Urine-derived bladder cancer organoids (urinoids) as a tool for cancer longitudinal response monitoring and therapy adaptation

Background Bladder cancer is one of the most common cancer types worldwide. Generally, research relies on invasive sampling strategies. Methods Here, we generate bladder cancer organoids directly from urine (urinoids). In this project, we establish 12 urinoid lines from 22 patients with non-muscle and muscle-invasive bladder tumours, with an efficiency of 55%. Results The histopathological features of the urinoids accurately resemble those of the original bladder tumours. Genetically, there is a high concordance of single nucleotide polymorphisms (92.56%) and insertions & deletions (91.54%) between urinoids and original tumours from patient 4. Furthermore, these urinoids show sensitivity to bladder cancer drugs, similar to their tissue-derived organoid counterparts. Genetic analysis of longitudinally generated tumoroids and urinoids from one patient receiving systemic immunotherapy, identify alterations that may guide the choice for second-line therapy. Successful treatment adaptation was subsequently demonstrated in the urinoid setting. Conclusion Therefore, urinoids can advance precision medicine in bladder cancer as a non-invasive platform for tumour pathogenesis, longitudinal drug-response monitoring, and therapy adaptation.


Introduction
Bladder cancer ranks among the top ve and ten most common cancers worldwide in men and women, respectively 1 .Over 573.000 patients were diagnosed with bladder cancer worldwide in 2022 (IARC).
Urothelial carcinoma (UC) is the predominant histopathological subtype of bladder cancer (BC) and on initial staging presents as non-muscle invasive bladder cancer (NMIBC; 73% of total) or muscle-invasive bladder cancer (MIBC) [2][3][4] .In general, NMIBC has a high recurrence rate (up to 84%), but rarely metastasizes, whereas MIBC is an aggressive disease with a high risk for relapse and metastases (up to 50%) 3,4 .Current neoadjuvant cisplatin-based combination chemotherapy regimens result in a complete pathological response in around 25% of the cases 5 .Thus, evaluation of tumor biology and assessment of chemosensitivity of the individual bladder tumor is needed to guide personalized bladder cancer treatment 6-11 .Large-scale genetic analyses of bladder cancer have identi ed drivers such as TP53, ARID1A, PIK3CA, FGFR3, STAG2 and ERBB2 and have yielded several molecular classi cation methods 12 , with consensus markers such as TP63 (Transitional/Intermediate cells), KRT5 (Basal class), KRT20 (Luminal class) and UKP3A (Urothelial differentiation) 11,[13][14][15][16] .However, it remains challenging to apply such information in clinical practice to guide treatment decision-making 59,17 .In addition, drug treatment itself may induce changes in tumor characteristics, such as genetic instability, and cause enrichment of speci c molecular subclones, which may result in altered drug-sensitivity and acquired drug-resistance 17- 20 .Platforms that allow drug-response monitoring longitudinally may therefore have great value in developing personalized and adaptive treatment strategies.
Recent developments in new technologies allow to grow living three-dimensional (3D) tumor structures on a patient speci c base (organoids).These in vitro multicellular 3D structures resemble key features of their original tumor tissue, are self-organizing and self renewing.Thereby, organoids allow in vitro patient speci c tumor analysis and screening [21][22][23][24] .Organoids derived from bladder tumor tissue (surgical biopsy) re ect the heterogeneity and molecular subclassi cation of the corresponding original tumor samples and can be used as a platform for testing drug-sensitivity 16 .Urine offers a potential alternative source for generating organoids from the urinary tract.Previously, benign kidney-and urothelium-derived organoids have been established from the urine of healthy individuals 16 .Importantly, urine from bladder cancer patients generally contains exfoliated viable tumor cells, providing a potential source for noninvasive tumor sampling 4 .
In this proof-of-concept study, we investigate whether urine-derived tumor cells can be used for the generation of organoids ('urinoids') from bladder cancer patients.In addition, we test the hypothesis that urinoid generation provides an effective non-invasive strategy for tumor subtyping and longitudinal drugresponse monitoring.We show that a single urine sample from a bladder cancer patient can be used as non-invasive sampling method for the generation of bladder cancer organoids with 55% e ciency.We show that these urinoids recapitulate histopathological features and genetic mutational status of the original tumor tissue.We show that urinoid clonality is similar to matched tumor tissue-based organoids ('tumoroids').Furthermore, we use both tumoroids and urinoids to identify the increase of patient speci c structural variations after immunotherapy.We nd a signi cant increase of mutational burden in among others microtubule-based processes and de novo chromoplexy events in both the urinoids and tumoroids after immunotherapy.Finally, we nd that both urinoids and tumoroids after immunotherapy show a signi cant increase of sensitivity to the microtubule targeting agents vinblastine and vincristine, which is not seen in pre-immunotherapy tumoroids.Overall, we conclude that urinoids are a non-invasive tool for following tumor pathogenesis, longitudinal drug-response monitoring, and therapy adaptation in bladder cancer.

Urine as a source for non-invasive generation of bladder cancer organoids
Bladder cancer patients from the University Medical Center Utrecht were invited to participate in this prospective proof-of-concept project by means of informed consent by the treating urologist or the nurse practitioner.Tumor tissue biopsies and urine samples were obtained from consenting bladder cancer patients, who underwent transurethral bladder-tumor resection (TUR-BT) or a radical cystectomy.The urine samples were collected using a transurethral catheter (Fig. 1).Urine-derived cells were subsequently cultured in bladder cancer organoid medium to generate organoids (urinoids) 16 .In parallel, bladder cancer organoids were generated from resected tumor tissue (tumoroids), which were also fresh-frozen for histopathological and genetic analyses.In total 22 bladder cancer patients were included in this study, resulting in a living biobank of 12 urinoid lines (culturing e ciency 55%; Fig. 1b; Table 1; Supplementary Table 1).Urinoids were generated from patients with various disease-stages including NMIBC (Ta (8.3%); CIS (16.7%);T1 (50%)) and MIBC (T2 (16.7%);T3 (8.3%)) (Fig. 1b; Table 1; Supplementary Table 1).
Urinoid cultures capture the histopathological features of the original tumor tissue All urinoids, paired tumoroids and the original non-cultured tumor tissues were analyzed by hematoxylineosin (H&E) staining and by immunohistochemistry (IHC) to detect expression of bladder cancer subtypespeci c and proliferation markers (Fig. 2; Supplementary Fig. 1a-c).H&E staining revealed that the histopathological morphology of the original tumor tissues was recapitulated in both the respective tumoroid and urinoid cultures (Fig. 2; Supplementary Fig. 1c).IHC analysis of Ki67 showed that the percentage of positive (proliferative) tumor cells was highly similar among the original tumor tissues and paired tumoroids and urinoids (Fig. 2).Furthermore, expression of p53, p63, CK5 and CK20 was highly concordant among the original tumor tissues and paired tumoroids and urinoids (Fig. 2; Supplementary Table 2).All original tissues, tumoroids and urinoids were negative for the nuclear kidney marker PAX8 (Supplementary Fig. 2).All tumor tissues and organoid lines were negative for the urothelial differentiation marker uroplakin 3a (UKP3A)(Supplementary Fig. 2).Urinoids were passaged for up to one year (~ 15 passages) and successfully biobanked via cryopreservation as a resource for future studies.Successful urinoid generation was not related to tumor stage (Supplementary Table 1).In several patients (n = 3), during surgery, malignancy was suspected, but upon further histopathological investigation revealed only benign tissue (pT0N0).Attempts for urinoid and tumoroid establishment from these patients all failed, highlighting the selectivity of the organoid growth medium for bladder cancer cells.

Mutational status of the urinoids resemble original tumor tissue of patient 4
Genetic resemblance of urinoids to the original tumor tissue was studied using bulk whole genome sequencing (WGS).For this in-depth genetic analysis, samples from patient 4 were analyzed.This patient was selected, as establishment of both urinoids and tumoroids were successful and this patient underwent a novel immunotherapy treatment, as part of the NABUCCO-trial, which consisted of 3 cycles of ipilimumab (1mg/kg) and nivolumab (3mg/kg) (Fig. 3a) 25 .At the start of this trial, patient 4 was diagnosed with a high-grade cT4aN1 invasive bladder cancer and suspected of invasion of the prostate.At the transurethral bladder tumor resection (TUR-BT), a tumor tissue sample was collected and a preimmunotherapy tumoroid line (BTOR4.1)was generated.The patient had a radiological reduction in tumor size (53 to 21mm diameter) in response to the pre-operative immunotherapy treatment, with no signs of invasion of the prostate.Subsequently, a radical cystoprostatectomy was performed.At the cystoprostatectomy, samples were harvested from the tumor for both the unaltered post-immunotherapy baseline tumor sample, original tumor sample (Original 4.2), and for the establishment of both postimmunotherapy urinoid (UBTOR4.2) and tumoroid (BTOR4.2) lines.
All urinoid and tumoroid samples of patient 4 were analyzed by WGS, to evaluate genetic resemblance to the original bladder tumor.Similarity of all post-therapy samples was based on single nucleotide polymorphisms (SNPs).Highly similar counts of SNPs were observed from the urinoid UBTOR4.2 and tumoroid BTOR4.2 (Supplementary Fig. 3a) when compared to the original tumor sample 4.2.SNP similarity was further con rmed by correlating the individual SNPs to those in the original tumor tissue.
SNPs were found to be either unique to the organoid lines or shared with the original tumor tissue (Fig. 3b).A high degree of shared SNPs were found for UBTOR4.2(92.56%) and BTOR4.2 (92.15%), compared to the original tumor sample 4.2.Next, the absolute number (Supplementary Fig. 3b) and the degree of shared (Fig. 3c) insertions and deletions (InDels) were analyzed.Similar absolute counts and a high degree of shared InDels were observed between UBTOR4.2 (91.54%),BTOR4.2 (90.69%) with original tumor sample 4.2.The predicted class and impact of mutations was analyzed with the snpEff tool.All post-immunotherapy organoids were highly similar to the original tumor sample 4.2 in counts (Supplementary Fig. 3c) and relative degree of shared missense and silent mutations (Fig. 3d).Interestingly, higher percentile variance was found in the nonsense mutations (Supplementary Fig. 3e).These were disregarded as signi cant differences because of the low absolute counts of nonsense mutations.The predicted impact variance of all SNPs and InDels on affected genes was found mainly in modifying gene impact mutations (Supplementary Fig. 3d).The degree of shared modi er (Fig. 3e), low, moderate and high impact mutations (Supplementary Fig. 3f) was high for all post-therapy organoids compared to the original tumor sample 4.2.Interestingly, high similarity of SNPs, InDels, class and impact of mutations were found in the pre-immunotherapy tumoroid line BTOR4.1 versus the postimmunotherapy original tumor sample 4.2 (Supplementary Fig. 3g-j, respectively).This suggests that any difference between pre-and post-therapy should be found within the smaller unique percentages or outside of the SNP and Indel mutations.Overall, urinoids from patient 4 were found to be highly similar on the genetic level compared to the original tumor tissue.

Clonality in urinoid cultures similar to tissue-based tumoroids
To determine if urine-based organoid establishment selects for genetic clonal lines, single cell karyotype sequencing 26 was performed for BTOR4.1,BTOR4.2 and UBTOR4.2(Supplementary Fig. 4a-c, respectively).Though several single cells show copy number variations (CNVs), the consensus of the cells for each line is diploid (Supplementary Fig. 4a-c, bottom plots).Only larger CNVs were found in the Y-chromosome, which is often disregarded 26 .Diploid bladder tumors have previously been reported in several bladder cancer cell lines and patients 27,28 .Clustering all cells using a principal component analysis (PCA), most cells cluster together for all organoid lines (Supplementary Fig. 4d&e).All organoid lines derived from patient 4 showed few CNVs and no genetic subclones within the organoid populations.Overall, no treatment induced large copy number variations were detected.Furthermore, no subclones were introduced by urine-based organoid establishment.

Increase of structural variations in patient 4 after immunotherapy
Using the GRIDSS-PURPLE-LINX whole genome analysis pipeline, larger mutations and overall genetic makeup were analyzed in all organoid lines derived from patient 4. Based on the CNV and B-allele frequency (BAF) (Supplementary Fig. 5a-c), the majority of the genome was found to be diploid copy numbers with only several smaller CNVs.This con rms previous ndings of single-cell karyotype sequencing.Interestingly, differences between pre-and post-immunotherapy were found when studying the structural variations (SVs).The absolute number (Fig. 4a) of structural variations (SVs) in the organoid lines of patient 4 show an increase of SV counts in both the post-immunotherapy organoid lines UBTOR4.2 and BTOR4.2.This increase in SVs might be caused due to a recent catastrophic genetic rearrangement [29][30][31] .Studying the fusion genes found in BTOR4.1,BTOR4.2 and UBTOR4.2(Supplementary Table 3), gene fusion FGFR3-TACC3 was found in all lines.FGFR3-TACC3 fusion has been previously linked to a poor prognosis and bladder cancer tumor progression 32 .Interestingly, a small decrease of fusion proteins was detected in the post-immunotherapy organoids BTOR4.2 (n = 8) and UBTOR4.2(n = 8), compared with the pre-immunotherapy BTOR4.1 (n = 11) (Supplementary Table 3).
Increase of SNP mutational burden in microtubule-based processes in patient 4 after immunotherapy It was hypothesized whether further genetic differences between the pre-and post-immunotherapy organoid lines can be found.As SNP counts and overlap are similar, SNPs only found in both postimmunotherapy lines were analyzed using an unbiased SNP pathway enrichment analysis (Fig. 4b; Supplementary Fig. 5d).Taking the top 50 quality SNP enrichments of all Gene Ontology (GO) segments, signi cant enrichment of SNPs (p < 0.05) was found in (among others) pathways relying on the microtubule network and other key mitotic processes.Previously, similar defects in microtubule-based and mitotic processes were shown to cause defects in structural stability and induce large chromosomal copy number variations 33,34 .

Chromoplexy events in patient 4 after immunotherapy
As genetic differences between pre-and post-immunotherapy were found in SVs, the unique SVs in both UBTOR4.2 and BTOR4.2 were studied further.Several complex link SV rearrangements were observed in both tumoroid BTOR4.2 and urinoid UBTOR4.2 (Fig. 4c&d; Supplementary Fig. 5e-g) 35 .This complex multi chromosomal (> 2) linked SVs have previously been documented as variant of chromoanagenesis called 'chromoplexy' in esophagus and prostate cancer 36,37 .No multi chromosomal complex linked SVs were found in BTOR4.1.Overall, these results indicate that a chromoplexy event emerged during the ipilimumab and nivolumab immunotherapy of patient 4.

Increased sensitivity to microtubule destabilizing agents after immunotherapy and chromoplexy events
The drug-responses of the longitudinally generated urinoids and tumoroids from patient 4 were evaluated, by exposure to a variety of commonly used DNA-intercalating drugs (cisplatin, gemcitabin, carboplatin, doxorubicin, epirubicin), microtubule destabilizing agents (vinblastine, vincristine) and a novel broblast growth factor kinase inhibiting agent (erda tinib) (Fig. 5; Supplementary Fig. 6a-h).Erda tinib was tested given the FGFR3-TACC3 fusion and to evaluate urinoid application for novel drug e cacy evaluation.In particular, the microtubule destabilizing agents were tested based on the discovered increased SNP load in microtubule-based processes in all post-immunotherapy organoids.The relative sensitivity per drug concentration was summarized in total areas under the curves (AUCs), measured using the relative ATP levels (Fig. 5).Overall limited responses were found for cisplatin, carboplatin and erda tinib in all organoids derived from patient 4. Moderate responses were found to doxorubicin and epirubicin, and overall high sensitivity to gemcitabine.Interestingly, a signi cant difference in sensitivity to microtubule destabilizing agents vincristine (p < 0.0001) and vinblastine (p < 0.0001) was seen between pre-treatment BTOR4.1 and both post-treatment lines BTOR4.2 and UBTOR4.2.Overall, these results underline the potential of urinoids as non-invasive method to evaluate drug-response in bladder cancer patients during treatment.

Discussion
This proof-of-concept study demonstrates that non-invasive patient-derived urine-based bladder cancer organoids (urinoids) can be successfully generated from urine samples.These urinoids recapitulate the histopathological features of the original tumor tissue, such as degree of proliferation and molecular subtype differentiations, found in both patients with non-muscle invasive and muscle invasive bladder cancer.In-depth genetic analysis shows that mutational pro ling of the urinoids closely resembles between the original tissue and the correlating tissue-based tumoroids (patient 4).Finally, the urinoids respond similarly to their respective tissue-based tumoroids in drug e cacy evaluation of standard and novel bladder cancer therapies.
Of special interest is the difference in the SNP load that was found post-compared to pre-immunotherapy.
The post-immunotherapy urinoid and tumoroid lines contain a signi cantly enriched SNP load in the mitotic machinery and microtubule network, compared to pre-immunotherapy samples.This increased SNP load could have caused an increase in structural variations and a catastrophic chromoplexy event, which were detected in both of the post-immunotherapy urinoid and tumoroid lines -38 .Previously observed chromoplexy has not yet been directly linked to a speci c moment or treatment in a singular patient [36][37][38][39] .As a signi cant increase of SNP load was found in microtubule-based processes, drug sensitivity to two microtubule destabilizing agents was evaluated.Compared to the pre-immunotherapy tumoroid line, both post-immunotherapy chromoplexic urinoids and tumoroids showed a signi cant increase in sensitivity to microtubule targeting agents (i.e.vincristine and vinblastine These drugresponse results support the de novo increase of the SNP load in the microtubule-based processes for both post-immunotherapy urinoid and tumoroid lines of patient 4. We hypothesize that these mutations lead to both de novo increased SVs and targeted treatment sensitivity that occured in the short timeframe of patient 4's 84 day treatment window.Whether these de novo increase of SNPs is immunotherapy induced or due to a treatment selection on already present mutations in a small clonal population could not be determined.Defective mitotic machinery is often the cause for chromosomal rearrangements with chromosomal instability [40][41][42][43][44] .Further specialized studies are needed to understand the underlying mechanisms of these chromoplexy events in the context of this immunotherapy.If this event is found to be common, the combination of immunotherapy with microtubule destabilization drugs (e.g.vincristine and vinblastine) could potentially have a synergistic therapeutic effect.Overall, this proof-of-concept study shows that urinoids can be generated for in vitro monitoring and drug-response prediction in bladder cancer.Both urine-derived and tissue-derived post immunotherapy organoids showed speci c treatment-induced de novo drug sensitivity for microtubule destabilization drugs.This similarity in sensitivity highlights the feasibility and value of sequential (follow-up) monitoring of a patient's speci c bladder tumor characteristics using the non-invasive urinoid approach.The currently reported urinoid culturing success rate of 55% has already led to the following practical improvements: (I) collecting more than one single urine sample, (II) urine collection in culture medium instead of PBS and (III) addition of antibiotics to collection volume.
Urinoids provide a unique opportunity to culture sequential follow-up samples from bladder cancer patients during their treatment.These insights will subsequently steer the development of current and novel combination therapies targeting drug-resistance pathways, or acquired vulnerabilities.Ultimately, urinoids are a non-invasive tool for following tumor pathogenesis, longitudinal drug-response monitoring, and therapy adaptation in bladder cancer.

Methods
Approval of studies involving human tissue and patient-inform consent.Human bladder tissue was obtained from the University Medical Center Utrecht (UMCU).Ethical approval was granted by the Biobank Research Ethics Committee (TCBio) of the UMCU.Written informed consent was obtained from all patients involved in this project.Bladder tissue was obtained through transurethral bladder tumor resections (TUR-BT) or cystectomy procedures.Urine samples were obtained through transurethral catheterization at the time of surgery.All samples were examined by a dedicated uro-pathologist.
Human bladder organoids establishment and culture.Human bladder tissue was examined and selected on malignancy by dedicated uro-pathologists.In both the TUR-BT and the cystectomy cases, we obtained a sample of tumor tissue from the patient.The tissue was cut into smaller pieces (1 mm to − 2 mm) with a surgical blade, of which half was frozen as original tumor sample.The remaining half was digested with Liberase (Sigma-Aldrich, 5401135001) in Advanced DMEM/F-12 (ThermoFisher, 12634028) with ROCK inhibitor (Y-27632, 10 µM) for 60 minutes at 37°C.This resulted in the generation of tissue-based organoids called tumoroids.Urine samples were collected at the start of the TUR-BT or cystectomy operation via catheterization into 25 mL DPBS (Corning Life Sciences, 21-031-CVR), with urine volume ranging from 5-75 mL samples.Urine tumor cells were collected by centrifugation and washed with DPBS for a minimum of 5 times.After centrifugation, the cell pellet was resuspended in ∼200 µL of RGF BME (R&D Systems Europe, 3533-010-02) and plated into one to two individual wells of a prewarmed 6well plate.This resulted in the generation of urine-based organoids called urinoids.In both tissue-and urine-based establishments, when the BME was solidi ed, human bladder organoid media was added.
Immunohistochemistry. Organoids and tissue were xed in 4% paraformaldehyde for 1 hr. to 6 hrs., dehydrated, and para n-embedded according to standard histology procedures.Sections were stained with Hematoxylin & Eosin (H&E) or the following antibodies: Keratin 5 (Abcam, ab52635), Ki67 (Dako, M7240), TP53 (Santa Cruz, sc-126), TP63 (Abcam, ab735) and Uroplakin III (SFI-1, Abcam, ab78196) according to the manufacturer's protocols.For Keratin 20 (DakoCytomation, M7019), EDTA antigen retrieval was used in combination with a Ultravision Protein Block (Fisher Scienti c, TA-125-PBQ) for 30 minutes at room temperature.The following secondary antibodies conjugated with HRP were used: bright vision poly-hrp-anti rabbit igg (VWR international, VWRKDPVR110HRP) and bright vision poly-hrp-anti mouse igg (VWR international, VWRKDPVM110HRP) for 1hr.at room temperature, and Goat IgG HRPconjugated Antibody (R&D Systems Europe, HAF017) for 1 hr.at 37°C.Staining was performed using 3, 3'-diaminobenzidine (DAB) for 10 minutes exactly at room temperature.Images were acquired by high resolution scanning of the slides and analyzed using the NDP.view (v2.7.39) software (Hamamatsu Photonics K.K.).These sections were evaluated in a blinded test by dedicated uro-pathologists.Regrettably, due to a clerical error, the original tissue of UBTOR5 was not available for immunohistochemical analysis and thus excluded for the comparison.
DNA isolation and library preparation.DNA was isolated using the QIAamp® DNA mini kit according to the manufacturer's instructions.Per sample, 500-1000 ng of DNA was used for DNA library preparation.Library preparation was performed following the Truseq DNA nano protocol.
Whole genome sequencing.Whole genome sequencing was performed by the Utrecht Sequencing Facility using the Illumina NovaSeq 6000 set-up and analyzed using the nf-core/sarek pipeline (v2.7.1) with reference genome assembly GRCh38 and with the GRIDSS-PURPLE-LINX pipeline (v1.3.2).Paired-end whole genome sequencing was performed with an average coverage of 30x.For both pipelines, normaltumor mode could not be used, but all modules were run with tumor-only, single-sample and/or multiplesample variant calling work ows when possible.
Single Cell karyotype sequencing.Single cell karyotype sequencing was provided by Single-Cell Core of the Oncode Institute, Utrecht, the Netherlands. 26Nuclei in 384-well plates are digested with NlaIII, after which the genomic fragments (following end processing) are ligated to barcoded-adapters containing a unique molecular identi er (UMI), cell-speci c barcode, and T7 promoter allowing linear ampli cation by in vitro transcription (IVT).Libraries were sequenced on an Illumina Nextseq500 with 2×75-bp paired-end sequencing.The data preprocessing has been performed using the SingleCellMultiOmics package      See above image for gure legend.

Figures
Figures

Figure 1 See
Figure 1

Figure 2 See
Figure 2

Figure 3 See
Figure 3

Table 1 .
Overview of all urinoids generated in this project giving an overview of all patients

Table overview
of all urinoids generated in this project giving an overview of all patients, urinoids generated, TNM classification, pathological stage of the tumor, gender and passage of the organoid line.UCC = Urothelial carcinoma, SCC = Squamous cell carcinoma, PI = Peritoneal invasion.All lines were stocked at lower passage numbers than the indicated passage number.Lines were taken out of long term culture at indicated passage number.