion is required to generate sufficient quantities of adult-derived MSCs for therapeutic purposes and that this process impairs their therapeutic potential, the use of embryonic stem cells or induced pluripotent stem cells as an unlimited stem cell source may overcome this limitation [47,48].

Data Availability
The data used to support the findings of this study are available from the corresponding author upon request.