Cotransplantation of Mesenchymal Stem Cells With Neonatal Porcine Islets Improve Graft Function in Diabetic Mice

Despite the initial optimism after the successes of the Edmonton Protocol in 2000 (1), many hurdles still prevent islet transplantation from replacing insulin as the gold standard treatment for patients with diabetes. However, despite early insulin independence, long-term graft attrition gradually reverts recipients to exogenous insulin dependency (2). Loss of islet graft function is partly due to a significant loss of β-cell mass in the first hours to days after infusion, mediated by a nonspecific inflammatory response characterized by proinflammatory cytokines (3). Cytokines are damaging to islet structure and function; coculture of islets and cytokines results in β-cell death and impairment of glucose-stimulated insulin secretion (GSIS) (4). A second contributing factor to islet death is the hypoxic conditions immediately posttransplantation. Pancreatic islets have a dense native capillary network, and β-cells receive 10–15 times more blood flow than the surrounding exocrine tissue (5), which amounts to ∼15–20% of the pancreatic blood supply despite comprising only 1–2% of the pancreas volume (6). During isolation and culture, this vasculature is destroyed, leaving the islets avascular before transplantation (5). Moreover, several days are needed to reestablish islet vascularization, thereby exposing islets to hypoxia (5).

The challenges of hypoxic and inflammatory-mediated islet destruction could be ameliorated by cotransplantation with mesenchymal stem cells (MSCs). MSCs are multipotent progenitor cells that can be isolated from a variety of tissues, including bone marrow, adipose tissue, Wharton’s jelly, umbilical cord blood, pancreas, and likely every adult tissue (7,8). MSCs have been reported to have anti-inflammatory, proangiogenic, and immunoregulatory effects (9,10) through the secretion of trophic factors such as hepatocyte growth factor, transforming growth factor β, interleukin-6, vascular endothelial growth factor (VEGF) (11), and annexin A1 (12). They may be capable of controlling inflammation within and surrounding the islet graft in the posttransplantation period as well as stimulating rapid graft vascularization. MSCs have already been demonstrated to modulate the diabetic milieu in humans. Carlsson et al. (13) reported preserved C-peptide secretion in patients with newly diagnosed type 1 diabetes given intravenous (IV) autologous bone marrow–derived MSCs. A second trial reported that patients with type 1 diabetes given a combination of allogenic umbilical- and autologous bone marrow–derived MSCs through the pancreatic artery had lower HbA_1c and fasting glycemia than patients who did not receive MSCs and required less exogenous insulin 1 year after MSC delivery (14). Of note, no severe adverse events were reported in either trial.

An additional challenge to islet transplantation becoming a widespread therapy is that cadaveric donor islets are in limited supply, and human islet quality can range greatly between donors. As previously reported, neonatal porcine islets (NPIs) may be a safe and effective alternative to human donor islets (15,16). These islets are in virtually unlimited supply, are easy to isolate, can proliferate after transplantation, and are resistant to hypoxia (17), proinflammatory cytokines (18), and hyperglycemia (19). Although the effect of rodent MSCs and islets have been examined in mouse transplant models (20–25), further studies are needed to examine the effects of clinically relevant human MSCs and β-cell sources, such as NPIs, on islet engraftment and functional outcome.

MSC human donor characteristics are known to have an effect on the therapeutic potential of MSCs (26). For example, MSCs from older donors have altered morphology and reduced proliferative ability, secretion of trophic factors, angiogenic potential, viability, and wound healing ability compared with younger donors (27–30). Sex-related differences also exist in trophic factor secretion (31). Few data exist on the effects of disease status of donors on the characteristics of MSCs, but some studies indicate that donors with autoimmune diseases have abnormal MSCs (26). Donors with certain autoimmune diseases have been shown to have MSCs with altered morphology (32), decreased proliferative capacity (32–34), altered gene expression (35), decreased expression of trophic factors (35,36), and reduced inhibitory effect on T cells (32,35). With the promise of novel therapies using MSCs for autoimmune diseases such as type 1 diabetes, learning more about the effect of autologous MSCs in the treatment of autoimmune diseases is critical.

In this clinically relevant study, we investigated the function of transplanted NPIs in the presence of human MSCs. We report that the cotransplantation of MSCs results in faster normalization, improved glucose tolerance, and improved early angiogenesis in a diabetic mouse model. We also report that MSCs from a donor with an autoimmune disease produced dramatically different outcomes when cotransplanted with NPIs.

Porcine pancreases were obtained from 1- to 3-day-old Duroc neonatal piglets from the University of Alberta Swine Research Centre (1.5–2.0 kg body weight), and NPIs were isolated and cultured for 5–7 days as previously described (16). Six independent NPI isolations were used for the transplant studies.

To prepare bone marrow–derived MSCs, human bone marrow was extracted from seven patients during orthopedic surgery (Division of Orthopedic Surgery, University of Alberta) after informed consent. For expansion, cells were plated in minimum essential medium α (Cellgro) supplemented with 2.5 ng/mL basic fibroblast growth factor (Millipore), 10% FBS (Gibco), 1 mmol/L sodium pyruvate (Gibco), 10 mmol/L HEPES (Gibco), and 100 units penicillin/1,000 units streptomycin (BioWhittaker) at a density of 166,000 cells/cm². Nonadherent cells were removed by changing the medium every 2–3 days. Once confluent, the cell monolayer was washed with Versene and detached with 0.05% volume for volume trypsin-EDTA (Invitrogen). Cells were counted and reseeded into supplemented minimum essential medium α culture at a density of 5,000–10,000 cells/cm² and underwent five passages before transplantation. As we previously reported (37), MSCs isolated with this protocol express by FACS analysis the classic MSC surface antigens CD29, CD44, CD73, CD90, and CD105.

Bone marrow MSCs were enzymatically detached from culture plates and counted, and 2 × 10⁶ cells were added to a 100-mm low-adherence culture dish (Corning) with 6,000 NPIs in a total volume of 10 mL culture media. Controls included 6,000 islets cultured alone. Cells were cultured for 48 h in DMEM low glucose (5.6 mmol/L) (Gibco) with 1% FBS, 20 ng/mL epidermal growth factor, 20 ng/mL basic fibroblast growth factor, 10 mmol/L HEPES, 100 units/mL penicillin, 0.1 mg/mL streptomycin, and 71.5 μmol/L β-mercaptoethanol.

To determine whether MSC coculture had an effect on the cellular composition or β-cell proliferation of the NPI grafts, immunostaining was performed by using previously published methods (15,16). Primary antibodies included insulin (1:1,000; Dako), glucagon (1:5,000; Sigma-Aldrich), proliferating cell nuclear antigen (PCNA) (1:300; Dako), and appropriate species-specific secondary antibodies Alexa Fluor 488 or 594 (1:200; Molecular Probes, Eugene, OR). Insulin, glucagon, and double insulin and PCNA-stained cells were quantified with the use of ImageJ software (http://rsbweb.nih.gov/ij). Before transplantation, total cellular insulin content of the grafts was determined, and a static incubation assay was used to assess GSIS (previously published methods [15,16]). RNA was also extracted from grafts (RNeasy Mini Kit; QIAGEN). cDNA was synthesized by using the High-Capacity Reverse Transcription Kit (Thermo Fisher Scientific), and relative quantification (RQ) was performed by using TaqMan Gene Expression Assays (Thermo Fisher Scientific) with SDS software on the ABI PRISM 7900HT. Validated primer sets were as follows: PDX1 (Ss03373351_m1), insulin (Ss03386682_u1), glucagon (Ss03384069_u1), pancreatic polypeptide (Ss03375477_m1), somatostatin (Ss03391856_m1), and GAPDH (Ss03375629_u1), a housekeeping gene. Analysis by RQ software (ABI 7900HT) used the ΔΔ Ct method, and data were plotted as RQ. Controls were no template reverse transcription control, no template quantitative PCR control, and human MSC cDNA to verify probe specificity to porcine cDNA.

Male inbred B6.129S7-Rag1^tm1Mom/J mice (The Jackson Laboratory) were used as transplant recipients. Animals were maintained under virus antibody–free conditions in climatized rooms with free access to sterile tap water and pelleted food. Mice were rendered diabetic by IV injection of 185 mg/kg streptozotocin freshly dissolved in acetate buffer (Sigma) 2–4 days before transplantation. Blood samples were obtained from the tail vein for glucose measurement (OneTouch UltraMini glucose meter). Grafts consisting of 3,000 NPIs alone or 3,000 NPIs + 10⁶ MSCs were transplanted under the left-side kidney capsule in Rag mice (16). Grafts were aspirated into polyethylene tubing (PE-90), pelleted by centrifugation, and then gently placed under the kidney capsule with the aid of a micromanipulator syringe. To examine whether NPI-MSC cell-to-cell contact is essential, in two separate transplant cohorts, mice were implanted under the kidney capsule with 3,000 NPIs, and immediately after implantation, they were injected IV through the tail vein with 10⁶ MSCs obtained from two independent bone marrow donors. Controls included mice transplanted with 3,000 NPIs under the kidney capsule but with no IV MSC injection.

All mice were monitored for nonfasting blood glucose levels at 3 days posttransplantation and once a week thereafter. When the blood glucose level was ≤11.1 mmol/L for 2 consecutive weeks, mice were deemed normalized. After normalization, an oral glucose tolerance test (OGTT) was performed on transplanted mice (16). After a 12-h fast, d-glucose (3 mg/g) was administered as a 50% solution intragastrically into nonanesthetized mice. Blood samples were obtained from the tail vein at 0, 15, 30, 60, and 120 min. All mice subsequently underwent a survival nephrectomy of the graft-bearing kidney, which was taken for morphological analysis or assessment of cellular insulin content (16). Nephrectomized animals were subsequently monitored to confirm a return of hyperglycemia.

The graft-bearing kidneys were prepared for immunohistological analysis by fixation in 4% weight for volume paraformaldehyde (BDH Laboratory Supplies), embedded in paraffin, and cut into 5-μm sections for processing and immunostaining. After rehydration, antigen retrieval for tissue samples was performed with Tris-EDTA buffer (pH 9.00). The samples were then blocked with 20% normal goat serum (NGS) (Jackson ImmunoResearch) for 1 h. Tissues were stained with a guinea pig anti-insulin antibody diluted at 1:1,000 (Dako) and a rabbit anti-CD31 antibody diluted at 1:50 (Abcam) in 5% NGS. Secondary antibodies used were Alexa Fluor 594 goat anti-rabbit and Alexa Fluor 488 goat anti-guinea pig (Molecular Probes) diluted at 1/200 in 5% NGS. Slides were cover slipped with ProLong Gold Antifade Mountant (Invitrogen) to preserve fluorescence. Negative controls included sections of the same tissues incubated without primary antibodies and only the secondary antibody, whereas positive controls included sections of neonatal porcine pancreas for both insulin and CD31 staining. Separate negative and positive controls were used for each independent staining procedure and subsequent imaging. Slides were visualized with an Axioscope 2 microscope equipped with an AxioCam MRc digital camera and analyzed with AxioVision 4.6 software (Carl Zeiss).

To assess the effects that MSCs may have on the degree of graft vascularization in the early transplant period, a cohort of mice were implanted with either NPI alone or NPI + MSCs, and then the grafts were harvested at 3 weeks posttransplantation and immunostained for insulin and CD31⁺ cells. In these grafts, the CD31⁺ area and the DAPI-positive cell number were measured by using the histogram feature of ImageJ and hand counting, respectively. The ratio of CD31⁺ vasculature was then calculated by dividing the number of CD31⁺ pixels by the number of DAPI-positive cells in the graft.

The grafts also were measured for total cellular insulin content (16). Extracted kidneys were homogenized and sonicated at 4°C in 10 mL of 2 mmol/L acetic acid containing 0.25% BSA. After 2 h at 4°C, tissue homogenates were resonicated and centrifuged at 10,000g for 25 min, and supernatants were collected. Pellets were further extracted by sonication in an additional 5 mL of acetic acid. The second supernatant was collected after centrifugation and combined with the first supernatant, the total volume was measured, and the samples were assayed for insulin content (Mouse/Rat Insulin Kit; Meso Scale Discovery).

Data are expressed as the mean number of independent observations unless otherwise specified, with individual biological replicates shown as points. Statistical significance of differences was calculated by either Student t test with the Holm-Sidak method for correction for multiple comparisons where appropriate if the sample approximated a normal distribution or Mann-Whitney U test if a normal distribution could not be assumed. Median time to normoglycemia was compared by using the Mantel-Cox log-rank test.

Quantification of the proportion of β- and α-cells in the grafts revealed no difference when NPIs were cultured alone or with MSCs (Table 1). Moreover, no differences were found in the number of proliferating insulin/PCNA double-positive β-cells. In contrast, coculture of NPIs with MSCs resulted in significantly more cellular insulin content than NPIs cultured alone (34.60 ± 0.75 vs. 27.43 ± 3.22 μg/pancreas, respectively; P < 0.05) (Table 1). During a GSIS assay, a significant difference (P < 0.01) was found in the amount of insulin secreted between the two culture groups at low glucose (2.8 mmol/L), high glucose (20.0 mmol/L), and stimulation index (Table 2). Gene expression analysis of the NPI and NPI + MSC cocultures revealed that cocultures had significantly higher levels (P < 0.01) of pancreatic polypeptide (1.6-fold increase). No difference in insulin, glucagon, somatostatin, and PDX1 gene expression (NPI + MSC vs. NPI, respectively) (Fig. 1) was observed.

Metabolic follow-up of glycemia and weight was measured on the transplanted mice weekly, and an OGTT was administered once recipients reached and maintained normoglycemia. MSC and NPI cotransplanted recipients (n = 14) exhibited significantly lower glycemia at weeks 18–20 and 22 posttransplantation compared with those with islets alone (n = 14) (Fig. 2A). Average weight decreased immediately after transplantation, and subsequently increased after week 3 posttransplantation. Weight was comparable between cotransplanted and islet-alone recipients (Fig. 2B). Mice with cotransplants reached normoglycemia significantly earlier than islet-only recipients, with a median time to normoglycemia of 19.5 weeks versus 25 weeks, respectively (P < 0.001) (Fig. 2C). Furthermore, at 20 weeks, 64% of mice in the cotransplant group had obtained normoglycemia compared with 0% transplanted with islets alone; cotransplanted mice demonstrate a more rapid return to glycemia. At the end of the follow-up period, 100% of the mice in both groups obtained normoglycemia. Individual experimental metabolic data are shown in Supplementary Fig. 1. After nephrectomy of the graft-bearing kidney, all mice returned to hyperglycemia within 24 h, confirming that resolution of hyperglycemia was due to the graft and not to regeneration of β-cells.

Mice transplanted with islets alone were less glucose tolerant compared with cotransplant recipients (Fig. 2D). During an OGTT, mice with NPI + MSC grafts had significantly lower glycemia than the islet-alone grafts at 15 min (15.6 vs. 19.7 mmol/L; P < 0.05, respectively) after glucose challenge. Furthermore, the glucose response curve was significantly lower in mice with NPI + MSC grafts than in those with NPI-alone grafts (area under the curve [AUC] 1,278 ± 74 vs. 1,072 ± 50 mmol/L · min; P < 0.05). Individual experimental OGTT data are shown in Supplementary Fig. 1. Examination of whether cell-to-cell contact is required for MSCs to exert an effect on NPIs revealed that mice transplanted with NPIs alone under the kidney capsule and given IV administration of MSCs exhibited no significant difference in weekly glycemia, time to normalization, glucose tolerance, or graft cellular insulin content compared with mice not given IV MSCs (Fig. 3A–D).

Immunohistochemical examination of the grafts at 35 weeks posttransplantation revealed highly vascularized tissue that consisted predominantly of insulin-positive cells (Fig. 4). Insulin staining at 3 weeks posttransplantation was minimal (Fig. 4A and C) compared with 35 weeks posttransplantation (Fig. 4E and G), although no differences between the two groups at either time point was observed. In addition, immunohistological analysis of the grafts for TUNEL showed no evidence of apoptosis in either NPI or NPI + MSC at week 2 or 3 posttransplantation (data not shown).

To assess graft vascularization, grafts were obtained at 3 or 35 weeks posttransplantation and immunostained for CD31⁺ vasculature (Fig. 4). We hypothesized that the more-rapid normalization in the cotransplant mice is a result of earlier revascularization due to MSC secretion of trophic factors. In a separate cohort of mice transplanted with or without MSCs, grafts were collected at 3 weeks posttransplantation to assess vascularization. Compared with the islet-alone graft (Fig. 4B), the NPI + MSC graft exhibited a markedly increased CD31⁺ vasculature (Fig. 4D) at week 3. An abundance of CD31⁺ vasculature was observed throughout the cotransplant graft, whereas a CD31⁺ vasculature was comparatively sparse in the islet-only graft. At week 35, grafts were observed to have similar amounts of CD31⁺ vasculature in the two groups (Fig. 4F and H). Moreover, quantification of comprehensive images at week 3 posttransplantation demonstrated a significantly more-CD31⁺ vasculature in cotransplant grafts versus NPI-alone grafts (117.4 ± 24.7 vs. 35.6 ± 6.3 area/cell; P < 0.05) (Fig. 4I).

Grafts were removed after mice reached normoglycemia, and a subset of the grafts were homogenized to measure total cellular insulin content (Fig. 5). Recipients in the cotransplanted group had nearly 1.5 times more insulin than islet-only recipients (17.5 ± 2.37 vs. 12.0 ± 0.77 μg, respectively; P < 0.05).

One paired experiment with a single NPI preparation and MSC donor was found to have contrasting results from the other experiments (Fig. 6). A review of the MSC donor information revealed that the donor had an autoimmune disease (psoriatic arthritis [PA]). In this paired experiment, cotransplant recipients (n = 6) exhibited no difference in glycemia than islet-only recipients (n = 4) in weeks 1–32 posttransplantation. No difference in weight was observed between groups (Fig. 6B), although the weights increased steadily at week 5 posttransplantation after an initial decline. Of note, mice with islets alone began to reach normoglycemia significantly earlier than mice with cotransplants (median time to normoglycemia 25.5 vs. 32 weeks, respectively; P = 0.05) (Fig. 6C). Glucose tolerance (Fig. 6D) and total graft insulin content (9.3 ± 1.73 and 10.3 ± 0.83 μg for NPI and NPI + MSC, respectively) did not differ between cotransplant and islet-only recipients (data not shown).

In this clinically relevant study, we demonstrate that cotransplantation of human bone marrow–derived MSCs with NPIs under the kidney capsule of streptozotocin-induced diabetic mice resulted in better functional outcomes than NPIs alone. Mice receiving cotransplants achieved normoglycemia significantly sooner, exhibited lower glycemia, and were more glucose tolerant, and significantly more graft cellular insulin content was recovered posttransplantation. Before transplantation, NPI and MSC cocultures contained significantly more total cellular insulin content and exhibited greater GSIS compared with NPIs cultured alone.

Several studies have shown that cotransplantation of rodent MSCs and islets result in lower blood glycemia (20–23,38), which is in accordance with the current results because we found significantly lower glycemia for several weeks midway through the transplant period (Fig. 1A). The current results also agree with evidence that cotransplantation reduces the time to achieve normalization compared with islet transplantation alone (22,23). Similarly, Ito et al. (21) reported superior glucose tolerance in mice receiving rodent islet and MSC cotransplants, which concords with the current findings. Because the proportion of time a patient spends in a hyperglycemic state is directly related to the risks of developing long-term complications and comorbidities (39), this finding is of great clinical significance. The mechanisms by which MSCs improve glycemic control are not yet known, but some evidence supports the secretion of trophic factors as a potential explanation. In particular, annexin A1 appears to play a significant role in enhancing GSIS, and annexin knockdown severely impairs MSC function (12). Moreover, the current data support this observation because NPIs cocultured with MSCs exhibited significantly increased GSIS as well as total cellular insulin content. Furthermore, MSCs have been found to upregulate certain genes involved in insulin secretion and synthesis in β-cells (40). However, when NPIs were cocultured with MSCs, we observed no significant differences in the insulin, glucagon, somatostatin, and PDX1 transcripts; however, significantly more pancreatic polypeptide transcripts were found in the NPIs cocultured with MSCs compared with NPIs alone.

Rapid development of a vascular system is important for islet function after transplantation (5), and earlier angiogenesis could contribute to improved NPI function as evidenced by the increase in CD31⁺ vasculature in the cotransplanted grafts. MSCs secrete VEGF, a trophic factor that induces angiogenesis (41). Many studies involving transplanted MSCs have found an increase in capillaries (20,21), endothelial cell–positive areas (38), or VEGF production (42) in graft areas, although one showed no increase in vascularization (23). In addition, mice deficient in pancreatic VEGF expression exhibited impaired glucose tolerance (41), indicating that VEGF may improve glucose tolerance. Fabryova et al. (42) used rodent MSCs to enhance vascularization of an implanted scaffold intended to create adequate vascularization in previously unsuitable sites for transplantation, such as subcutaneous and the greater omentum (42). MSCs were unable, however, to provide adequate vascularization for the intramuscular site (40). The lack of any effect when MSCs are administered IV (Fig. 3) indicates that MSC action may be site dependent, and cell-to-cell contact may be necessary. Further investigation is required to identify the specific mechanisms by which MSCs affect the graft microenvironment.

Cotransplant recipients also exhibited superior glucose tolerance (Fig. 2D) and higher total preculture and posttransplantation graft cellular insulin content (Table 2 and Fig. 5). We have previously reported that human MSCs prevent β-cell apoptosis in human islets in vitro (37), and in vivo studies have suggested that rodent MSCs can prevent rodent β-cell apoptosis (23). A lack of immunostaining for TUNEL on early time point grafts indicate no apoptosis in grafts from either group at 3 weeks posttransplantation (data not shown), suggesting that MSCs do not prevent β-cell apoptosis in NPIs. This observation does not contradict previous findings, however. NPIs are known to be more resistant to hypoxia than human islets (17) and are able to proliferate after transplantation (16). Therefore, that NPIs are less sensitive to hypoxic conditions in the initial posttransplantation period is not unexpected.

MSC donor characteristics such as age and sex have an impact on the therapeutic potential of MSCs (26). Less is known about how the specific disease status can affect characteristics of MSCs (26), but autoimmune diseases as a whole have been associated with some dysfunction (26), including altered morphology (32), decreased proliferative capacity (32–34), altered gene expression (35), decreased expression of trophic factors (33,35), and reduced inhibitory effect on T cells (32,35). In one of the paired experiments, we discovered that the MSC donor had PA, an autoimmune disease in which joints are targeted for immune attack and inflammation. The experiment using the PA MSCs from this donor was conducted with the same protocol as the experiments discussed above. In this experiment, mice in the cotransplant group had no differences in glycemia (Fig. 6A), significantly increased time to normalization (Fig. 6C), no difference in glucose tolerance (Fig. 6D), and no difference in insulin content (data not shown). The positive effects that were exemplified in the previous experiments disappeared, and in fact, PA MSCs had a detrimental effect on the ability of NPIs to reverse diabetes, suggesting that these MSCs had a functional impairment, such as a reduction in VEGF expression or reduced anti-inflammatory effects. Some studies show that MSCs from donors with diabetes may have decreased proliferative capacity and differentiation potential (26,28). Furthermore, MSCs from rodent and human donors with type 2 diabetes have been shown to have reduced proliferative capacity, higher levels of senescence and apoptosis, and decreased differentiation potential (28,43) likely caused by the diabetes microenvironment of high glucose and increased concentrations of advanced glycation end products and reactive oxygen species (28). Therefore, donors with type 1 diabetes may have impaired MSCs for two reasons: autoimmune disease status and diabetic microenvironment. Learning more about the functional capacity of MSCs from donors with type 1 diabetes is critical because use of autologous MSCs from these patients may not be as effective.

In this study, the results demonstrate that MSCs have a beneficial effect on the efficacy of NPI transplantation and in vitro function. MSCs cotransplanted with NPI encourage islets to be curative faster and increase the glucose tolerance of islets after reversal of diabetes potentially as a result of earlier development of a vascular system and enhanced glycemic control.

Funding. This study is supported by the Alberta Diabetes Institute, the Canadian Institutes of Health Research (grant #MOP 119500), and the Canadian Stem Cell Network. The University of Alberta Swine Research Centre provided the neonatal pigs for the study.

The funders had no role in the study design, data collection and analysis, decision to publish, or publication of the manuscript.

Duality of Interest. No conflicts of interest relevant to this article were reported.

Author Contributions. J.A.H. drafted the manuscript, collected data, and contributed to the data analysis and interpretation. C.E.E. contributed to the data analysis and critical revision of the manuscript. K.S. contributed to the isolation of NPIs, data research, and critical revision of the manuscript. T.L. contributed to the image analysis and cell composition quantification. B.S. contributed to the isolation of islets and transplantation. A.M.-S. contributed to the data research. P.K. contributed to the characterization of the islet grafts. A.A. contributed to the data research and study concept and design. G.S.K. contributed to the study concept and design, data analysis and interpretation, critical revision of the manuscript, and securing of funding. G.S.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.