Administration of mesenchymal stem cells in diabetic kidney disease: a systematic review and meta-analysis

Search strategy

We searched the Embase, Cochrane Library, ISI Web of Science, PubMed, and US National Library of Medicine (NLM) databases through January 2020 for original papers that assessed the effects of MSC administration on DKD animal models or patients without language restrictions. Keywords in this research included the following: (mesenchymal stem cells OR MSC OR multipotent stromal cells OR mesenchymal stromal cells OR mesenchymal progenitor cells OR Wharton jelly cells OR adipose-derived mesenchymal stem cells OR bone marrow stromal stem cells) AND (diabetic nephropathy OR DN OR diabetic kidney disease OR DKD).

Randomized controlled trials, comparative studies, or controlled trials that assessed the efficacy or safety of MSC therapy for treatment as an intervention in DKD animal models (without species limitations) or patients with DKD were included. The included studies were required to contain biochemical data on renal function or adverse events and to report albuminuria and impaired renal function in patients or animals with DM. The precise distinction between DKD and diabetic nephropathy (DN) was outside the scope of this paper, and both were included. Reviews, case reports, meta-analyses, comments, and letters were excluded. Articles that studied embryonic stem cells, induced pluripotent stem cells, or MSC components (rather than actual MSCs) for the treatment of DKD were excluded. In addition, studies that lacked a control arm or did not provide essential data such as renal function and sample size were excluded. We also searched for additional relevant reports by browsing the references of the articles.

Data extraction

The main features of the included studies were summarized, and the data were extracted independently by two authors using a standardized datasheet. Adverse events and biochemical indicator data were extracted from the articles, such as blood glucose, CCr, SCr, BUN, U-albumin/U-creatinine ratio (U-ACR), microalbuminuria, urinary albumin excretion, urine protein/Cr, kidney weight, body weight, and kidney weight/body weight ratio. If a paper contained no specific information, data were obtained by measuring the chart in the paper or by contacting the primary authors. Any disagreements in the extracted data were resolved by the third author.

Validity and quality assessment

For clinical trials, quality assessment was performed using 4 items based on the Jadad scale [12]: randomization, concealment of allocation, blinding method, and description of withdrawals and dropouts. A total score of ≥ 3 was considered high quality.

For animal studies, the methodological quality assessment was carried out using a risk of bias (RoB) tool by the Systematic Review Centre for Laboratory Animal Experimentation (SYRCLE), which is based on the Cochrane RoB tool and adjusted for animal experiments. The following ten items were assessed. (1) Sequence generation: Were the subjects randomly assigned to the case or control groups with an adequately generated allocation sequence? (2) Baseline characteristics: Were the baseline characteristics of the two groups comparable? (3) Allocation concealment: Was the allocation of all the subjects adequately concealed? (4) Random housing: Were all the subjects randomly housed in the same environment during the experiment? (5) Researcher blinding: Were the researchers blinded to which subjects had received treatment (in this case, MSC treatment)? (6) Random outcome assessment: Were the animals selected in random order for outcome assessment? (7) Blinding of outcome assessors: Were the outcome assessors blinded to the group information? (8) Incomplete outcome data: Were incomplete outcome data or dropouts adequately addressed? (9) Selective outcome reporting: Was the study free of selective outcome reporting for significant results? (10) Other sources of bias: Was the study apparently free of other problems that could result in a high risk of bias, such as contamination of MSCs, inappropriate influence of the funder, errors in units of analysis, design-specific risk of bias, and additional animals to replace dropouts? An answer of “yes” means a low risk of bias, while “no” means a high risk of bias, and “unclear” means the risk of bias cannot be assessed for the lack of sufficient information. Disagreements were resolved by consensus-oriented discussion.

Statistical analysis

Stata SE 12 was used for statistical analysis. For continuous variables, standard mean differences (SMDs) were obtained by pooling the mean values, standard deviations, and sample sizes. For binary data, the odds ratio (OR) was calculated. Moreover, 95% confidence intervals (95%CIs) were calculated between the MSC-treated groups and the control groups. If there were multiple MSC-treated groups in an article, the data in the control group were reused. Heterogeneity across studies was quantified using I² and was considered significant at a p value of < 0.1. The data were pooled using a fixed-effect model without heterogeneity or a random-effect model. A p value of < 0.05 was regarded as statistically significant for all analyses. Potential publication bias was assessed with Begg’s test, Egger’s test, and the trim-and-fill method.

Search results

In total, 33 studies in 29 publications were included, among which 28 publications were based on animal studies [13–40] and 1 was based on a clinical trial [41]. In addition, there are 4 ongoing clinical trials registered with the NLM.

Among 32 animal studies, 24 studies used rat models, 7 used mouse models, and 1 used a rhesus macaque model. A single method or a combination of multiple methods was used to induce DM, including streptozotocin (STZ) injection, high-fat diet dietary induction, nephrectomy, and natural development of models. However, the dosage and frequency of STZ injection and the time when the animals were tested for the establishment of DN were different. Although MSCs were used in all the included studies, the details of the source, dosage, frequency, administration, and point in time varied. The sources of MSCs were bone marrow mesenchymal stem cells (BM-MSCs) in 22 studies, adipose-derived stem cells (ADSCs) in 4 studies, human umbilical cord blood-derived mesenchymal stem cells (hUCB-MSCs) in 5 studies, and stem cells from exfoliated deciduous teeth in 1 study. Allogeneic administration was used in 23 studies, xenoplastic administration was used in 8 studies, and autologous administration was used in 1 study. The characteristics of the included animal studies are summarized in Table 1.

The only clinical trial was a multicenter, randomized, double-blind, dose-escalating, sequential, placebo-controlled study, finished in 2016. Thirty patients were randomized to receive one of two doses of mesenchymal precursor cells or placebo, and the efficacy and adverse events were observed. The main features of the clinical trial are shown in Table 2.

None of the animal experiments reported the occurrence of graft rejection after administration, but 2 MSC-treated human patients developed antibodies specific to the donor HLA in the clinical trial, one of these cases occurred transiently, whereas the other presented at baseline and persisted throughout the observation period without the appearance of adverse events. Strangely, however, antibodies specific to the donor HLA were also found in one placebo-treated patient. Six animal experiments specified the deaths or dropouts. Lang and Dai [27] reported the deaths of 6 model rats during the construction of the diabetes model (21.4%, 6/28), and Wang et al. [21] reported 1 death each in the MSC-treated group (8.3%, 1/14) and the DN group (10%, 1/10) as well as 2 deaths because of anesthesia. In the study of Li et al. [32], no rat died in the DN group (0.0%, 0/14), and 2 died in the MSC-treated group (18.2%, 2/11). During a 12-week observation, the MSC-treated group (25%, 3/12) had lower mortality than the DN-treated group (66.7%, 8/12) [33]. Similarly, Xian et al. [34] found 2 deaths in the hUCB-MSC group (16.7%, 2/12), making for a markedly lower mortality rate than the T1DM group (40%, 6/15) at the end of the study. An et al. [39] found no marked change in the immune system of rhesus macaque DN models in response to hUCB-MSC treatment.

Quality assessment

Quality assessments of animal experiments and clinical trials were performed (Tables 3 and 4). Table 3 shows a number of “unclear” judgments in the quality assessment of animal experiments; in particular, outcome assessment in a random order, concealment of allocation and blinding of outcome assessors in all included experiments were rated “unclear,” largely due to a lack of awareness of randomization and blinding methods in animal experiments. As shown in Table 4, a total score of 7 suggested the high methodological quality of the included clinical trial.

Assessment of glucose

Glucose was detected after MSC treatment in all but 2 studies [32, 37]. Sixteen studies measured glucose once at the end of the experiment [18–21, 23–29, 31, 33–36]. Seven studies conducted blood glucose monitoring at several points in time [14–17, 22, 30, 40]. Five studies, 7 studies, 5 studies, 12 studies, 17 studies, 7 studies, and 2 studies were included to assess the effect of treatment on blood glucose levels at 1 week, 2 weeks, 3 weeks, 1 month, 2 months, 3 months, and 6 months, respectively, all of which showed a highly significant hypoglycemic effect in the MSC-treated group (1 week: SMD = − 1.484, 95%CI − 2.586 to − 0.381, p < 0.001; I² = 80.6%; 2 weeks: SMD = − 2.312, 95%CI − 3.743 to − 0.882, p = 0.002, I² = 89.6%; 3 weeks: SMD = − 4.007, 95%CI − 6.472 to − 1.541, p = 0.001, I² = 92.1%; 1 month: SMD = − 1.740, 95%CI − 2.660 to − 0.821, p < 0.001, I² = 83.8%; 2 months: SMD = − 1.830, 95%CI − 2.633 to − 1.028, p < 0.001; I² = 86.0%; 3 months: SMD = − 1.649, 95%CI − 2.838 to − 0.461, p = 0.007; I² = 84.6%; 6 months: SMD = − 3.045, 95%CI − 5.895 to − 0.195, p = 0.036; I² = 76.4%). The total hypoglycemic effect was also analyzed (SMD = − 1.954, 95%CI − 2.389 to − 1.519, p < 0.001; I² = 85.1%) (Fig. 1).

Assessment of serum creatinine

There were 4 studies, 2 studies, and 5 studies that assessed SCr at 1 month, 2 months, and 3 months, respectively. All of them showed significantly reduced creatinine values in the MSC-treated group (1 month: SMD = − 4.126, 95%CI − 7.936 to − 0.315, p = 0.034; I² = 94.9%; 2 months: SMD = − 3.505, 95%CI − 4.746 to − 2.264, p < 0.001; I² = 1.8%; 3 months: SMD = − 6.736, 95%CI − 10.311 to − 3.162, p < 0.001; I² = 89.0%). The total effect on SCr was also analyzed, suggesting that MSCs decreased SCr and improved renal function (SMD = − 4.838, 95%CI − 6.789 to − 2.887, p < 0.001; I² = 90.8%) (Fig. 2).

Assessment of blood urea nitrogen

BUN was evaluated at 5 different time points, each of which was used by relatively few studies. At 2 weeks (2 studies included), 3 weeks (2 studies included), 1 month (2 studies included), 2 months (3 studies included), and 3 months (4 studies included), BUN decreased in the MSC-treated group, although no statistical significance was seen at 3 weeks or 1 month (2 weeks: SMD = − 2.514, 95%CI − 3.582 to − 1.447, p < 0.001; I² = 37.3%; 3 weeks: SMD = − 4.432, 95%CI − 9.220 to − 0.356, p = 0.070; I² = 92.0%; 1 month: SMD = − 10.392, 95%CI − 21.247 to 0.464, p = 0.061; I² = 95.6%; 2 months: SMD = − 3.389, 95%CI − 6.679 to − 0.099, p = 0.044; I² = 89.8%; 3 months: SMD = − 5.902, 95%CI − 8.988 to − 2.815, p < 0.001; I² = 85.0%). The total effect on BUN was also analyzed, suggesting that MSCs decreased BUN (SMD = − 4.912, 95%CI − 6.402 to − 3.422, p < 0.001; I² = 89.3%) (Fig. 3).

Assessment of creatinine clearance rate

The data of six studies were pooled to evaluate CCr at 2 months after MSC treatment; CCr was significantly decreased in the MSC-treated group compared to the DKD group (2 months: SMD = − 1.881, 95%CI − 2.842 to − 0.921, p < 0.001; I² = 79.7%) (Fig. 4).

Assessment of blood insulin level

Two studies assessed insulinemia. The insulin level increased at 3 months after MSC treatment, although the significance was not notable (3 months: SMD = 3.051, 95%CI − 0.091 to 6.193, p = 0.057; I² = 90.3%).

Assessment of urine protein

The measurement of urine protein varied in the included studies. Microalbuminuria, urinary albumin excretion, the urinary albumin/urinary creatinine ratio, and the urinary protein/creatinine ratio were used to assess urine protein excretion in the DKD animals.

Urinary albumin excretion levels at 1 month (2 studies included) and at 2 months (7 studies included) were observed to be lower in the MSC-treated group than in the DKD group, although no significance at 1 month was observed (1 month: SMD = − 6.507, 95%CI − 17.935 to 4.921, p = 0.264; I² = 98.3%; 2 months: SMD = − 4.386, 95%CI − 5.891 to − 2.881, p < 0.001; I² = 85.5%). The total effect on urinary albumin excretion was also analyzed, suggesting that MSCs decreased urinary albumin excretion (SMD = − 4.830, 95%CI − 6.602 to − 3.058, p < 0.001; I² = 92.5%).

Microalbuminuria was detected at 3 weeks and 3 months; each of these time points was addressed by 2 studies that satisfied the inclusion criteria. Microalbuminuria was found to be decreased in the MSC-treated group at 3 months (3 weeks: SMD = − 9.112, 95%CI − 21.627 to 3.404, p = 0.154; I² = 95.3%; 3 months: SMD = − 4.431, 95%CI − 5.771 to − 3.091, p < 0.001; I² = 0.0%). The total effect on microalbuminuria was analyzed, suggesting that microalbuminuria was significantly lower in the MSC-treated group than in the DKD group (SMD = − 5.791, 95%CI − 8.681 to − 2.901, p < 0.001; I² = 86.3%).

The urinary albumin/urinary creatinine ratios at 1 month (6 studies included) and at 2 months (10 studies included) were observed to be significantly lower in the MSC-treated group than in the untreated DKD group (1 month: SMD = − 2.419, 95%CI − 3.070 to − 1.769, p < 0.001; I² = 0.0%; 2months: SMD = − 2.648, 95%CI − 3.454 to − 1.842, p < 0.001; I² = 58.9%). The total effect on the urinary albumin/urinary creatinine ratio was analyzed, and the analysis suggested that the urinary albumin/urinary creatinine ratio was significantly lower in the MSC-treated group than in the DKD group (SMD = − 2.539, 95%CI − 3.075 to − 2.003, p < 0.001; I² = 42.6%).

Urinary protein/creatinine ratios were not significantly different between the groups at 2 weeks (2 studies included) after MSC treatment (SMD = − 2.779, 95%CI − 7.617 to 2.059, p = 0.260; I² = 92.6%).

Assessment of kidney weight

Kidney weight and the kidney weight/body weight ratio were used to assess kidney hypertrophy. No significant intergroup difference in kidney weight was found between the MSC and untreated DKD groups at 1 month (2 studies included; SMD = − 0.674, 95%CI − 2.052 to 0.704, p = 0.337; I² = 67.0%).

The kidney weight/body weight ratio was found to be significantly decreased in the MSC-treated group at 2 months (8 studies included, SMD = − 1.364, 95%CI − 2.164 to − 0.565, p = 0.001; I² = 79.7%), while no significant difference was found between the two groups at 3 months (2 studies included, SMD = − 10.012, 95%CI − 29.753 to 9.729, p = 0.320; I² = 97.0%). The total effect on the kidney weight/body weight ratio was analyzed, and the analysis suggested that a reduced kidney weight/body weight ratio was found in the MSC-treated group (SMD = − 1.624, 95%CI − 2.594 to − 0.655, p = 0.001; I² = 86.9%).

Assessment of body weight

There were 3 studies and 5 studies that assessed body weight at the 1-month and 2-month time points, respectively. No significant difference in 1-month body weight was found between the two groups (SMD = 2.634, 95%CI − 0.730 to 5.999, p = 0.125; I² = 95.5%). At 2 months, the body weight of the MSC-treated groups significantly increased compared to that of the DKD groups (SMD = 0.903, 95%CI 0.346 to 1.459, p = 0.001; I² = 40.2%). An overall effect of MSC treatment on body weight was also found (SMD = 1.499, 95%CI 0.461 to 2.536, p = 0.005; I² = 87.3%).

Assessment of renal fibrosis

Four included studies evaluated the percentage of glomerulosclerosis at 2 months after MSC treatment, and no significant difference was found (SMD = − 0.350, 95%CI − 4.173 to 3.473, p = 0.858; I² = 96.2%).

Transforming growth factor-β (TGF-β) was measured at different time points using different methods. According to polymerase chain reaction (PCR) assays at 1 month (2 studies included) and 2 months (3 studies included) as well as western blot (WB) assays at 2 months (2 studies included), TGF-β was significantly decreased in the MSC-treated group (1-month PCR: SMD = − 3.281, 95%CI − 4.225 to − 2.337, p < 0.001; I² = 4.2%; 2-month PCR: SMD = − 7.594, 95%CI − 13.274 to − 1.915, p = 0.009; I² = 93.6%; 2-month WB: SMD = − 9.329, 95%CI − 11.569 to − 7.089, p < 0.001; I² = 16.2%). The same was true for the total expression of TGF-β (SMD = − 6.839, 95%CI − 9.367 to − 4.312, p < 0.001; I² = 90.5%).

Collagen I (Col-I) was detected by immunohistochemistry (IHC) and PCR. According to PCR at 2 months (3 studies included), Col-I was significantly decreased (SMD = − 11.856, 95%CI − 14.887 to − 8.826, p < 0.001, I² = 41.3%) in the MSC-treated group, although no significant intergroup difference was found by IHC at 2 months (2 studies included; SMD = − 4.714, 95%CI − 10.670 to 1.242, p = 0.121; I² = 95.3%). An overall effect on Col-I expression was also found (SMD = − 9.081, 95%CI − 14.233 to − 3.929, p = 0.001; I² = 95.1%).

Three included studies evaluated fibronectin (FN) by IHC at 2 months after MSC treatment, and a statistically significant decrease was found in the MSC-treated group (SMD = − 7.781, 95%CI − 10.680 to − 4.881, p < 0.001; I² = 71.3%).

Two studies evaluated α-smooth muscle actin (α-SMA) by WB at 1 month after MSC treatment, and 3 studies quantified its expression by PCR at 2 months. Both measures of α-SMA expression were significantly decreased in the MSC-treated group (1-month WB: SMD = − 2.514, 95%CI − 3.550 to − 1.479, p < 0.001; I² = 0.0%; 2-month PCR: SMD = − 2.098, 95%CI − 3.721 to − 0.476, p = 0.011; I² = 83.4%). An overall effect of MSC treatment on the expression of α-SMA was found (SMD = − 2.249, 95%CI − 3.311 to − 1.186, p < 0.001; I² = 72.1%).

E-cadherin was quantified by WB at 1 month (2 studies included) after MSC treatment; the treatment was associated with a significant and notable decrease in E-cadherin deposition (SMD = 3.600, 95%CI 2.338 to 4.861, p < 0.001; I² = 0.0%).

Assessment of inflammatory mediators

Monocyte chemokine protein-1 (MCP-1) was detected by IHC at 2 months (2 studies included) after MSC treatment, and no significant difference was found between the two groups (SMD = − 8.913, 95%CI − 20.994 to 3.167, p = 0.148; I² = 93.1%).

Tumor necrosis factor-α (TNF-α) was detected by enzyme-linked immunosorbent assay (ELISA) at 2 weeks (3 studies included) and by PCR at 1 month (2 studies included) after MSC treatment, both of which showed statistically significant decreases in the MSC-treated group (2-week ELISA: SMD = − 3.853, 95%CI − 7.207 to − 0.499, p = 0.024; I² = 90.4%; 1-month PCR: SMD = − 4.369, 95%CI − 6.835 to − 1.903, p = 0.001; I² = 57.5%). An overall effect of MSC treatment on the expression of TNF-α was found (SMD = − 4.027, 95%CI − 5.955 to − 2.098, p < 0.001; I² = 84.9%).

Risk of bias

Given sufficient data to assess publication bias, 2-month blood glucose was used for measurement. There was some degree of bias, indicated by a moderate asymmetry of the funnel plot, and Egger’s test showed p = 0.013. However, the trim-and-fill method did not identify any missing studies (Fig. 5).