Animal research study ethical approval and care

The care and treatment of all animals used in these studies were conducted in accordance with the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Research, National Research Council, U.S. Department of Health and Human Services) and with the approval of the Massachusetts General Hospital Institutional Animal Care and Use Committee (Protocols 2017N000216 and 2017N000214). Adult cynomolgus monkeys (Macaca fascicularis) were obtained from Charles River Laboratories, Alpha Genesis, Inc., and BC US LLC. Monkeys were typically socially housed except as necessary for study activities and received High Protein Monkey Diet (LabDiet, 5045) as well as produce and other dietary enrichment. Porcine donors were typically individually housed and received Minipig Grower Diet (LabDiet, 5081). Euthanasia criteria for recipients included serum creatinine of greater than or equal to 6 mg/dL or when a humane endpoint was met as assessed by veterinary staff. Recipients were euthanized with a mixture of pentobarbital sodium and phenytoin sodium.

Genotype and production of donor animals

Donors were produced and provided by eGenesis. All donors were engineered to eliminate three known glycan xenoantigens using CRISPR-mediated non-homologous end joining to deactivate the enzymes responsible for their generation (3KO: alpha-1,3-galactosyltransferase [GGTA1], cytidine monophospho-N-acetylneuraminic acid hydroxylase [CMAH], beta-1,4-galactosyltransferase 2 [B4GALNT2]). All kidney xenografts expressed human proteins that regulate important pathways related to xenotransplantation success, including complement, coagulation, and innate immunity. To increase the cohort size of this study, and because a plausible involvement of the different transgenes in the pathways under investigation seemed unlikely, multiple different genotype donors were included in this study (Supplementary Tables 1, 2). Retroviral inactivation of porcine endogenous retroviral elements using CRISPR/Cas9 mediated non-homologous end joining was performed in the case of the clinical candidate genetics, EGEN-2784. Production of gene-edited porcine donors generally involved: gene editing of primary cells from wildtype Yucatan minipigs in vitro, confirming desired genetic alterations, and somatic cell nuclear transfer to generate cloned pigs expressing genes of interest^30,31. In some cases, serial cloning and editing were conducted to layer edits (e.g., retroviral inactivation after insertion of human transgenes). For this study, all donors were female and were produced by cloning.

Life sustaining kidney xenotransplantation

On day 0, kidney procurement was performed under general anesthesia using a deceased donor multi-organ protocol as in clinical transplantation including heparin administration and in situ flush with University of Wisconsin preservation solution. For the recipient operation, after establishing appropriate general anesthesia, a midline incision was made and the porcine kidney was transplanted intra-peritoneally by anastomosing renal artery and renal vein to the abdominal aorta and vena cava, respectively. Typically, donor warm ischemia time was <1 min, cold ischemia time ranged from 30–60 min and anastomosis time was <30 min. Ureterovesical anastomosis was performed by Lich-Gregoir method without stenting. Bilateral native nephrectomy was performed simultaneously in 13/17 cases. In four cases, due to inadequate urine output or marginal recipient status during surgery, a single native kidney was left in place and later removed around day 20. In 12/17 xenotransplantation cases, central venous access was established by internal jugular vein catheterization on day −5 and removed after the first post-operative week. For the first few weeks postoperatively, the transplanted kidney was monitored by urine output, ultrasound, and serum creatinine measurement twice a week then weekly thereafter.

Immunosuppression and post-operative management

Induction therapy included anti-CD20 [2B8R1F8]-Afucosylated antibody (NIH Nonhuman Primate Reagent Resource [NIH-NPRR], Cat# PR-8288, RRID:AB_2819341, 20 mg/kg) on day −5, Anti-rhesus thymocyte (rhATG, NIH-NPRR, Cat# PR-0000e, RRID:AB_2716327, 5 mg/kg) on days −1 and 0, anti-CD154 mAb (5C8H1 or TNX-1500, NIH-NPRR, Cat# PR-1547, RRID:AB_2716324 or Tonix Pharmaceuticals, respectively, 25 mg/kg) two doses on day 0, then a single dose of 20 mg/kg on days 2, 5, 7, and 12. Daily solumedrol (Pfizer) was tapered to off during the first 30 days post-transplant and tacrolimus (Astellas) was administered intramuscularly daily for 60 days targeting trough levels of 8–13 ng/mL. Weekly anti-CD154, and daily mycophenolate-mofetil were continued as maintenance therapy (MMF, Genentech; 200 mg/day) [Fig. 1a]. Recipient erythropoiesis was supported by daily, three-times weekly, or weekly exogenous epoetin alfa (Epogen, Amgen) to maintain hemoglobin >9.5 g/dL. To minimize the chance of contamination of anti-pig natural antibodies, transfusion was performed only for hematocrit lower than 20%. Prophylaxis included ganciclovir (5 mg/kg) on days 0-90 and enrofloxacin until day 14 or trimethoprim/sulfamethoxazole continued throughout.

Animal weights were obtained weekly and a loss greater than 25% from pre-transplant was an end of study criteria. Clinical pathology (CBC, chemistries) was conducted using Catalyst Dx Chemistry Analyzer (Idexx) and HemaTrue Veterinary hematology analyzer (Heska) with blood collected on days 2, 5, 8, 12, and then every 1–2 weeks or if clinically indicated. Endocrine testing was performed by Michigan State University Veterinary Diagnostic Laboratory and consisted of clinical measurement of aldosterone, calcifediol, calcitriol, ionized calcium, parathyroid hormone (PTH), and parathyroid hormone-related peptide (PTHrP). For urine collection, a clean cotton/polyester mesh was placed over the pan at the bottom of the housing location of each recipient, urine was able to pass through without contamination of feces or food products. The animals were observed until urination, after which the liquid was collected and sent for analysis by Idexx BioAnalytics. In the perioperative period, xenograft ultrasonography was performed using a LogiQe Ultrasound System (GE Healthcare) on days 2, 5, 8, 12 and then every 1–2 weeks to monitor for vascular or ureteral complications and to monitor xenograft size. Xenograft length was measured in the long-axis on cross section and hydronephrosis described by the Society for Fetal Urology (SFU) grading schema³². Protocol xenograft biopsies were obtained every 2–4 months in recipients with stable function or for cause if a rise in serum creatinine occurred. Tissue was processed for light microscopy and another portion processed for RNA-seq as described below. At end of study, a complete necropsy was performed for histopathologic examination of the renal xenograft, lymph nodes, heart, lung, liver, pancreas, thymus, and skin. Parathyroid and thyroid tissue was obtained in the case of one long-term survivor (M2519). Additional tissue from the xenograft was processed for RNA-seq at necropsy.

Renin activity (plasma and serum) assays

The plasma renin activity (PRA) and serum renin activity (SRA) assays were based on previously reported methods³³. Briefly, citrated plasma or serum was mixed with 8-hydroxquinolone hemi-sulfate and EDTA and incubated for 0 or 6 h at 37 °C before the reaction was halted with BSA buffer resulting in a final dilution of 1:20. Production of Angiotensin-I (AngI) was measured using an Angiotensin-I ELISA kit (Novus Biological, NBP 262134) according to the manufacturer’s instructions with absorbance read on a FilterMax F5 microplate reader (Molecular Devices). Values below the level of detection were considered as zero pg/mL. PRA or SRA was calculated as AngI at 6 h – AngI at 0 h divided by length of incubation, e.g., pg/mL/6 h. Activity was not adjusted for dilution associated with the addition of reaction buffers.

Beta-C-terminal telopeptide (CTx) assay

NHP recipient serum pre- and post-transplant and pooled cynomolgus serum were diluted 1:5 and run in the Serum CrossLaps® beta-C-terminal telopeptide (CTx) ELISA (Immunodiagnosticsystems, AC-02F1) according to the manufacturer’s instructions with absorbance read on a FilterMax F5 microplate reader (Molecular Devices).

Isolation of RNA from porcine kidneys and Bulk Illumina RNA-sequencing (RNA-seq)

Samples were collected, snap-frozen, and cryopreserved from contralateral untransplanted kidneys (CUK) at time of organ procurement, biopsies at time of protocol biopsy or kidney at necropsy. On the day of protocol biopsy collection, the mean serum creatinine value was 0.7 mg/dL and mean blood urea nitrogen was 22.9 mg/dL indicating normal renal function. Sequencing runs were minimized to reduce batch effect. For a given batch, snap-frozen tissue was pulverized using a pre-chilled BioPulverizer (BioSpec, 59012MS). Total RNA was extracted using the Qiagen RNeasy kit (Qiagen, 74106) after passing the sample through homogenizing columns (Qiagen, 79656) and performing a DNAase digestion (Qiagen, 79256) according to manufacturer’s instructions. RNA integrity number and concentration were measured using an automated electronic electrophoresis system (Agilent, 4200 Tapestation) and reagents (Agilent, 5067-5578). For those samples passing quality checks, 500 ng of total RNA was carried into the Illumina Stranded mRNA kit library preparation (Illumina, 20040534). Libraries were characterized using the high-sensitivity dsDNA assay (Invitrogen, Q33231) for Qubit 3 (Invitrogen, Q33216). Libraries of appropriate concentration and insert length were diluted to loading concentration of 650 pM and loaded and sequenced using a P3 100 cycle kit (Illumina, 20040559) on Illumina NextSeq 2000.

Bioinformatic processing and computational analysis of Bulk Illumina RNA-sequencing

ENSEMBL Sscrofa11.1 FASTA assembly and v105 GTF annotation were prepared as a reference by removing unplaced contigs/scaffolds, and then adding intended genetic alterations/payload sequences as a separate contig. ENSEMBL Macaca fascicularis 6.0 FASTA assembly and v105 GTF annotations with unplaced contigs/scaffolds removed were appended to the pig reference for aligning reads from samples with NHP transcripts. Transcript sequences were generated with gffread (0.12.7)³⁴ on these synthetic genome assemblies. Indexing was performed on transcripts with salmon index (1.6.0)³⁵ applying standard settings and the whole genome as a decoy. Salmon quant in mapping-based mode was used to quantify RNA-seq read counts per transcript using custom settings: 100 bootstrap replicates, GC content and sequence bias correction applied. For this study, only transcripts aligning to Sscrofa11.1 were retained and analyzed. Gene level aggregation on transcript abundance was performed with tximport (1.22.0)³⁶. Protein coding and lncRNA genes were retained and normalized rlog counts calculated using DESeq2 (1.36.0)³⁷ with design including sample source (CUK, biopsy, necropsy) and pig identification number used as covariates. Heatmaps were created using ComplexHeatmap (2.12.1)³⁸. Pathway analysis was performed using pathfindR (1.6.4) on KEGG and GO pathways³⁹.

Single-cell isolation for single-cell RNA-sequencing (scRNA-seq)

Dissociation of kidney samples into single cell suspensions was performed using dissected cortex from necropsy and CUK samples or from approximately half of a biopsy sample. Tissue was minced into 1–2 mm³ pieces and digested with type IV collagenase (Sigma, C5138-500 mg; 1 mg/mL in 10 mL of Hank’s Buffered Salt Solution (HBSS) [ThermoFisher, MT-20023CV]). The resulting single-cell suspensions were passed through a 100 µm cell strainer (Miltenyi Biotec, 130-098-463), centrifuged and then passed through a 70 µm cell strainer (Miltenyi Biotec, 130-098-462) and centrifuged again. Red blood cell (RBC) lysis was performed by resuspending these pellets in 4 mL RBC lysis buffer (Roche, 11814389001) and placing on ice for 2 mins RBC lysis was quenched by adding 25 mL 20% fetal bovine serum in HBSS followed by centrifugation and the pellets resuspended in HBSS and filtered through 70 µm filters. A dead cell removal step was performed using a Dead Cell Removal Kit following the manufacturer’s instructions (Miltenyi Biotec, 130-090-101). After dead cell removal, cells were counted and resuspended at ~600–1200 cells/µL. These suspensions were applied to the 10X Genomics Chromium Next GEM Single Cell 3′ v3.1 kit to prepare scRNA-seq libraries, per manufacturer’s instructions (10X Genomics, 1000121).

Bioinformatic processing and computational analysis of 3′ scRNA-seq

Assembly construction and annotation were as described for bulk RNA-seq and was indexed here using STARsolo (2.7.9a) genomeGenerate on standard settings⁴⁰. 10 × 3′ scRNA-seq was quantified using custom settings for STARsolo –soloCellFilter EmptyDrops_CR, –soloMultiMappers EM, –clipAdapterType CellRanger4, –outFilterScoreMin 30, –soloCBmatchWLtype 1MM_multi_Nbase_pseudocounts, –soloType Gene, –soloUMIdedup 1MM_CR, –soloUMIfiltering MultiGeneUMI_CR, and –soloUMIlen 12. Counts were read into R (4.2.0) with DropletUtils (1.14.2)⁴¹ and cell quality checking and control performed using a set of filtering criteria: number of unique molecular identifiers (UMI), detected features, %mitochondrial UMI, %protein-coding and lncRNA UMIs, complexity, and doublet removal using scDblFinder (1.8.0)⁴². Count normalization was performed using scran (1.22.1) per recommendations for quickCluster, computerSumFactors, and logNormCounts⁴³. Cell-cycle analysis and annotation was performed with Seurat’s (4.1.0) CellCycleScoring⁴⁴. scVI was used to model and remove batch effect using covariates: %mitochondrial UMI, Seurat Cell Cycle S and G2M phase scores, pig donor identification, and kidney region (i.e., cortex vs medulla)⁴⁵. The 20 latent dimensions from scVI were passed to scran’s clusterCells and clustering performed using the bluster (1.4.0) Louvain algorithm on shared nearest neighbors (k = 3; https://www.bioconductor.org/packages/release/bioc/html/bluster.html). Uniform manifold approximation and projection (UMAP) was created via scater’s (1.22.0) runUMAP with 20 nearest neighbors and minimum distance 0.2⁴⁶. Cell types were assigned comparing known markers⁴⁷ and the molecular signature database⁴⁸ cell type signature enrichment from fgsea (1.20.0; for a function on top 50 features per cell cluster)⁴⁹ against cluster marker genes from scran’s scoreMarkers. UMAP plots were generated using scater’s ggcells and ggplot2 (3.2.0).

Statistics

For clinical data, generalized linear models were constructed with appropriate distribution (gaussian, negative binomial, or zero adjusted gamma) in gamlss (5.4.12) using glm. Fixed effects models were constructed by adding dummy variables for recipient-ID to models of xenograft length over time using glm. For RNA-seq, P-value adjustment for multiple comparisons was performed using the FDR method described by Benjamini-Hochberg and implemented in DESeq2. All analyses were performed using RStudio 2022.02.2 Build 485, which interfaces with R (4.2.0) and tests were two-tailed using an alpha of 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.