Identification and removal of unexpected proliferative off-target cells emerging after iPSC-derived pancreatic islet cell implantation

Significance The unrestricted capacity of human induced pluripotent stem cells (iPSCs) to proliferate and differentiate is a major advantage for the development of cell-based therapies. However, this advantage is accompanied by a risk of off-target cell proliferation and tumorigenesis. Therefore, minimizing the risk of off-target cell contamination is imperative for the safe use of iPSC-derived products. In this study, we identified proliferative off-target cells, termed PMSCs (proliferative mesenchymal stem cells), that emerged unexpectedly after the implantation of iPSC-derived pancreatic islet cells. Additionally, we developed an in vitro detection system and effective removal methods for PMSCs, paving the way for safer clinical applications of PSC-derived islet-like cells.


Cell culture and differentiation
Cells were maintained on iMatrix-511 (Nippi)-coated dishes in StemFit AK03N (Ajinomoto) at 37 °C in a humidified 5% CO2 incubator.Cells were passaged every 3 or 4 days via non-enzymatic dissociation using 0.5 mM EDTA (Thermo Fisher Scientific) and subjected to differentiation experiments, usually after more than 2 weeks of running culture.For differentiation culture to generate s7-iPICs, we performed 3D stirred-floating aggregate culture based on our previous reports (1)(2)(3).The details of a typical differentiation culture are provided below.
The following day, CHIR99021 was removed from the medium and the culture was continued for another 2 days.

Type 1 diabetes mouse model
NOD.CB17-Prkdc-scid/J (NOD-scid) mice were obtained from the Charles River.Male mice between the ages of 8 and 9 weeks were intraperitoneally administered multiple low doses of streptozotocin (50 mg/kg/day for 5 days; Sigma-Aldrich).Mice that became hyperglycemic within 2-3 weeks of streptozotocin injection were subjected to implantation experiments as a type 1 diabetes mouse model.Individual mice were implanted only once at one site over an extended period.
Fibrinogen from human plasma (Merck Millipore) and thrombin (Sigma) were reconstituted in iMEM and PBS to prepare 10 mg/mL and 50 IU/mL solutions, respectively, and stored at −80 °C until use.
For implantation using alginate gel (Fig. 2A and B), cell aggregates were suspended in 3% alginate (NovaMatrix) in 25 mM HEPES buffer, cross-linked with 75 mM strontium chloride hexahydrate (FUJIFILM Wako) in 25 mM HEPES buffer, and formed into discs.We monitored the blood glucose levels of the implanted animals using the Accu-Chek Aviva system (Roche DC Japan) and collected plasma samples from the tail vein on the indicated days.For the oral glucose tolerance test, the mice were fasted overnight and orally injected with 2 g/kg glucose solution (Otsuka).Plasma samples were collected from the tail vein before and at 15, 30, 60, and 120 min after injection.

Tissue processing and immunostaining
Grafts were collected 23-30 weeks after implantation, fixed with 4% paraformaldehyde (FUJIFILM Wako) for over 24 h at 4 °C, and embedded in paraffin or frozen in OCT compound.Paraffin blocks were sectioned at 5 μm and used for hematoxylin and eosin (HE) staining and immunostaining.
Frozen blocks were sectioned at 10 μm and used for immunofluorescence staining.The primary antibodies are listed in Table S1.Secondary antibodies were conjugated to Alexa Fluor 488, 546, or 568 (Thermo Fisher Scientific or Jackson ImmunoResearch).The sections were counterstained with Hoechst (Thermo Fisher Scientific) to label nuclei.

Single-cell RNA sequencing library preparation, sequencing, and data processing
A total of seven samples (one sample of Vitro s6-iPICs, two samples of Vivo s6-iPICs, one sample of reference human islets, one sample of Vitro s6-iPICs cultured without PD-166866, and two samples of s6-iPICs after 4 weeks of extended culture, including EGF treatment) underwent scRNA-seq.Vivo s6-iPIC samples were de-crosslinked using 100 mM sodium citrate solution.Human islets were purchased from PRODO (HP-18304-01, Donor age: 21 years, Donor sex: Male, Donor BMI: 27.3 kg/m 2 , Donor HbA1c: 5.4%, Estimated purity: 85%, Estimated viability: 95%).Single-cell RNA-seq libraries were generated using the 10x Genomics Chromium TM controller and Chromium Single Cell 3 kit v2 (10x Genomics) according to the manufacturer's instructions.Successful cDNA amplification and library construction were ensured using high-sensitivity DNA kits on an Agilent 2100 Bioanalyzer (Agilent).The obtained libraries were sequenced using HiSeq (Illumina) with 150 bp paired-end reads at a depth of > 100,000 reads per cell.The sequencing reads of the in vitro s6-iPIC samples were aligned to the human GRCh38 genome reference, and gene counts were quantified as UMIs using Cell Ranger (10x Genomics).The sequencing reads of the in vivo s6-iPIC samples were first aligned to the human GRCh38 and mouse mm10 genome references.Cells with at least one UMI count of the mouse genes were collected.Thereafter, the sequencing reads were aligned to the human GRCh38 genome reference and the collected cells containing mouse genes were removed for further analyses.We imported UMI count matrices into the R software Seurat package (4,5), where normalization was performed according to the default settings.Cells with mitochondrial gene counts over 10% were regarded as dead or damaged and were removed before further analyses.The UMI count matrices were scaled by regressing the total number of UMI counts per cell and the percentage of mitochondrial gene counts.The genes for dimensional reduction were selected based on the average expression and dispersion of each gene, and a principal component analysis was performed.
Principal components were used for Seurat's shared nearest neighbor graph clustering, and tdistributed stochastic neighbor embedding (t-SNE) or uniform manifold approximation and projection (UMAP) dimensional reduction were used to visualize the data.We examined the similarity between clusters using hierarchical clustering analysis of average expression.The cell cycle was evaluated and scored based on the expression of genes known as S-phase, G1, and G2M markers.To estimate cell types and similarities within s6-iPIC samples with reference to known tissues or cell lines, we performed RCA using the RCA package (6).Differential gene expression analysis of each cluster compared with the others was performed using the likelihood-ratio test for single-cell gene expression in the Seurat.For pseudotime analysis, the processed UMI count matrices were imported into a singlecell dataset for the monocle package (7-9).We selected the genes for ordering the cells using 'dpFeature' in monocle, to set Cluster 15 as the root (pseudotime was zero) and contracted the singlecell pseudotime via the 'DDRTree' algorithm.Differential gene expression analyses along with pseudotime were performed using the likelihood-ratio test in monocle, and genes with q-values less than 0.05 were identified as differentially expressed genes along with pseudotime.Differentially expressed genes were classified into gene clusters using pseudotime expression patterns, and functional enrichment analysis of each gene cluster was performed using the clusterProfiler package (10).We obtained previously reported scRNA-seq data using in vivo grafts of ESC/iPSC-derived isletlike cells from publicly available database (GSE151117).The UMI count matrices from three ESC (HUES8)-derived islet-like cell grafts (GSM4567001, GSM4567002, and GSM4567003) and two iPSC (WS4 corr )-derived islet-like cell grafts (GSM4567004 and GSM4567005) were imported into Seurat package and low quality cells were removed.After normalization and scaling the data, the principal component analysis was performed and clustering and dimensional reduction were conducted.

Flow cytometry
Differentiation efficacy and quality were analyzed at the individual stages, based on developmental markers using immunostaining methods and LSRFortessa X20 flow cytometry equipment (BD), as described previously (2).The data were processed using FlowJo software.The primary antibodies are listed in Table S1.Secondary antibodies of the appropriate species were conjugated to Alexa Fluor 488, 546, 568, or 647 of the appropriate species (Thermo Fisher Scientific or Jackson).

Plasma glucose and hormone measurements
Plasma glucose, human C-peptide, and mouse C-peptide levels were measured using the Glucose Test C-II Wako (Fujifilm Wako), Mercodia Ultrasensitive C-peptide ELISA (Mercodia), and Mouse Cpeptide Measurement Kit (Morinaga), respectively, according to the manufacturer's instructions.(C) Top 10 significant genes (q-value < 0.05) and enriched gene ontology (GO) terms (q-value < 0.05) of each gene cluster.

SI References
TableS1

Fig. S3 .
Fig. S3.Detailed information of RCA.Related to Fig. 3. (A) Heatmap of similar tissues or cell lines for each color classified by the RCA in Fig. 3A.

Fig. S4 .
Fig. S4.Comparison of Clusters 15 and 9 using pseudotime analysis.(A) Estimated pseudotime from Cluster 15 to Cluster 9 on the t-SNE projection.(B) Expression heatmap of differentially expressed genes ordered by their common kinetics through pseudotime.

Fig. S5 .
Fig. S5.Tissue-specific gene enrichment analysis of the Cluster 15 specific genes.(A and B) Annotation of the top 20 differentially expressed genes in Cluster 15 (Fig. 3C) using TissueEnrich (https://tissueenrich.gdcb.iastate.edu/).(A) Human tissues extracted as related tissues from the Human Protein Atlas dataset.(B) Expression intensities of SFRP4, OGN, and IGF1 (causative genes of high relevance to the uterus in Fig. S5A) in each human tissue.

Fig. S6 .
Fig. S6.Additional information for tissue immunostaining.Related to Fig. 3E and F. (A) HE and immunohistochemical images of an s7-iPIC graft with abnormal outgrowth at 24 weeks post-implantation.Black and white scale bars indicate 500 μm.Images were taken from serial sections of the same sample as in Fig. 3E and F and are representative of dozens of samples showing similar results.

Fig. S8 .
Fig. S8.Proliferation-promoting activity of betacellurin and EGF on the non-endocrine population in s6-iPICs.(A) Schematic representation of s6-iPIC induction including betacellulin or EGF treatment, and subsequent flow cytometry analysis.(B) Flow cytometry plots illustrating the protein expression of s6-iPICs differentiated with betacellulin or EGF treatment.The numbers in each plot diagram show the percentage of each population.The CHGA-negative non-endocrine population, shown in red, increased in a dose-dependent manner by betacellulin or EGF treatment.

Fig. S9 .
Fig. S9.Additional information for single-cell RNA sequencing analysis containing extended culture samples.Related to Fig. 4. (A) Shared nearest neighbors clustering of the reanalyzed scRNA-seq data in Fig. 4D on the UMAP projection.Reanalysis with the additional samples re-assigned Clusters 15 and 9 in Fig. 2D to Clusters 24 and 19 in this figure, respectively.(B) Bubble plot of endocrine and non-endocrine signature genes in the clusters classified in Fig. S9A.Color intensity indicates average relative expression levels.The bubble size indicates the percentage of expressing cells.(C) Single-cell gene expression of PDX1 and CHGA on the UMAP projection in Fig. 4D.(D) Heatmap of similar tissues or cell lines for each color classified by the RCA in Fig. 4E.

Fig. S10 .
Fig. S10.Evaluation of the effects of cyclophosphamide, tamoxifen, anastrozole, and lapatinib treatment on putative PMSC and cyst populations.(A) Schematic representation of s6-iPIC derivatives induced with or without additional compound treatment and subsequent extended culture and flow cytometry analysis.(B) Relative cell numbers of PDX1 − /CHGA − , PDX1 + /CHGA − , and CHGA + populations in each s6-iPIC derivative post-extended culture.The number of cells in control s6-iPICs was set to "1".The number of cells in each population was calculated from flow cytometry results (percentage of each population) and live cell counts.Data are shown as the mean ± SD (n = 3, technical replicates).*P < 0.05, versus s6-iPICs, Dunnett's test.

Fig. S11 .
Fig. S11.Evaluation of the effect of kinase inhibitors, cisplatin, and docetaxel treatment on insulin-positive cell rates.(A) Schematic representation of s6-iPIC derivatives induced with or without additional compound treatment and flow cytometry analysis.(B and C) Representative flow cytometry plots illustrating insulin and NKX6.1 protein expression before an extended culture of s6-iPIC derivatives.