Targeting AKR1B10 by Drug Repurposing with Epalrestat Overcomes Chemoresistance in Non–Small Cell Lung Cancer Patient-Derived Tumor Organoids

More than 50% of surgically treated patients with non-metastatic non–small cell lung cancer (NSCLC) develop recurrences later on (1) that are directly linked to drug-resistant, micrometastatic cancer cells that are targets for perioperative systemic treatments (2). Patients with NSCLC undergoing surgery have been routinely administered adjuvant platinum-based doublet cytotoxic chemotherapy to eradicate this micrometastatic disease, but this therapy leads to ∼5% survival benefit only (1). Newer targeted agents are benefitting a small fraction of patients with NSCLC with lung adenocarcinoma (LUAC) subtypes (3), while personalized treatment options for lung squamous cell carcinoma (LUSC) subtypes are absent. In addition, these newer targeted (3) and immune therapies (4) lack sustained effects due to inherent or acquired drug resistance. Identifying targets to overcome drug resistance holds great promise to prevent lethal recurrences in patients with NSCLC (1). Exploring gene networks and molecular pathways for the identification of agents that are approved for other cancers or diseases can repurpose drugs with known pharmacological profiles that can be advanced efficiently into phase II and III clinical trials (5).

A major obstacle in determining drug resistances has been the inability to reliably culture viable patients’ tissues ex vivo. Patient-derived xenograft (PDX) mouse tumor models have been used widely to study drug resistance patterns in NSCLC (6). However, PDX models are inefficient with low engraftment rates and lack the tumor microenvironment. Patient-derived tumor organoids (PDTO) are self-organizing and grow efficiently in a three-dimensional (3D) matrix and retain tumor microenvironmental cells in early passages (7, 8) with ongoing investigations to preserve/mimic the tumor microenvironment (e.g., co-culture with autologous peripheral blood lymphocytes) (9). The drug responses observed in PDTOs also match the ones observed in patients with cancer and PDX models in vivo (7, 8).

In the present study, we established and molecularly characterized PDTOs from surgically resected tumor tissues of patients with non-metastatic NSCLC. Drug resistance toward standard-of-care doublet platinum-based chemotherapy was determined and associated with druggable target aldo-keto reductase 1B10 (AKR1B10). Although AKR1B10 overexpression has been shown to correlate with smoking in NSCLC (10), its role in chemotherapy resistance is understudied. Moreover, investigations using cohorts of patients with NSCLC to elucidate the role of AKR1B10 in promoting resistance to commonly used cytotoxic agents are scarce and warrant further exploration for potential clinical translation (11). Based on our observation of AKR1B10 overexpression in chemoresistant PDTOs, an already clinically used AKR1B10 inhibitor, epalrestat, was repurposed to overcome chemoresistance in PDTOs. Proven pharmacokinetics, oral availability, rapid absorption, and distribution, efficacy, and safety with minimal toxicity underline the near-term translational relevance of repurposing epalrestat in clinical trials in patients with drug-resistant NSCLC (12–14).

Detailed methodology is described in the Supplementary Methods S1.

The study was approved by the University of Missouri Institutional Review Board (IRB#: 2010166) and conducted in accordance with the Declaration of Helsinki. Study subjects were prospectively recruited at Ellis Fischel Cancer Center at the University of Missouri between June and September 2021. All patients gave written informed consent. The study was registered at ClinicalTrials.gov (Identifier: NCT02838836; date of registration: July 20, 2016). Patients with treatment-naïve NSCLC were enrolled, whereas patients with concurrent occurrence of another malignancy were excluded. Staging was performed based on the TNM staging manual of the American Joint Committee on Cancer (AJCC), eighth edition. All patients recruited were eligible for surgical resection for non-metastatic, loco-regional stages (I–IIIA) as determined by a multidisciplinary thoracic oncology team. All lung cancer resections were performed by specialty-trained thoracic surgeons in alignment with the International Association for the Study of Lung Cancer and Union for International Cancer Control. Clinical and outcome data were gathered by reviewing the hospital electronic medical records from our cancer survivorship program or direct communication with the patients or families or their treating physicians.

Surgically resected tumor tissues from patients with treatment-naïve NSCLC were enzymatically digested and single cells were seeded in Matrigel (Corning, 354230) to generate PDTOs that were extensively characterized as outlined below.

Histology by hematoxylin & eosin (H&E) staining and pathological biomarker expression analysis by immunohistochemistry (IHC) was performed using patient matched tumor tissues and PDTOs. In all cases, IHC staining was scored by a pathologist blinded to the study. Staining intensity was scored using the 0 to 3 scale where scores of 0 or 1 were considered negative, 2 was moderate positive, and 3 was strong positive. To qualify for 2 and 3 scores, staining of more than 10% of tumor cells had to be observed.

Immunofluorescence staining was performed in patient matched tumor tissues and PDTOs to determine the cellular heterogeneity and abundance of different cell types.

Tumor organoids from passage-2 were treated in triplicates with carboplatin (1 μg/mL) and paclitaxel (0.5 μg/mL) after determining doses by titration. On days 0 (treatment initiation), 3, and 6 (treatment termination) images were taken with the z-stack method (15). Percentage of growth/regression and viability (3D CellTiter-Glo; Promega) was determined against vehicle controls. Chemoresistant PDTOs were treated with carboplatin/paclitaxel with/without epalrestat (SML0527; Sigma-Aldrich). Based on dose titration, a noncytotoxic epalrestat concentration of 319.4 ng/mL (1 μmol/L) was used for PDTO testing. Organoid growth measurements, 3D cell viability assay, and IHC staining for proliferation marker Ki67 (rabbit, ab16667, 1:200; Abcam, RRID: AB_302459) and apoptosis marker cleaved caspase 3 (Rabbit, 9664, 1:100; Cell Signaling Technology, RRID: AB_2070042) determined the therapeutic effect of epalrestat.

Genomic DNA extracted from PDTOs and matched primary lung was used for library preparation. The library was quantified with Qubit (RRID: SCR_018095), pooled and sequenced on Illumina platforms with PE150 strategy to 30× coverage depth.

Clean reads were mapped to the reference genome (b37/hg19/hg38) by Burrows-Wheeler Aligner software (16). GATK was used to call SNPs and InDels and ANNOVAR was used to perform variant annotation (17). Cancer-relevant somatic mutations were identified by filtering against mutations reported in COSMIC (18).

Total RNA from PDTOs and matched primary lung tumor tissues were extracted and used for Illumina stranded library preparation and sequencing. Libraries were sequenced on NovaSeq 6000 S2 PE100 flow cell to 50× coverage depth. Following MultiQC, FASTQ files were aligned to human reference genome (GRCh38) using STAR alignment method (19). Principal component analysis (PCA) was used to determine variability between samples. Differentially expressed genes were identified by DESeq2 R package (20) and cellular populations quantified by microenvironment cell populations (MCP) counter method (21). Gene and pathway enrichment analysis was performed (22).

Real-time qPCR and IHC were performed to validate AKR1B10 expression in PDTO-matched tumor tissues and an additional internal validation cohort of chemoresistant and chemosensitive patient tumors.

For external validation, RNA sequencing normalized counts of AKR1B10 gene expression were determined using The Cancer Genome Atlas (TCGA).

NSCLC LUAC (n = 6) and LUSC (n = 6) cell lines were used to study the correlation of AKR1B10 expression with chemosensitivity and the impact of AKR1B10 knockdown on chemotherapy outcome. LUAC cell lines A549 (RRID: CVCL_0023), H1437 (RRID: CVCL_1472), H1573 (RRID: CVCL_1478), H838 (RRID: CVCL_1594), H1373 (RRID: CVCL_1465), and HCC827 (RRID: CVCL_2063) and LUSC cell lines H647 (RRID: CVCL_1574), SKMES1 (RRID: CVCL_0630), H226 (RRID: CVCL_1544), H1703 (RRID: CVCL_1490), H2170 (RRID: CVCL_1535), and SW900 (RRID: CVCL_1731) were used. All LUAC (except A549) and LUSC cell lines were cultured and maintained in complete RPMI medium with 10% FBS and 1% penicillin-streptomycin. A549 cell line was cultured and maintained in complete DMEM medium with 10% FBS and 1% penicillin-streptomycin. Cell lines were originally obtained from ATCC and authenticated by short tandem repeat sequencing. Cell lines used were between passages 5 and 10, and mycoplasma testing was performed by standard PCR just before freezing at every passage. All cell lines were passaged for at least two generations before using for cell proliferation assay or Western blotting.

All animal research protocols were reviewed and approved by the Institutional Animal Care and Use Committee (ACUC Protocol # 37592). A549 cells were subcutaneously injected in NSG mice (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ, RRID: IMSR_JAX:005557). Mice were administered carboplatin plus paclitaxel doublet therapy or cisplatin monotherapy without or with epalrestat. Subcutaneous tumors from all the experimental groups were used for mass spectrometric analysis to determine the intratumoral concentration of epalrestat.

Statistical analyses were performed using one-way ANOVA (Kruskal–Wallis and Tukey’s multiple comparisons test) and Student t test (Multiple t tests). Receiver operating characteristic (ROC) and area under the curve (AUC) analyses were performed using Prism v8.00; GraphPad. A P value of <0.05 was considered significant. Bioinformatical analysis is outlined in Supplementary Methods.

All raw whole exome and RNA sequencing data were deposited in NCBI Sequence Read Archive (SRA) and are accessible through project numbers PRJNA1020294 and PRJNA935652, respectively. Data are also provided within the manuscript and Supplementary Information Files. All other datasets generated and analyzed in the current study are available from the corresponding authors upon request.

Ten patients with treatment-naïve NSCLC with non-metastatic stages I–IIIA undergoing surgical resection were prospectively enrolled and observed for ≥2.5 years (Table 1). All resected tumor tissues were successfully established as PDTOs (Supplementary Fig. S1A) within a median of 12 days (range 4–16; Fig. 1A). Tumoroid growths were determined with 3D z-stack imaging (Fig. 1; Supplementary Fig. S1B). Histological architecture and pathological biomarker expressions of primary LUAC (CK7, TTF-1) and LUSC (CK5/6, P40) were consistently preserved in PDTOs (Supplementary Fig. S1C–L).

Mutational signatures of six matched pairs of PDTOs/tumor tissues were determined by whole exome sequencing (Supplementary Fig. S2). Hierarchical clustering based on somatic mutations demonstrated strong similarities between matched tumor tissues/PDTOs (Supplementary Fig. S2A). Strong overlap of cancer-relevant somatic mutations (such as missense/nonsense mutations, frameshift indels) confirmed conserved genomic fidelity of PDTOs (Supplementary Fig. S2B; Supplementary Files S1 and S2). Somatic mutations identified in oncogenes, tumor suppressor genes, and relevant pan-cancer and NSCLC-specific genes broadly overlapped (Supplementary Fig. S2C–E; Supplementary Table S1).

Whole transcriptome analysis of matched PDTOs/tumor tissues determined the global gene expression and cellular heterogeneity profiles (Supplementary Fig. S3). PCA involving transcriptome of PDTOs and matched tumor tissues along with PDX tumors derived from an independent set of samples revealed close clustering of PDTOs and tumor tissues (Supplementary Fig. S3A). As expected, in the PCA unmatched PDX tumors that were generated from a different set of samples from patients with NSCLC did not cluster closely with the PDTOs and their matched tumor tissues. PCA findings were further substantiated by Pearson’s correlation between PDTOs and matched tumor tissues (Supplementary Fig. S3B). Cell type abundance was determined using the MCP counter method (Supplementary Fig. S4). The epithelial cell composition was conserved in all PDTOs compared to matched tumor tissues, except for patient MU380. While immune, stromal, and endothelial cell type abundances were retained in some PDTOs, these tumor microenvironmental cells were not preserved in all PDTOs (Supplementary Fig. S4). MCP counter findings were further validated by multiplex immunofluorescence staining for immune, stromal, and endothelial cells (Supplementary Fig. S5). Epithelial cell subtype-specific marker (Supplementary Table S2) expression analysis confirmed the presence of epithelial cell types predominantly found in LUAC and LUSC, respectively. As such, alveolar epithelial cell type II (AT2) was identified in all LUAC-derived PDTOs and matched tumor tissues (Supplementary Fig. S6). Similarly, basal cells were present in all LUSC-derived PDTOs and matched tumor tissues (Supplementary Fig. S6).

PDTOs were treated with a standard-of-care, platinum-based doublet chemotherapy (carboplatin/paclitaxel) regimen given routinely perioperatively to patients with NSCLC (Fig. 1B). Doses of carboplatin and paclitaxel were based on dose titration curve determined using the first three (out of a total of 10) PDTOs used in the study (Supplementary Fig. S7). Responses to chemotherapy in PDTOs were monitored (Fig. 1C–L left) and quantified based on growth changes (Fig. 1C–L middle) and cell viability (Fig. 1C–L right) on days 3 and 6 in comparison to vehicle controls. PDTOs were categorized as chemosensitive if statistically significant reduction (P < 0.05) of growth and viability was observed. Five (50%) PDTOs were found to be chemoresistant (Fig. 1C–L). As the radiographically visible tumor was surgically resected at the time of administration of adjuvant chemotherapy (Table 1), response rates observed in PDTOs could not be directly compared to the primary lung tumor in the patients. However, we observed a correlation between PDTOs and matched patient recurrence patterns in three patients (MU375/MU380/MU383) who received adjuvant platinum-based doublet chemotherapy. Patient MU383 developed tumor recurrences while receiving adjuvant carboplatin/paclitaxel, and patient MU375 developed tumor recurrences 24 months after surgery following adjuvant cisplatin/pemetrexed. Both PDTOs from MU375 and MU383 were chemoresistant, aligning with the patients’ tumor recurrences despite chemotherapy treatments. In contrast, MU380 PDTO was chemosensitive, reflecting the patient’s clinical course as he remained recurrence-free for 31 months following surgery and chemotherapy (Fig. 1; Table 1).

Differential gene expression analysis showed significant upregulation of several genes in chemoresistant PDTOs (Fig. 2A; Supplementary File S3). Gene and pathway enrichment analysis of these upregulated genes revealed enrichment of metabolite conversion enzymes (Fig. 2B) and biotransformation pathways (Fig. 2C) that are critical in detoxification of cytotoxic agents. These enrichments were due to upregulated AKR1B10 and AKR1B15 in chemoresistant PDTOs. In alignment with gene expression, AKR1B10 protein expression was high in resistant PDTOs versus sensitive PDTOs (Supplementary Fig. S8A–C). The AKR1B family of enzymes is known to have a role in detoxification of cytotoxic agents leading to resistance toward platinum- or taxol-based drugs (23). An inhibitor of AKR1B10/15, epalrestat, was identified for drug repurposing, as epalrestat is already in clinical use in Japan for patients with diabetic polyneuropathy (13), and in the United States a phase III clinical trial in pediatric patients with phosphomannomutase 2-congenital disorder of glycosylation is ongoing (NCT04925960). Epalrestat can be given orally with low toxicity profile (24). The dose of epalrestat for PDTO drug testing was based on dose titration curves (Supplementary Fig. S9A–C). Treatment of chemoresistant PDTOs with epalrestat media concentration of 1 μmol/L (319.4 ng/mL) demonstrated significant sensitization to carboplatin/paclitaxel (Fig. 2) evidenced by the inhibition of PDTO growth and viability (P < 0.05; Fig. 2D–F). Epalrestat alone did not have any significant effects on PDTO growth. IHC for cell proliferation (Ki67) and apoptosis (cleaved caspase-3) markers revealed a significant reduction in number of proliferating cells and increased number of apoptotic cells in PDTOs treated with epalrestat and carboplatin/paclitaxel compared to other groups (P < 0.05; Fig. 2G–I). To determine the effective concentration (EC₅₀) of carboplatin/paclitaxel doublet drug concentrations in PDTOs, we tested increasing concentrations in PDTOs where we observed matched responses in the patients (MU375/MU380/MU383). EC₅₀ of the platinum-doublet was found to be 3- to 6-fold higher in resistant PDTOs MU375/MU383 versus sensitive PDTO MU380 (Supplementary Fig. S10). This fold difference of the EC₅₀ also correlated to the AKR1B10 expression in chemoresistant versus chemosensitive PDTOs (Supplementary Fig. S8). Further, sensitization of resistant PDTOs MU375/MU383 by epalrestat to different concentrations of platinum-doublet treatment was correlated with AKR1B10 expression (Supplementary Figs. S11 and S8).

Confirming findings in chemoresistant PDTOs, matched tumor tissues of patients with NSCLC also showed overexpression of AKR1B10 determined by differential gene expression analysis (Fig. 3A; Supplementary File S4). Further, statistically significant AKR1B10 gene and protein overexpression was confirmed in matched tumor tissues (Fig. 3B–D; Supplementary Fig. S8). Additionally, an internal validation set of resistant (n = 7) and sensitive (n = 7) patients’ NSCLC primary tumors to platinum-based doublet chemotherapy also showed differential AKR1B10 protein expression. The majority (85.71%) of resistant tumors showed high expression of AKR1B10, whereas all sensitive tumors showed no expression of AKR1B10 (Fig. 3E and F; Supplementary Fig. S12A and S12B; Supplementary Table S3). To determine the clinical utility of AKR1B10 as a biomarker to stratify patients as responders and nonresponders, we performed ROC and AUC analysis (Fig. 3G). AKR1B10 IHC score cut off ≥2 showed high sensitivity (85.71%), specificity (100%), and AUC (0.93; Fig. 3G). To further validate the AKR1B10 expression pattern in an external, larger cohort of patients with NSCLC, we analyzed TCGA data. Substantial AKR1B10 overexpression was observed in both LUAC/LUSC tissues of patients with NSCLC (n = 1,048) compared to normal lung (n = 110; Fig. 3H). Subsequently, 59.0% of the NSCLC tissues had >200 AKR1B10 transcripts per million (TPM) base pairs, with 39.4% of patients having very high (>2,000 TPM) AKR1B10 expression (Fig. 3I).

To identify the mechanism through which AKR1B10 promotes chemotherapy resistance, we tested a library of 12 NSCLC cell lines (n = 6 LUAC and LUSC, respectively). Differential expression of AKR1B10 in cell lines correlated with nuclear factor erythroid 2 (NRF2) expression (Fig. 3J and K). We categorized AKR1B10 high-, medium-, and non-expressing cell lines and performed titration with commonly used cytotoxic agents in NSCLC (Supplementary Fig. S13A and S13B; Supplementary Table S4). There was a positive correlation between AKR1B10 expression and IC₅₀ values of carboplatin, cisplatin, and paclitaxel, whereas there was no correlation with the IC₅₀ values of gemcitabine and pemetrexed (Fig. 3L; Supplementary Fig. S13A and S13B; Supplementary Table S4). shRNA-mediated knockdown of AKR1B10 in high-expressing human NSCLC cell line A549 (Fig. 3M and N) showed a significant difference in cell viability against IC₂₅ concentration of carboplatin and cisplatin (Fig. 3O; Supplementary Fig. S14). A trend in differential cell viability was observed against paclitaxel, yet not statistically significant (Fig. 3O; Supplementary Fig. S14). Knockdown of AKR1B10 showed no difference in A549 response to gemcitabine and pemetrexed (Fig. 3O; Supplementary Fig. S14). These results confirmed the role of AKR1B10 in promoting resistance to platinum-based drugs in NSCLC. Finally, the potential of epalrestat to sensitize NSCLC to chemotherapy was validated in vivo by xenografting A549 NSCLC cells to immunodeficient mice (Fig. 4A). Mice injected with A549 cells were treated with carboplatin-paclitaxel doublet therapy or cisplatin monotherapy followed by oral or intraperitoneal administration of epalrestat. Epalrestat was effective in sensitizing A549 xenograft tumors to both platinum-based doublet (Fig. 4B–E) or monotherapy (Fig. 4F–I). As determined by mass spectrometry, epalrestat administered twice at a concentration of 10 mg/kg body weight via intraperitoneal route resulted in a mean intratumoral concentration of 41.92 ± 12.62 μg/g, whereas oral epalrestat resulted in intratumoral concentrations of 39.92 ± 6.58 μg/g (Fig. 4E and I). The inhibition of tumor growth by platinum-based doublet (Fig. 4E) or monotherapy (Fig. 4I) was directly proportional to the intratumoral epalrestat concentration.

Accurate prediction of personalized drug responses is essential to reduce the persistently high recurrence rates in patients with surgically resected non-metastatic NSCLC (1). Lack of high-throughput screening platforms for standard-of-care, repurposed and novel drugs is a clinical gap in improving NSCLC patient outcomes (7, 25). In the present study, we prospectively generated PDTOs from patients with treatment-naïve NSCLC undergoing curative surgery for non-metastatic disease. We observed PDTOs to be efficient and viable tumor-derived models that conserve histopathological features, genomic fidelity, transcriptomic and cellular signatures that are close to the matched primary lung tumor tissues. As demonstrated by other groups, PDTOs reflect tumor heterogeneity, genetic alterations, metabolism, and drug and radiation response (8, 26). All our NSCLC PDTOs conserved LUAC/LUSC subtype-specific histomorphology and pathological biomarker expressions. A critical component in preclinical validation of tumor-derived models is the need for sustained conservation of complex genetic polymorphisms and mutational heterogeneity, cellular, molecular, and other biological properties of the matched parental tumors. The genomic landscape between primary tumors and PDTOs was retained, identifying somatic variants commonly found in our rural, Midwestern America NSCLC study population consisting of heavy smokers (27). Our PDTOs also recapitulated the transcriptomic signature of matched tumor tissues. Concerning epithelial cancer cell heterogeneity, LUAC- and LUSC-derived PDTOs showed abundance of AT2 cells and basal cells, respectively, that are cancer histology–specific epithelial subtypes contributing to NSCLC onset/progression.

PDTOs have been shown to serve as efficient ex vivo platforms that reflect drug responses in matched patients while conserving components of the tumor microenvironment (28). We applied a dynamic 3D growth analysis method via bright-field microscopy z-stack imaging including cell viability of PDTOs to measure the real-time growth rates and the impact of drugs on PDTO growth over 6 days (15). Blockade of pro-tumorigenic mechanisms and pathways can lead to increased drug sensitivity, possibly resulting in potentiation of therapeutic efficacy of drugs. To identify druggable targets, we screened PDTOs and tumor tissues by whole transcriptome differential gene expression analysis. AKR1B10 was upregulated in chemoresistant PDTOs. AKR1B10 overexpression has been reported in some solid cancers, including NSCLC (23). Moreover, the AKR1B family of enzymes is known for their role in detoxifying platinum- and taxol-based compounds, causing them to be ineffective (23). The observed impact of AKR1B10 in promoting resistance to platinum-based drugs in our study aligns with previous reports (23). Further, the strong correlation of AKR1B10 expression with NRF2 expression observed in NSCLC cell lines indicates the involvement of the well-studied NRF2 pathway in regulating AKR1B10 expression (29, 30). AKR1B10 upregulation in chemoresistant PDTOs provided a druggable target for further study to overcome resistance to carboplatin/paclitaxel in PDTOs, as the AKR1B10 inhibitor epalrestat is a drug approved in Japan for patients with diabetic neuropathy (13). While epalrestat is not FDA-approved yet, in the United States there is an ongoing phase III clinical trial in pediatric patients with phosphomannomutase 2-congenital disorder of glycosylation (NCT04925960). Importantly, epalrestat can be given orally and has clinically proven safety with a low toxicity profile (24). A pharmacokinetic study performed in healthy subjects revealed a favorable, rapid biodistribution of epalrestat with an elimination half-life of a maximum of 2 hours (12). A bioequivalence study of 50 mg oral epalrestat revealed peak plasma concentrations of >2,500 ng/mL within 2 hours (12). In our study, the epalrestat media concentration in NSCLC PDTO required to overcome chemoresistance was much lower (319.4 ng/mL). We also demonstrated that an intratumoral epalrestat concentration of ∼40 μg/g of tumor was effective in sensitizing NSCLC xenograft tumors to platinum-based therapy. Our results on PDTO and intratumoral epalrestat concentrations indicate that the standard and well-tolerated 3 × 50 mg epalrestat dosage given to diabetic neuropathy patients is likely sufficient to achieve effective bioavailability and intratumoral drug concentrations in a future clinical trial in patients with NSCLC. In a phase II clinical trial evaluating epalrestat in the treatment of metastatic triple-negative breast cancer, the standard oral dosage of 3 × 50 mg was chosen (NCT03244358). The safety/efficacy of epalrestat has been demonstrated in two independent studies in short-term (12 hours; single-center) and long-term (3 years; multicenter) study settings (12, 13). While short-term administration is associated with minor grade 1 adverse effects (such as nausea, skin rashes, lower extremity weakness/edema, lightheadedness, vertigo), epalrestat has also been well tolerated in clinical trials with up to 3-year duration with no severe (grade 3–5) adverse effects even upon long-term administration (13). As another potential benefit, epalrestat penetrates the brain (24) as one of the most common and devastating metastatic sites of NSCLC.

One of the limitations of the present study is that the surgical resection of radiographically visible tumors prior to the administration of adjuvant chemotherapy precludes direct comparisons of drug responses between patients’ tumors and PDTOs. Therefore, drug response rates observed in PDTOs could not be directly matched to responses of the matched, resected primary NSCLC tumor used for PDTO generation. However, PDTO responses matched recurrence patterns during the surveillance period in three patients with NSCLC who received adjuvant platinum-based doublet chemotherapy. Further, we observed a drug response rate of 50% to carboplatin/paclitaxel treatments in PDTOs—similar to the ∼50% response rate observed in patients with NSCLC receiving this regimen (1, 31). While TCGA analysis identified AKR1B10 overexpression at a rate in 59.0% of patients with NSCLC, TCGA does not provide chemotherapy response data. However, AKR1B10 expression correlated with chemoresistance in PDTO-matched tumor tissues and in a small, but well characterized, internal cohort of tissues of patients with NSCLC with known chemoresistance. Future studies need to validate whether AKR1B10 overexpression and drug resistance correlate in larger cohorts of patients with NSCLC.

In summary, NSCLC PDTOs preserve molecular and cellular characteristics of matched NSCLC tumor tissues and have the potential to serve as efficient preclinical platforms to test drugs to overcome drug resistance. In the present study, following confirmation of AKR1B10 overexpression, epalrestat was repurposed to overcome chemoresistance in NSCLC PDTOs. While there are detailed toxicity and biodistribution data for epalrestat available, the present study provides promising in vivo data on intratumoral bioavailability in patient-derived NSCLC tissues. The study provides a rationale for targeting AKR1B10 by repurposing epalrestat in a clinical trial to avoid unnecessary drug development time and pharmacological analyses to prevent and/or treat recurrences and reduce NSCLC mortality from drug resistance.

M.A. Ciorba reports grants from Incyte, Pfizer, and Janssen as well as personal fees from AbbVie and from Geneoscopy outside the submitted work. J.T. Kaifi reports grants from Department of Veterans Affairs during the conduct of the study as well as other support from Extract Biologics, LLC, outside the submitted work; in addition, J.T. Kaifi has a patent for US 11,890,616B2 issued and licensed to Kaifi/Kwon. No disclosures were reported by the other authors.

K.N. Suvilesh: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. Y. Manjunath: Conceptualization, resources, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. Y.I. Nussbaum: Data curation, software, formal analysis, visualization, writing–review and editing. M. Gadelkarim: Methodology. M. Raju: Validation. A. Srivastava: Conceptualization, formal analysis, writing–review and editing. G. Li: Conceptualization, formal analysis, supervision, funding acquisition, writing–original draft, writing–review and editing. W.C. Warren: Formal analysis, supervision, writing–original draft, writing–review and editing. C.-R. Shyu: Formal analysis, supervision, writing–original draft, writing–review and editing. F. Gao: Formal analysis, writing–original draft, writing–review and editing. M.A. Ciorba: Conceptualization, formal analysis, supervision, writing–original draft, writing–review and editing. J.B. Mitchem: Conceptualization, supervision, funding acquisition, visualization, methodology, writing–original draft, writing–review and editing. S. Rachagani: supervision, writing–original draft, writing–review and editing. J.T. Kaifi: conceptualization, formal analysis, supervision, funding acquisition, investigation, visualization, methodology, writing–original draft, writing–review and editing.

We are exceedingly grateful to all patients for their voluntary participation. The authors thank Nathan Bivens (MU Genomics Technology Core) for technical support and Brian Mooney (Charles W. Gehrke MU Proteomics Center) for constant inputs in experiments involving mass spectrometry. We are grateful to David Pittman (MU Department of Pathology) for reviewing and scoring histopathological and immunohistochemical staining. We thank Leonard Maggi, Washington University in St. Louis, for providing NSCLC cell lines. The authors thank all the funding support: J.T. Kaifi received funding from the Department of Veterans Affairs (CX002498-01A2). This study was also supported by an Ellis Fischel Cancer Center Pilot Award (J.T. Kaifi, S. Rachagani). J.B. Mitchem received funding from the Department of Veterans Affairs (K2BX004346-01A1). M.A. Ciorba has support from the Washington University DDRCC (NIDDK P30 DK052574) and the Barnes-Jewish Hospital Foundation Siteman Investment Program Award (Grant 5897). The content is solely the responsibility of the authors and does not necessarily represent the official views of the Department of Veterans Affairs. The funding bodies had no role in study design, collection, analysis, interpretation of data, or writing the manuscript. This research project was also supported by Endowments from the University of Missouri [Paul K. and Dianne Shumaker (C.-R. Shyu) and Margaret Proctor Mulligan (J.T. Kaifi) Professor Endowments].