β–TC–6 and 293T (CRL–3605 and CRL–3216) cell lines were obtained from ATCC, EndoC–βH1 cell line was obtained from Human Cell Designer. 293T cells were maintained in DMEM (D6429, Sigma), supplemented with 10% FBS (S1520, Biowest), and penicillin/streptomycin (L0022, Biowest) in a 37°C incubator with 5% CO2. β–TC–6 cells were maintained in low bicarbonate DMEM (P04–03596, PAN Biotech), supplemented with 15% FBS (S1520, Biowest), and penicillin/streptomycin (L0022, Biowest) in a 37°C incubator with 5% CO2. EndoC–βH1 cells were maintained in Opti–β1 (Human Cell Design) after coating with βcoat (Human Cell Design) in a 37°C incubator with 5% CO2. All cell lines were routinely tested for mycoplasma contamination on a weekly basis and confirmed to be negative.

To generate NT, Hivep2 KO, Ins2 KO, and Rnls KO β–TC–6 cells, three gRNAs were selected from Brie mouse gRNA database (Supplementary Table 1) (34) and cloned into pLentiCRISPRv2 vector. After an initial screen of T7EI and TIDE analysis, one gRNA was selected for each gene as follows: NT gRNA (5′–GCGCTTCCGCGGCCCG–3′), Hivep2 gRNA (5′–TAAGGCGGATGACTCTCACA–3′), Ins2 gRNA (5′–GGACTCCCAGAGGAAGAGCA–3′), and Rnls gRNA (5′– CTACTCCTCTCGCTATGCTC–3′) were cloned into pLentiCRISPRv2 vector with puromycin antibiotic resistance gene (Addgene #98290) (35). The gRNA containing lentiviruses were used to establish these cell lines.

To generate NT, HIVEP2 KO, INS KO, and RNLS KO EndoC–βH1 cells, similarly, three gRNAs were selected from Brunello human gRNA database (Supplementary Table 1) (34) and cloned into pLentiCRISPRv2 vector. After an initial screen of T7EI and TIDE analysis, one gRNA was selected for each gene as follows: NT gRNA (5′–GACGGAGGCTAAGCGT–3′), HIVEP2 gRNA (5′–GACAAGATGTCAGACCTAGG–3′), INS gRNA (5′–GAAGCTCTCTACCTAGTGTG–3′), and RNLS gRNA (5′–TCCCACACAGCAAGGTACAA–3′) were cloned into pLentiCRISPRv2 vector with puromycin antibiotic resistance gene. All plasmid sequences were verified by Sanger sequencing before transduction and transfection (Supplementary Figure 1).

293T cells were seeded at 4.0 × 106 cells on a 10–cm plate 1 day before transfection. The cells were transfected with 2.5 μg of a vector, 2.25 μg of a transfer plasmid (psPAX2, #12260; Addgene), and 0.25 μg of a VSV–G vector (pMD2.G, #12259; Addgene) using the polyethylene imine (PEI) transfection reagent (Thermo Fisher Scientific, Waltham, MA). Viral particles were harvested 48 h post transfection and purified using a 0.45–μm Millex–HV filter (SLZHVR33RS; Merck Millipore, Burlington, MA).

For viral transduction, 5 × 10⁶ cells were seeded in 10–cm tissue–culture dishes and allowed to attach for 24 h. The cultures were then exposed overnight to 10 µL of 100–fold–concentrated lentiviral stock supplemented with 10 µg/mL protamine sulfate (P3369, Sigma–Aldrich) to enhance viral adsorption. Fresh growth medium was provided the next morning. Forty–eight hours later, robust fluorescence in parallel GFP–virus controls confirmed successful transduction, and antibiotic selection was initiated. β–TC–6 cultures received 4.5 µg/mL puromycin, whereas EndoC–βH1 cultures were treated with 2.5 µg/mL. Selection was maintained for four days, after which surviving cells were expanded for an additional two days in antibiotic–free medium to recover. All downstream assays were performed with these fully selected, puromycin–resistant populations.

CRISPR–Cas9 editing in both β–TC–6 and EndoC–βH1 cells were detected by T7 endonuclease I mismatch cleavage assay. Genomic DNA was purified from NT and gRNA transduced cells using the Macherey–Nagel™ NucleoSpin™ Tissue, Mini kit for DNA from cells and tissue. The Hivep2, Rnls, and insulin gRNA targeting site was amplified using the Phusion High–Fidelity PCR Kit. The primers for the gRNA site PCR were as follows: Hivep2 site forward, 5′–CTCCCTGCTGAGAAGTTGCC–3′; reverse, 5′– GATGTGGCTGTTCGGGTAAG –3′, Ins2 site forward, 5′–CGTGAAGTGGAGGACCCACA–3′; reverse, 5′–AAACTGTGGGTCCTCCACTTCACG–3′, Rnls site forward, 5′–TATCCCATGTGGCTTGGAGT–3′; reverse, 5′–GAGGTGAAGAATCGGTCCACT–3′, HIVEP2 site forward, 5′–GTTGTGTTGCCGAACTAGCC–3′; reverse, 5′–GGGGACAGGGTTGGGTATGA–3′, INS site forward, 5′–CCACCCTCTGATGTATCTCGG–3′; reverse, 5′–AGACTATAAAGCCAGCGGGG–3′, RNLS site forward, 5′–GAGTGGAATCAATCTTAGCAGTTG–3′; reverse, 5′–GAGTGGAATCAATCTTAGCAGTTG–3′. The 400 ng PCR products were used to form heteroduplexes by denaturing at 95°C for 5 min and then reannealing the products in a thermocycler using the following protocol: ramp down to 85°C at −2°C per s; ramp down to 25°C at −0.1°C per s; hold at 4°C. Ten units of T7 endonuclease I (#M0302) were added to the annealed PCR products and the reaction was incubated at 37°C for 15 min. The digestion reaction was stopped by 1 μl of 0.5 M EDTA and immediately applied to a 1% agarose gel to visualize digested and undigested products by electrophoresis.

β–TC–6 and EndoC–βH1 cells were treated with Macherey–Nagel™ NucleoSpin™ RNA, Mini kit for RNA purification for RNA extraction according to the manufacturer’s protocol. Purified RNA was reverse transcribed into complementary DNA (cDNA) using the iScript™ cDNA Synthesis Kit. Gene expression levels of genes were analyzed by QuantiNova Probe PCR Kit (Qiagen). The primers for the qPCR were as follows: Hivep2 site forward, 5′–GTTGAGTCTTGGACATTGAAGC–3′; reverse, 5′–CACAGGACATCAGAATACTACTGG–3′, Ins2 site forward, 5′–CGTGGCTTCTTCTACACACCCA–3′; reverse, 5′–TCCAGTGCCAAGGTCTGAAGGT–3′, Rnls site forward, 5′–AAGGCTGGGGATATAGGGGG–3′; reverse, 5′–GGGAGACTTCTGCACCTGAC–3′, HIVEP2 site forward, 5′–ACAGCCACTAGATTCCCAGA–3′; reverse, 5′–TTGTCCTGACGCTTCCTGTT–3′, INS site forward, 5′–AGGCCATCAAGCAGATCACT–3′; reverse, 5′–TTCCCCGCACACTAGGTAGA–3′, RNLS site forward, 5′–GGGAGCACGGAACCAAAGA–3′; reverse, 5′–AGTCATTCTTCCCCCTGAGTC–3′. The relative mRNA expression levels were normalized to GAPDH mRNA expression. All RT–qPCR assays were performed using a LightCycler^® 480 Instrument II (Roche Diagnostics).

Single–cell suspensions were prepared in complete growth medium (DMEM, 10% FBS, 1% penicillin/streptomycin) at 1 × 10⁷ cells/mL. Thirty–microliter droplets containing either 300, 500, or 1000 cells were dispensed onto the inner surface of 100–mm Petri–dish lids. Lids were inverted over dishes containing 10 mL sterile PBS to maintain humidity and incubated for 72 h at 37°C, 5% CO₂. Spheroids that formed at the bottom of the droplet were harvested by gentle pipetting with pre–warmed medium.

For higher throughput, homemade ULA dishes were fabricated by pouring 5 mL of sterile 1% (w/v) agarose in PBS into 100–mm Petri dishes, allowing the gel to solidify for 30 min, and rinsing once with growth medium. Subsequently, 5 × 10⁶ single cells in 10 mL medium were added per dish and cultured on an orbital shaker (50 rpm) for 3 days at 37°C, 5% CO₂. Gentle agitation promoted rapid self–assembly into compact spheroids with minimal size heterogeneity.

Spheroids from both protocols were washed twice with warm PBS and incubated for 3 min with fluorescein diacetate (FDA, 2 µg/mL) at 37°C to label metabolically active cells. Propidium iodide (PI, 1 µg/mL) was then added for an additional 5 min to stain membrane–compromised cells. After two PBS washes, spheroids were transferred to glass–bottom dishes and imaged immediately by epifluorescence microscopy (FITC and Texas Red filter sets). Aggregate diameters were measured from bright–field micrographs using ImageJ.

Immunofluorescence was performed to assess insulin production in both 2D monolayer cultures and 3D spheroids. For 2D cultures, EndoC-βH1 cells were fixed 72 h after seeding. For 3D cultures, spheroids were harvested by centrifugation at 100 × g for 5 min and the resulting pellets were fixed with 4% paraformaldehyde (PFA) for 15 min. Following fixation and washing with PBS, samples were permeabilized with 0.1% Triton X-100 and blocked with bovine serum albumin (BSA) to prevent non-specific binding. Samples were incubated overnight at 4°C with a polyclonal guinea pig anti-insulin primary antibody (1:50; PA1-26938, Thermo Fisher Scientific). After washing, samples were incubated with an Alexa Fluor 488-conjugated anti-guinea pig secondary antibody (1:250; ab150185, Abcam) for overnight at room temperature. Nuclei were counterstained with DAPI (1:5000; D1306, Thermo Fisher Scientific). Spheroids were transferred to 8-well chambered glass slides and imaged using a confocal laser scanning microscope.

For glucose–stimulated insulin secretion assay, the cells were plated in a 6–well plate as 1.5 × 106 cells per well. After 48h, cells were washed twice very gently using Krebs–Ringer Buffer (KRB). Before starting GSIS, cells were starved for 60 minutes by leaving cells in 2 mL of KRB buffer and incubating at 37°C, 5% CO2. After starvation, KRB buffer was removed and cells were incubated using 2.8 mM D–Glucose KRB (low glucose, LG) at 37°C, 5% CO2 for 60 minutes. Supernatant was collected and then cells were finally incubated using 16.7 mM D–glucose KRB (high glucose, HG) for 60 minutes at 37°C, 5% CO2, and the supernatant was collected. Supernatant was used to perform insulin quantification using the insulin ELISA kit from Mercodia.

For β–TC–6 the KRB contained (in mmol/L): 115 NaCl, 5.3 KCl, 2.6 CaCl₂, 1.2 MgSO₄, 1.0 KH₂PO₄, 24 NaHCO₃ and 10 HEPES, supplemented with 0.1% (w/v) fatty–acid–free BSA and either 1.67 mM (LG) or 16.7 mM (HG) glucose. The solution was equilibrated in 5% CO₂ to maintain pH 7.35–7.40.

For EndoC–βH1 the KRB containing (in mmol/L): 138 NaCl, 5.4 KCl, 2.6 CaCl₂, 0.5 MgCl₂, 5 NaHCO₃ and 10 HEPES, again with 0.1% fatty–acid–free BSA and either 2.8 mM (LG) or 28 mM (HG) glucose.

To enable in vivo bioluminescence tracking, both β–cell lines were first transduced with a lentiviral construct with the firefly luciferase (Fluc). Fluc coding sequence is driven by the human ubiquitin–C (ubC) promoter with a selection marker of Neomycin (G418) or hygromycin B (Supplementary Figure 14). Stable cell lines were obtained by antibiotic selection, 400 µg/mL G418 for β–TC–6 cells and 200 µg/mL hygromycin B for EndoC–βH1 cells, resulting in uniform Fluc expression. To ensure absolute parity in reporter expression across experimental groups, all gene-edited lines were generated from the same parental stable luciferase-expressing cell pool, thereby ensuring an identical reporter pedigree prior to CRISPR-Cas9 editing.

For in vivo studies, 5 × 10⁶ Fluc–cells from each knockout line, together with a NT control, were resuspended in 200 µL Matrigel (CLS354234, Merck) and injected subcutaneously into the left and right flank of CD–1 mice (designated Day 0). Bioluminescence imaging (BLI) was performed with a Caliper IVIS Spectrum system (PerkinElmer) following established protocols for the longitudinal monitoring of β-cell graft survival in murine models (36). Ten minutes before image acquisition, animals received an intraperitoneal injection of d–luciferin (150 mg/kg of animal; P1042, Promega) prepared at 15 mg/mL in calcium– and magnesium–free Dulbecco’s PBS and sterile–filtered through a 0.22–µm membrane. Imaging parameters were kept constant throughout the study (exposure = 1 min, binning = medium, f/stop = 1, field of view = 12.5 cm) to permit quantitative comparison. Photons emitted from the graft region were integrated with Living Image software and expressed as radiance (photons/scm²sr).

All in vivo procedures were approved by the Koç University Animal Research Facility Ethical Committee (KUARF, approval no. 2024.HADYEK.038) and conducted in accordance with the ARRIVE guidelines and Turkish Animal Welfare regulations. Female CD–1 or SCID mice (8–10 weeks, 25–30 g) were housed in individually ventilated cages (22 ± 2°C, 50 ± 10% humidity, 12 h light/dark cycle) with ad libitum access to standard chow and water.

For each genetic condition, NT control, Hivep2/HIVEP2 knockout, Rnls/RNLS knockout, and Ins2/INS knockout, n = 4 mice were used in both the allogeneic (β–TC–6 → CD–1) and xenogeneic (EndoC–βH1 → CD–1) cohorts. On day 0 mice were anesthetized with 2% isoflurane, and spheroids from 5 × 10⁶ luciferase–expressing cells suspended in 200 µL growth–factor–reduced Matrigel were injected subcutaneously into each flank (left, NT; right, gene–edited) and visualized via IVIS. Mice were euthanized by CO₂ inhalation followed by cervical dislocation.

One–way ANOVA with Tukey comparison test on GraphPad Prism Software (Graph–Pad Software version 10.5, CA, USA) was used. A value of p ≤ 0.05 was considered statistically significant. The significance levels are denoted as follows: n.s. for p > 0.05, * for p ≤ 0.05, ** for p ≤ 0.01, *** for p ≤ 0.001, and **** for p ≤ 0.0001.

Prior to the gene–editing experiments, 3D β–cell culture was optimized to obtain size–controlled, high–viability spheroids that more closely mimic the cell–cell interactions of native islets. The first approach relied on the hanging–drop technique. Single–cell suspensions of β–TC–6 cells were adjusted to 300, 500, or 1000 cells per 30 µL droplet, arrayed on the inverted lids of Petri dishes and cultured for 72 h at 37°C under 5% CO₂. During this period, cells settle at the bottom of the droplet due to gravity, where they aggregate into spheroids. Bright–field microscopy revealed a direct relationship between initial seeding density and final aggregate size. Droplets seeded with 300 cells produced spheroids averaging 120 µm in diameter, whereas those initiated with 1000 cells routinely exceeded 200 µm. Live/Dead staining demonstrated that aggregates below 150 µm were uniformly viable, while the larger spheroids displayed a central necrotic zone, consistent with oxygen– and nutrient–diffusion limitations (Figures 1A, B) (37). Although the method can be considered as simple, the hanging–drop format proved labor–intensive, and difficult to scale beyond a few hundred spheroids, making it impractical for the cell quantities required for transplantation.

To overcome these constraints, a suspension culture protocol was developed using home–made ultra–low–attachment dishes. Petri dishes were coated with 1% (w/v) agarose solution and allowed to solidify, creating a non–adhesive surface that prevents cell attachment. 4 × 10⁶ single cells were added to each dish in 8 mL of complete medium and cultured for 72 h on an orbital shaker set to 50 rpm. Under these conditions, cells aggregated spontaneously into hundreds of uniformly round spheroids (Figures 2A, B). Image analysis of 100 aggregates showed a narrow size distribution centered at 140 ± 8 µm, while confocal Live/Dead staining confirmed ≥ 95% viability throughout the spheroid cross–section (Figure 2C). Notably, the suspension approach yielded a higher number of spheroids per dish compared to the hanging–drop method, making it the preferred method for downstream experiments [Costa et al., 2018].

Given its superior throughput, precise size control, and absence of a necrotic core, the agarose–based suspension culture was selected for all subsequent CRISPR/Cas9 editing, functional assays, and transplantation studies. While the hanging–drop data provided a valuable reference for correlating spheroid diameter with diffusion limits, this information guided our decision to maintain aggregate diameters below 150 µm in all subsequent experiments.

To perform targeted gene knockouts, three guide RNAs (gRNA) were selected from the Brie and Brunello databases, including the guide RNA previously validated by Cai et al. as a control for Rnls knockout (31). The aim was to target exons as early as possible to maximize the probability of premature stop codons triggering nonsense-mediated decay (NMD) or producing truncated non-functional protein fragments, ensuring the disruption of the gene’s biological activity. As an internal control for functional knockout validation, the insulin gene (Ins2 for mouse, INS for human) was targeted in both cell lines. Each selected guide RNA was cloned into the pLentiCRISPRv2_puro plasmid, and successful insertion was confirmed by Sanger sequencing. Lentiviral particles were then produced by co–transfection of HEK293T cells with the guide RNA–cloned pLentiCRISPRv2_puro plasmids along with packaging (psPAX2) and envelope (VSV–g) plasmids. GFP–encoding lentivirus was produced similarly and used as a control to monitor infection efficiency. Following lentiviral production, β–TC–6 and EndoC–βH1 cell lines were infected with lentiviruses encoding gRNAs targeting RNLS, HIVEP2, and INS. After antibiotic selection to enrich for successfully transduced cells, cell pellets were collected and used for subsequent validation assays, including quantitative PCR (qPCR), genomic DNA isolation for T7 endonuclease I (T7E1), Tracking of Indels by Decomposition (TIDE) assays to confirm successful genomic editing at the intended target sites. TIDE analysis was specifically employed to quantify the ratio of frame-shift mutations, ensuring that the edited populations were dominated by out-of-frame indels that preclude the formation of functional proteins.

To validate successful genome editing in β–TC–6 cells, T7E1 mismatch assays were conducted for the Hivep2, Ins2, and Rnls gene targets. Distinct cleavage products were observed for gRNA2 targeting Hivep2, all three gRNAs targeting Ins2, and gRNA3 targeting Rnls, confirming the presence of CRISPR/Cas9–induced insertions and deletions (indels). No cleavage bands were detected in the corresponding non–targeting (NT) control lanes, confirming editing specificity (Figure 3A). Consistent with the T7E1 results, TIDE analysis of Hivep2–edited β–TC–6 cells using gRNA2 revealed editing efficiency of approximately 43.7%, further validating robust gene disruption at the target site (Figure 3B).

In parallel, genome editing in EndoC–βH1 cells was assessed using the same approach. T7E1 assays revealed clear cleavage products for all three gRNAs targeting HIVEP2, INS, and gRNA2 targeting RNLS, indicating efficient CRISPR–mediated gene knockout events (Figure 3C). TIDE analysis of the HIVEP2 locus edited with gRNA1 demonstrated a high indel frequency of approximately 94.3%, corroborating the T7E1 data and confirming moderate editing efficiency in human β–cells (Figure 3D).

To evaluate whether these high-frequency genomic indels resulted in corresponding changes at the transcript level, we conducted quantitative PCR (qPCR) analysis for the targeted loci (Supplementary Figure 2). The resulting data revealed a complex transcriptional landscape consistent with the phenomena of transcriptional adaptation or genetic compensation. While Rnls knockout in murine cells led to a significant reduction in mRNA levels, likely reflecting nonsense-mediated decay (NMD)-mediated transcript instability (38), we observed an unexpected upregulation of Hivep2 mRNA levels in murine cells. Such outcomes are indicative of cellular attempts to maintain genetic robustness by upregulating a mutant locus or its functional homologs following the loss of a functional protein product, even when the transcripts carry frame-shift mutations that preclude functional protein synthesis. Given the potential for these complex feedback loops to confound mRNA quantification, we prioritized high-resolution genomic indel analysis and functional physiological assays as the most definitive evidence of successful gene inactivation in this study.

Glucose−stimulated insulin secretion (GSIS) was next examined to determine whether gene editing affected β–cell function (Figure 6). In mouse β−TC−6 cells, the NT control exhibited a stimulation index (SI) close to 1, indicating that these cells are not responsive to different glucose challenges. Disruption of Hivep2 had no impact, yielding the SI statistically non-significant from control. Additionally, the knockout of Ins2 resulted in the sharp decrease of insulin production. However, SI was calculated as 1 because the ratio of insulin at high glucose over low glucose is similar due to the disrupted insulin production mechanism (excluding unresponsiveness of murine β–cell). Notably, Rnls−edited cells showed significantly reduced SI compared to control cells, indicating that loss of Rnls compromises GSIS in the murine line (Figure 6A).

Human EndoC−βH1 cells showed the much higher SI characteristic of this more mature model. Again, HIVEP2 deletion left glucose responsiveness intact, whereas INS knockout dramatically suppressed secretion. In contrast to the mouse data, RNLS editing produced only a modest, statistically non−significant reduction in SI, suggesting that loss of RNLS does not compromise insulin secretion of human β–cells (Figure 6B). Together, these functional assays reinforce the idea that HIVEP2 deletion is not dampen insulin secretion, while RNLS loss exerts a cell−line−dependent decrease in insulin secretion, prominent in mouse β−TC−6 cells.

Stimulation index (ratio of insulin released at HG versus LG; mean ± SD, 30 spheroids per independent experiment) for A) murine β–TC–6 (n=5 independent experiment) and B) human EndoC–βH1 spheroids (n=3 independent experiment) carrying non–targeting (NT) or gene–specific guides. In mouse cells, cells showed no response against different glucose challenges, whereas Hivep2 deletion was neutral and Rnls deletion produced a moderate but significant reduction (P < 0.05) relative to NT. In human cells, INS knockout again eliminated GSIS (*** P < 0.001) while neither HIVEP2 nor RNLS knockout altered the stimulation index (ns, not significant).

Before initiating the gene-knockout in vivo transplantation experiments, a pilot experiment was done to verify whether EndoC–βH1 spheroids can survive long-term in an immunodeficient environment. Luciferase-labelled spheroids (5 × 10⁶ cells, Matrigel embedded) were implanted subcutaneously into the right flank of either CD-1 mice (fully immunocompetent) or severe combined immunodeficiency disease (SCID) mice (T and B cell deficient). Longitudinal bioluminescence imaging (d-luciferin, 150 mg/kg) revealed rapid loss of graft signal in CD-1 animals, falling below background by day 7, whereas identical grafts in scid mice persisted and expanded over a 35-day observation period (Figure 7). These data confirm that innate immune mechanisms alone are insufficient to eliminate the graft, and that adaptive immunity is the dominant driver of xenogeneic rejection in our system.

Bioluminescence images of a CD-1 mouse (top) and a scid mouse (bottom) bearing identical luciferase-expressing EndoC–βH1 spheroids. Corresponding photon-flux curves (right) plot residual graft viability as a percentage of day-0 signals. In CD-1 mice the xenograft is rapidly rejected, with complete loss of luminescence by day 7. In contrast, the same graft in scid mice persists and progressively enlarges, demonstrating that adaptive immunity is required for xenogeneic clearance in this model.

To evaluate whether targeted gene knockouts influence β–cell survival in an allogeneic transplantation context, luciferase–labelled β–TC–6 cells genetically edited for Hivep2, Ins2, and Rnls, as well as NT controls, were transplanted subcutaneously into the flank region of immunocompetent CD1 mice using Matrigel as a matrix (Figure 8). Cellular survival was assessed through bioluminescence imaging (BLI) following intraperitoneal administration of D–luciferin. The Hivep2 knockout cells followed a survival trajectory similar to the NT controls, resulting in complete signal loss by day 7. However, a modest and transient improvement in early post-transplant protection was observed during the initial four days, as evidenced by a slightly higher residual radiance compared to controls. Despite this early trend, it remains a robust observation that single Hivep2 deletion does not significantly extend long-term graft survival, as the protective effect was rapidly overwhelmed by the allogeneic immune response.

Interestingly, Rnls knockout cells exhibited prolonged survival relative to Hivep2 knockout cells, maintaining comparable bioluminescence intensity through day 10 and exhibiting a slower decline thereafter. These findings collectively reveal that both Rnls and Hivep2 did not affect β–cell survival in allogeneic transplantation, indicating that Rnls protective effect on autoimmune attack does not translate in allorejection, which is consistent with previously reported findings by Cai et al. (31).

Luciferase–expressing spheroids (5 × 10⁶ cell per flank, embedded in Matrigel) were implanted subcutaneously on day 0; left flanks carried a non–targeting (NT) control, right flanks received a gene–edited counterpart. Serial ventral images (days 0–14; identical color scale) are shown for Hivep2 (top) and Rnls (bottom) knockouts. Line plots on the right depict residual photon flux, expressed as percentage of day–0 signal, for NT (black) versus each edit (orange). All grafts declined progressively and were undetectable by day 14; neither HIVEP2 nor RNLS deletion altered the clearance curve.

To test the effect of gene knockouts in xenotransplantation, luciferase−labelled EndoC–βH1 cells, harboring HIVEP2 or RNLS knockouts or a NT control, were embedded in Matrigel and implanted into flank region of immunocompetent CD−1 mice (Figure 9). Across all groups, signal decay was markedly faster than in the allogeneic β−TC−6 transplantation, reflecting the increased severity of xenorejection. HIVEP2−edited grafts lost >80 % of their luminescence by day 4 and were undetectable by day 7, indicating no xenoprotective effect.

RNLS−edited grafts displayed a modest delay in clearance, retaining ~30 % of baseline signal on day 7 versus <10 % in controls, but were nonetheless fully rejected by day 10. Thus, in the xenogeneic CD−1 model, none of the single−gene edits conferred durable protection to human β–cells. These findings underline the greater immunological hurdle posed by xenotransplantation and suggest that more extensive engineering, potentially combining RNLS or HIVEP2 loss with additional immune−evasion strategies, will be required to achieve long−term engraftment of human β–cells in fully immunocompetent hosts.

Luciferase–labelled human EndoC–βH1 spheroids (5 × 10⁶ cells per flank, Matrigel embedded) were implanted subcutaneously on day 0; the left flank received a non–targeting (NT) control, the right flank the indicated CRISPR knockout. Sequential bioluminescence images are shown for HIVEP2 (top) and RNLS (bottom) edits, together with quantitative photon–flux decay curves (right; percentage of day–0 signal). All grafts, irrespective of genotype, lost >80% of baseline radiance by day 4 and were completely eliminated by day 7, illustrating the innate and adaptive barriers to human–to–mouse xenotransplantation. Neither RNLS nor HIVEP2 knockout conferred any detectable survival advantage and did not alter rejection kinetics.