Nano Pom-Poms Prepared Exosomes for Highly Specic Cancer Biomarker Detection

Extracellular vesicles (EVs), particularly nano-sized small EV exosomes, are emerging biomarker sources. However, due to heterogeneous populations secreted from diverse cell types, mapping exosome multi-omic molecular information specically to their pathogenesis origin for cancer biomarker identication is still extraordinarily challenging. Herein, we introduced a novel 3D-structured nanographene immunomagnetic particles (NanoPoms) with unique ower pom-poms morphology and photo-click chemistry for specic marker-dened capture and release of intact exosome. This specic exosome isolation approach leads to the expanded identication of targetable cancer biomarkers with enhanced specicity and sensitivity, as demonstrated by multi-omic exosome analysis of bladder cancer patient tissue uids using the next generation sequencing of somatic DNA mutations, miRNAs, and the global proteome. The NanoPoms prepared exosomes also exhibit distinctive in vivo biodistribution patterns, highlighting the highly viable and integral quality. The developed method is simple and straightforward, which is applicable to nearly all types of biological uids and amenable for enrichment, scale up, and high-throughput exosome isolation. biomarker discovery and precisely dening the onset and progression of cancer diseases 30 . In this paper, we demonstrated such capability for analyzing exosomes derived from bladder cancer patient tissue uids including urine, plasma, and tumor tissue, by the next generation sequencing (NGS) of somatic DNA mutations, miRNAs, and the global proteome for achieving non-invasive, ultra-sensitive diagnosis of bladder cancer. The results showed improved specicity and sensitivity for detecting urological tumor biomarkers from NanoPoms isolated exosomes compared to other ultracentrifugation or bead isolation approaches. We also identied a few new miRNAs and proteome cancer biomarkers highly enriched in urinary exosomes but was not reported yet, which could expand the landscape for discovering new EV cancer biomarkers for improving bladder cancer diagnosis. ROC analysis, and clustering analysis. The one-way ANOVA and t test were used. Differences are considered statistically signicant at P < 0.05. *P < 0.05; **P < 0.01; ***P < 0.001. The Venn diagram was analyzed using open software from the Bioinformatics & Evolutionary Genomics. The bioinformatic analysis of small RNAs was detailed in supplementary material.


Introduction
Despite the tremendous efforts made in developing cancer biomarkers and liquid biopsy for past decades, only a few (less than 25) cancer biomarkers have been approved by FDA for clinical practice 1,2 .
Extracellular vesicles (EVs) have been emerging biomarker sources for expanding the landscape of cancer biomarker discovery in promoting cancer diagnosis [3][4][5] , immunotherapy 6,7 , drug target and delivery 8 . Signi cant attention has been focused on the exosome type small EVs (sEVs) and their molecular components (e.g., proteins, DNAs, mRNA and miRNA), which has been found in association with a variety of physiological functions and pathological disease states 9 . Exosome secretion is exacerbated from tumor cells and enriched with a group of tumor markers, as evidenced by increased presence in plasma and ascites from patients in variable cancers 10 . However, currently there is no standardized puri cation method for processing body uids which often contain diverse EV types, and obtaining homogeneous exosome populations that are speci c to their cellular origin and molecular components is not attainable 11,12 . As evidenced, EVs are living cell-secreted membrane vesicles in multiple subpopulations, including membrane shedding microvesicles (100 nm-1000 nm), endosomal multivesicular body released exosomes (30 nm-150 nm), and apoptotic cellular fragment vesicles (≥ 1000 nm) [13][14][15][16] . Due to such large heterogeneity and signi cant size overlap between vesicle populations, the consensus has not yet emerged on precisely de ning EV subtypes, such as endosome derived exosomes 17 which is highly relevant to the disease pathogenesis. The generic term of EVs is recommended by complying with 2018 guidelines from the International Society for Extracellular Vesicles (ISEV) proposed Minimal Information for Studies of Extracellular Vesicles ("MISEV") 17 . Current puri cation methods that recover the highest amount of extracellular materials, no matter with the vesicle or non-vesicular molecules, are mainly the precipitation polymer kits and lengthy ultracentrifugationbased (UC) approach 18,19 . Such isolation approach is not scalable and unable to differentiate the exosome populations from different cellular origin or other EV subtypes (e.g., microvesicles and apoptotic bodies), neither free proteins 20 or viruses, in turn, posing a signi cant concern for studying cancer biomarkers from tumor cells derived exosomes. The bulk measurement of a mixture of vesicle populations could potentially mask the essential biosignatures, which severely impairs the investigations of associated pathological mechanism [21][22][23][24] . As the perfectly enriched biomarker sources, using exosomes for mapping multi-omic molecular information speci cally to their pathogenesis in cancer biomarker identi cation is still extraordinarily challenging.
Herein, we introduced a novel approach using 3D-structured nanographene immunomagnetic particles (NanoPoms), which possesses unique ower pom-poms morphology and photo-click chemistry for speci c marker-de ned capture and release of intact exosomes from nearly all types of biological uids, including human blood, urine, cow's milk, and cell culture medium, etc (Figs S1 and S2). Compared to current existing immunomagnetic beads based EV isolation either in small quality or bound to solid surface/particles [25][26][27] , NanoPoms enable on demand capture and release of intact exosomes, which leads to the expanded identi cation of targetable cancer biomarkers with enhanced speci city and sensitivity. Zhu etc. also reported the covalent chemistry mediated EV capture/release using click chip for EV puri cation and diagnosis of hepatocellular carcinoma (HCC) 28 and Ewing sarcoma 29 . The transcyclooctene (TCO) grafted antibody bound EVs with click chemistry motifs are presented after EV release which may introduce alteration of EV surface properties for in vivo applications. We tested in vivo biodistribution of our NanoPoms prepared exosomes after release and results support the retaining of biological surface properties with distinctive biodistribution patterns speci c to exosome cellular origin, which implies the great potential for therapeutic development.
The group of enriched biomarkers carried by exosomes, including DNAs, RNAs and proteins, could offer the unmatched possibility to integrate multi-omic data analysis for expending the landscape of cancer biomarker discovery and precisely de ning the onset and progression of cancer diseases 30 . In this paper, we demonstrated such capability for analyzing exosomes derived from bladder cancer patient tissue uids including urine, plasma, and tumor tissue, by the next generation sequencing (NGS) of somatic DNA mutations, miRNAs, and the global proteome for achieving non-invasive, ultra-sensitive diagnosis of bladder cancer. The results showed improved speci city and sensitivity for detecting urological tumor biomarkers from NanoPoms isolated exosomes compared to other ultracentrifugation or bead isolation approaches. We also identi ed a few new miRNAs and proteome cancer biomarkers highly enriched in urinary exosomes but was not reported yet, which could expand the landscape for discovering new EV cancer biomarkers for improving bladder cancer diagnosis.
Beads were washed with PBST then resuspend in 1 ml PBST and 0.09% NaN3 solution for storage at 4 ℃. In this study we used the pan capture with a mixture of CD9, CD63, and CD81 antibodies for beadconjugation. For in vivo biodistribution study, we used CD9 antibody conjugated NanoPoms to prepare HTB9 and HEK cells derived exosomes. After bead fabrication and conjugation, XPS analysis was used (PHI 5000 VERSA PROBE II) with an Al anode of the x-ray source (46.95 eV) and 100 µ X-ray beam size for operating at 23.2 W. The power of the source was reduced to minimize X-ray damage for analyzing exosomes on bead surface.
The EV isolation from patients' plasma, urine, or cow milk and conditioned cell culture media were performed by incubation of 100 μ L antibody-beads complex with 1 mL of samples at 4 ℃ overnight. After washing, the photorelease was performed using Analytikjena UVP 2UV Transilluminator Plus at 365 nm wavelength at 4 ℃ for 15 min (~6 mW/cm2). The UC isolation of EVs followed the well-documented protocols published previous 68 . Brie y, to remove any possible apoptotic bodies and large cell debris, the supernatants were centrifuged at 10,000g for 30 mins, then transferred to ultracentrifuge tube (Thermo Scienti c, USA) for ultracentrifugation at 100,000g for 70 min (Sorvall™ MTX150 Micro-Ultracentrifuge, USA), with second ultracentrifugation (100,000g for 70 min) for nally collecting EV pellets. The size characterization of EVs was performed using the nanoparticle tracking analysis (NTA) Nano-Sight LM10 (Malvern Panalytical). Post-acquisition parameters were adjusted to a screen gain of 10.0 and a detection threshold to 5. Standard 100 nm nanoparticles were used for calibration. Appropriate sample dilution in 1× PBS was evaluated before every measurement with ve repeats for each measurement.
Exosome DNA extraction and NGS sequencing. Frozen urine samples were thawed overnight at 4 °C and pre-centrifuged at 4 °C 10,000g for 30 min to remove cell debris. By using NanoPoms isolation, the extracted exosomes were treated with DNase I before DNA extraction. The QIAamp DNA Mini Kit (Qiagen, 51304) was utilized to extract DNA from all EV samples. The addition of 1 μL of an aqueous solution containing 10 μg of carrier DNA (poly dA) to 200 μL Buffer AL was used to ensure binding conditions are optimal for low copy number DNA according to the manufacturer's protocols. DNA was Eluted in 20 μL Buffer AE. DNA concentrations were measured using a Nanodrop platform at an absorbance at 260 and 280 nm subtracted by the background value of carrier ploy dA only.
The library preparation by targeted enrichment using Qiagen GeneRead QIAact AIT DNA UMI and GeneRead clonal Amp Q Kits, was subjected to next-generation sequencing (NGS) to generate FASTQ les The extracted DNAs were ampli ed by PCR to detect the EGFR (P00533:p.Thr790Met) mutation. The sequences of primers for PCR were as follows: Primer F, 5'-ATGCGTCTTCACCTGGAA-3'; primer R, 5'-ATCCTGGCTCCTTATCTCC-3'. Primers were designed by Primer3Plus online. The PCR assay was performed with Promega GoTaq Flexi DNA Polymerase kit in a 50-μL mixture containing 10μL of 5× PCR buffer, 0.25μL GoTaq Flexi DNA Polymerase, 10 μM of each primer (IDT, USA) and 20 μL of DNA in an ABI PCR instrument (Applied Biosystems). The PCR conditions were as follows: Initial denaturation at 95˚C for 2 min, followed by 35 cycles at 95˚C for 15 sec, 54˚C for 30 sec and 72˚C for 40 sec, then a hold at 72˚C for 5 min and a nal permanent hold at 4˚C. The 319 bp DNA size of PCR products were clari ed by 1% agarose gel electrophoresis using 5 μL PCR products and remained DNA were puri ed by QIAquick PCR Puri cation Kit (Qiagen, 28104). The puri ed PCR products were sequenced by Sanger Sequencing approach (GeneWiz, USA) using the same primers above.
Exosome RNA extraction and NGS sequencing. The miRNeasy Mini Kit (Qiagen, 217004) was used to extract total RNA from all EV samples per manufacture's protocols. The amount of 700 μL QIAzol lysis reagent was adapted according to the manual. To achieve a higher RNA yield, the rst eluate of 30 μL was applied to the membrane a second time. Isolated RNAs were quanti ed by High Sensitivity RNA ScreenTape Assay using Agilent TapeStation 2200 (Agilent, 5067-5579, 5067-5580). Total RNA was stored at -80 °C until small RNA Library preparation. The QIAseq miRNA Library is prepared for Single Read 75bp sequencing, with UMI tag per manufacture's protocols. After small RNA sequencing using Illumina MiSeq system, the Qiagen speci c UMI analysis per the kit instruction was performed with details in supplementary information.
Droplet digital PCR. A pair of probes and a pair of primers were designed to detect EGFR and TERT mutation respectively. Due to the short size of the probe, in order to increase the hybridization properties and melting temperature, Locked Nucleic Acid (LNA) bases were introduced on the bases indicated with a "+". One probe was designed to recognize wildtype (5′-TET/T+CATC+A+C+GC/ZEN/A+GCTC/-3′ IABkFQ). The second probe was designed to recognize the EGFR (P00533:p.Thr790Met) mutation loci, (5′-6FAM/T+CATC+A+T+GC/ZEN/A+GC+TC/-3′ IABkFQ). Primers were designed to cover both side of detection loci. For TERT, a probe was designed to detect both C228T and C250T mutation as both mutations result in the same sequencing string 70 , with (TERT Mut:/56-FAM/CCC+C+T+T+CCGG/3IABkFQ/). A second probe was designed to recognize the C228 loci, also containing LNA bases, (TERT WT, /5HEX/ CCCC+C+T+CCGG/3IABkFQ/). Probes and primers were custom synthesized by Integrated DNA Technologies (IDT). Ampli cations were performed in a 20 μL reaction containing 1 × ddPCR Supermix for Probes (No dUTP), ( Bio-Rad, 1863024), 250 nM of probes and 900 nM of primers and 8 μL exosome DNA template. Droplets were generated using the QX200 AutoDG Droplet Digtal PCR System (Bio-Rad). Droplets were transferred to a 96-well plate for PCR ampli cation in the QX200 Droplet Reader. Ampli cations were performed using the following cycling conditions: 1 cycle of 95°C for 10 minutes, then 40 cycles of 94 °C for 30 seconds and 60 °C for 1 minute, followed by 1 cycle of 98°C for 10 minutes for enzyme deactivation. Keep all ramp rate at 2°C/sec. QuantaSoft analysis software (Bio-Rad) was used to acquire and analyze data.
Western blotting and Proteomic analysis. The 5 mL of each urine sample for two patients and one healthy control were used for exosome isolation and subsequent Western blot analysis. 40 mL of HTB-9 conditional cell culture media and 40 mg cell pellets were also used as controls in this study. Samples were lysed in 1× RIPA buffer supplemented with protease inhibitors for 15 min on ice. Only cell sample were ultrasonicated for 1 min. Protein concentration was quanti ed using Micro BCA Protein Assay Kit (Thermo Fisher, 23235). The absorbances were read at 562 nm on a Synergy H1 reader (BioTek). All sample concentration were adjusted to 0.1 μg/μL. Western blotting was performed under reducing conditions (RIPA buffer, β-mercaptoethanol and Halt Protease Inhibitor Cocktail, EDTA-Free) at 95 °C for 5 min. 20 μL of protein lysate, each, were loaded onto 4-20% Mini-PROTEAN TGX Precast Protein Gels (BioRad, 4561093). The separated proteins were transferred to a PVDF membrane (BioRad, 1620218).
After blocking the membrane in Intercept (PBS) Blocking Buffer (LI-COR, 927-70001) for one hour at room temperature, it was incubated over-night with the primary antibody at 4 °C, followed by another incubation with the secondary antibody for half hour at room temperature. SEM and TEM. exosome-bead particle complex was resuspended in 200 μL cold PBS solution. For electron microscope evaluation, exosome-particle complexes were washed with pure water followed by the xation in a 2% EMS-quality paraformaldehyde aqueous solution. 5 μL of exosome-particle mixtures were added to cleaned silicon chips and immobilized after drying EVs under a ventilation hood. Samples on silicon chips were mounted on a SEM stage by carbon paste. A coating of gold-palladium alloy was applied to improve SEM image background. SEM was performed under low beam energies (7 kV) on Hitachi SU8230 led emission scanning electron microscope. For TEM, ~5 μL of each exosome-particle complex was left to adhere onto formvar carbon coated copper Grid 200 mesh (Electron Microscopy Sciences) for 5 mins followed by 5 mins of negative staining with 2% aqueous uranyl acetate. Excess liquids were blotted by lter papers. Total grid preparation was performed at room temperature till totally air-dried under a ventilation hood for 25 mins. Images were acquired on the same day at 75 kV using Hitachi H-8100 transmission electron microscope.
DiR-labelled exosomes or free DiR dyes were segregated using Amicon Ultra-15 Centrifugal Filter method.
The 2.0 ×10 9 particles/ml of isolated exosomes measured via NTA were used for each mouse injection. The 6-to 8-week-old female BALB/cJ mice were used. The animal IACUC protocols have been approved by the University of Kansas Institutional Animal Care and Use Committee with protocol number 258-01 and operated in the KU Animal Care Unit. Freshly puri ed DiR-labelled exosomes were injected through the tail vein for intravenous (i.v.) injection. The In-Vivo Systems (Bruker, USA) with high-sensitive CCD camera was used for collecting uorescence, luminescence and X-ray images. Iso urane sedated live mice were taken uorescence and X-ray images prior to the animals were sacri ced, then main organs (brain, heart, lung, liver, kidney and spleen) were harvested for uorescence imaging in 3 mins (excitation 730 nm, emission 790 nm), X-ray imaging (120 mm FOV, 1 min) and luminescence imaging (90 fov, 0.

Results
NanoPoms enable speci c capture and on-demand release of intact exosomes.
In this work, we introduce a novel 3D-structured nanographene immunomagnetic particles (NanoPoms) with unique ower pom-poms morphology and photo-click surface chemistry for speci c marker-de ned capture and release of intact exosomes (Fig. 1). Conventionally, the non-covalently assembled nanographene suffers from the instability in buffer solutions over time 31 . Our method interfaces Fe3O4/SiO2 core-shell particles (~800 nm) with graphene nanosheets via carboxamide covalent bonds, which leads to substantially improved stability in the aqueous samples. The ower pom-poms morphology produces the unique 3D nano-scale cavities in between for a nity capture of only nano-sized vesicles such as exosomes ( Fig. 1 and Fig. s1). The dense nano-graphene polydopamine sheet layers provide much larger surface area [32][33][34][35] for immobilization of a nity capture entities (e.g., antibodies, aptamers, and a nity peptides) as shown in Fig. 1B right panel in contrast to conventional beads. Most importantly, the conjugated photo-click chemistry on bead surface allows the release of intact, captured exosomes on demand, which further ensures the speci city for harvesting marker-de ned exosome subpopulations. Fig. 1C immune gold nanoparticle staining TEM imaging demonstrated the capture of exosomes, which exhibit much narrower size distribution than ultracentrifugation (UC) prepared EVs (Fig. 1D), and uniform and reproducible particle size around 100-150 nm smaller than ExoEasy prepared EVs (Fig. 1E). We observed the dense and round-shaped exosomes (~100 nm) completely covering the surface of NanoPoms and subsequently restoring the pom-poms surface morphology after light release (Fig. 1F).
The XPS analysis, TEM, uorescence binding analysis and BCA Protein Assay were also performed to evaluate exosome capture performance and capacity in Figs S1, S2 and S3. This NanoPoms method is applicable to nearly all types of biological uids, including human blood, urine, cow's milk, and cell culture medium, etc. (Fig. S3). The operation protocol is simple and cost-effective, amenable for scaling up, sterilization settings, and GMP operations (see Table S1).
NGS analysis of somatic DNA mutations carried by urinary exosomes. Detecting DNA mutations carried by urinary tumor exosomes is emerging, yet challenging, due to the needs of highly pure sample preparation for detection. We analyzed the bladder cancer (BC) patient urine samples prepared by both NanoPoms, UC, and commercial bead approaches for isolating urinary exosomes, with control group from healthy individuals. The NGS GeneRead AIT panel was used to identify the most cancer relevant 1,411 variants. UC preparation was found insensitive to cancer relevant variant detection, as it requires much larger urine sample input (4 mL) with more than 100 ng exosome DNAs to give detectable variant signals ( Fig. 2A). We suspect that UC isolated exosome DNAs contain more genes which are not speci c to cancers. The PDGFRA variant (c.1432T>C, p.Ser478Pro) with 56.8% frequency was detected from a healthy individual in the control group using UC preparation, but not from NanoPoms preparation. In the BC disease group, NanoPoms prepared exosomes enabled much enhanced detection sensitivity and speci city to BC relevant mutations including KRAS, PIK3CA, and ERBB2, which only consumed 1 mL urine sample with about 10-50 ng exosome DNAs. However, commercial bead isolated exosomes using the same input of urine samples did not yield su cient DNAs for sensitive detection of cancer relevant variants ( Fig. 2A). In order to validate whether the gene mutations found in urinary exosomes are from the urological tumor, we evaluated the matched patient tumor tissue. The NGS GeneRead analysis of tumor tissue cells showed the consistent mutations of KRAS and ERBB2 as carried by urinary exosomes from the same BC patient. Although as one might expect, more mutations were detected in the tumor tissue, including MTOR and BRCA1; however, the pathogenic PDGFRA variant (c.1939A>G, p.Ile647Val) was found in the urinary exosomes from both UC and NanoPoms preparations, but not in the tumor tissue cells (Fig. 2B). It is worth mentioning that the PDGFRA variant (c.1939A>G, p.Ile647Val) has been reported as the tumor marker from the bladder urothelial carcinoma and the gastrointestinal stromal tumor 36,37 .
We also analyzed urinary exosome-derived DNA mutations using droplet digital PCR (ddPCR) from both UC and NanoPoms preparations. A total of 30 bladder cancer patient urines were analyzed with 10 healthy individuals as the control group. With the same exosome DNA input (10 mg), EGFR (Thr790Met) and TERT (C228T and C250T) were both detected in Fig. S4. We observed much higher signal amplitudes from NanoPoms prepared exosome DNAs than that from UC approach (Fig. 2c and Fig. S5). The average patients' EGFR Wt copy number is 3185.4 ± 468.3 from NanoPoms approach, which is 12.8-fold higher than that from UC approach (248.9 ± 46.4) with 3-fold higher mutation detection e ciency (Fig. S4). The overall detection signal to base ratio from patient group is statistically higher than that from control group (Fig. 2C), indicating the signi cant diagnostic value (Fig. 2D) for developing liquid biopsy and noninvasive diagnosis of BC using urinary exosomes from NanoPoms preparation. In contrast, UC-based e preparation is unable to differentiate patient group from the healthy control group (p >0.05, Fig. 2C).
Interestingly, we also observed EGFR heterozygous mutation in three BC patients while conducting ddPCR analysis of NanoPoms prepared urinary exosome DNAs (Fig. S5). In contrast, UC isolates from the same patients 2 and 3 did not show such heterozygous mutation (Fig. 3A and Fig. S5). In order to further validate this observation, we obtained the matched patient plasma and buffy coat with white blood cells (WBC) as the control. NanoPoms preparation allows to pull out marker speci c exosome populations based on the exosomal surface markers (CD9, CD63, and CD81) to match urinary exosome populations, which avoids the interferences from other microvesicles or non-disease associated vesicles. Afterwards we used Sanger sequence to con rm the presence of the EGFR heterozygosity for three patients. Results were consistent with ddPCR analysis from NanoPoms preparation (Fig. 3B). As expected, the EGFR heterozygosity was not detected from wide-type control WBCs from matched patients. These results clearly support that marker speci c capture and release enabled by NanoPoms method can signi cantly enrich tumor-associated exosomes for sensitive mutation detection. Although the UC preparation yields larger numbers of vesicle particles, their speci city and purity to tumor-associated exosomes are much less than NanoPoms preparation. NGS analysis of urinary exosome RNAs. Analyzing RNAs within urinary exosomes has been emerging with needs for non-invasive, early detection, and timely medical checkup of BC 38,39 . Exosome long noncoding RNAs (lncRNAs) PVT-1, ANRIL and PCAT-1 have been reported as the novel biomarker in BC diagnosis [40][41][42][43] . However, NGS pro ling of microRNA from tumor derived urinary exosomes from BC patients has not been exploited. In this study, we analyzed urinary exosome microRNA NGS pro les from both BC and healthy individuals.
The distribution of exosome small RNA categories from NanoPoms preparation showed more lncRNAs in both the BC group and healthy control group (42% from NanoPoms vs. 18.9% from UC) (Fig. 4A, Table  S2). In contrast, UC preparation leads to the higher percentage of tRNA. Although the exact role of exosome lncRNAs is not well understood yet, several studies have showed exosomal lncRNAs are novel biomarkers in cancer diagnosis and are highly associated with cancer progression and cellular functions [44][45][46] . Currently, only a small number of lncRNAs has been investigated which partially due to the inconsistency imposed by exosome preparation methods 47 . We further look into the top 100 miRNAs expression pro les as shown in Fig. 4B. The heatmap clustering analysis indicates the clear differentiation between BC group and healthy control from NanoPoms exosome preparation, in contrast to UC preparation. We also investigated the in uence of photo cleavage process during the exosome harvesting on the integrity of overall exosome miRNAs ( Fig S6) and did not observe any signi cant differences or impair on miRNA pro les.
In order to further interpret urinary exosome miRNA pro les and characterize the in uences imposed by sample preparation steps, we used the volcano plot to analyze the statistical signi cance (P value) versus fold-gene expression changes from both UC and NanoPoms preparations. It is interesting to note that top 10 miRNAs were highly enriched from the NanoPoms preparation, including hsa-miR-3168, hsa-miR-92b-5p, hsa-miR-891a-5p, hsa-miR-934, and hsa-miR-6785-5p ( Fig. 4C and Table S3). We searched the reported miRNA functions and found those miRNAs were reported as the cancer relevant markers speci cally sorted into exosomes (Table S3). For instance, hsa-miR-3168 has been reported to be enriched in exosomes via a KRAS-dependent sorting mechanism in colorectal cancer cell lines 48 and is known as the melanoma mature miRNA 49 . The miR-92b-5p has been found to play a critical role in promoting EMT in bladder cancer migration 50 . The hsa-miR-934 is an essential exosomal oncogene for promoting cancer metastasis 51 . NanoPoms exosome preparation offered much higher molecular relevance for identifying tumor associated biomarkers, which is crucial for exploring more speci c targetable cancer biomarkers.
Proteomic analysis of urinary exosome proteins. The urinary protein biomarkers could enable highly signi cant clinical values for the cystoscopic evaluations in BC diagnosis. EDIL-3 (Epidermal growth factor (EGF)-like repeat and discoidin I-like domain-containing protein 3) and mucin 4 (MUC 4) both have been reported in exosomes puri ed from BC patient urines 52,53 . We selected generic exosome markers CD9, CD63, and TSG101, as well as the EDIL-3 and MUC4 for Western blotting analysis of urinary exosome proteins prepared by UC and NanoPoms methods, with the human bladder carcinoma cell line HTB9 as the control (Fig. 5A).
The generic exosomes markers CD9, CD63, and TSG101 were consistently expressed in urinary exosomes, HTB9 cells and their exosomes, which indicates consistent isolation of exosomes. The expression level of EDIL-3 is signi cantly higher in BC patients than healthy individuals, but not in the tumor cell line or their exosomes from conditioned media. MUC4 protein marker was only observed in the human urinary exosomes and HTB9 exosomes, but not in HTB9 cells. This observation supports the previous report that EDIL-3 and MUC 4 are highly promising biomarkers in developing urinary exosomebased BC diagnosis and prognosis tests 52 . The proteomic pro ling of urinary exosomes from NanoPoms preparation was shown in Fig. 5B, and identi ed proteins were compared with the ExoCarta Exosome Protein Database and the Urinary Exosome Protein Database. Several proteins associated with exosome biosynthesis were observed, such as proteins PIGQ and PAPD7 involved in Golgi apparatus, the cytosol protein S100-A7 and A9 found within the exosome lumen which is engaged with natural membrane budding process during multivesicular body formation. We also observed a diverse group of cytosolic enzymes (glyceraldehyde-3-phosphate dehydrogenase) and cytoskeletal constituents (actin, Beta-actinlike protein 2 ACTBL2, and myosin-9). Although the majority of proteins are shared identi cations between BC patient and healthy control groups (~65%), as well as the databases we used, interestingly, we found 10 proteins which are uniquely identi ed only from BC patient using NanoPoms preparation (Table S4). Those proteins have previously been reported to be associated with bladder cancer metastases, including IRAK4 54 , KRT23 55 , and RALGAPA2 56 (full list in Table S4). Also 4 proteins were found uniquely in the healthy group using NanoPoms preparation, but not reported by ExoCarta and Urinary Exosome Protein Databases. From the Human Protein Atlas database (https://www.proteinatlas.org/), those proteins are intracellular and associated with vesicles, Golgi apparatus, and secreted pathway. The identi cations are broadly consistent with that expected for exosomes and compatible with other researchers' investigations 57 . Approximately 35% of proteins do not overlapped between the BC patient and the healthy control, which further support the utility of NanoPoms prepared exosomes for diagnosis of BC.
Identi ed proteins were classi ed by encoding genes which indicate the majority are located within membranous vesicles, cytosol, cytoplasm, and the cytoskeleton, and some are located in Golgi (Fig. 5C). The biological processes associated proteome revealed signi cant associations with the regulation of biological process, metabolic process, response to stimulus, cell organization and biogenesis, transport, and the cell death. The protein binding molecular function from this proteome is dominant. Results exhibit good speci city to exosomal proteome, indicating NanoPoms preparation could provide a pure and high-quality exosome, which could facilitate the important research area in EV proteomics and multiomics.
In vivo biodistribution study of NanoPoms prepared exosomes. The NanoPoms preparation of exosomes via marker speci c capture and release is able to collect intact, pure, and homogenous exosome subtypes. Due to the on-demand, light-triggered release process, the molecular engineering, such as the surface modi cation, drug loading, or dye labelling, can be implemented to immunomagnetically captured exosomes before washing and releasing. This protocol avoids the redundant post puri cation of small molecules from isolated exosomes, which is often challenging and causes contaminations. For instance, the remaining free dye during in vivo tracking of exosomes could cause false signals with longer distribution half time, unspeci c staining, or tissue accumulation 58 . In this study, we prepared exosomes from bladder tumor HTB9 cells and non-malignant HEK cells with DiR labelling for intravenous tail injection into BALB/cJ mice. The buffer solution from beads washing step (without exosomes) was used as the negative control. From these representative images in 24, 48, and 72 hr time intervals post injection (Fig. 6A), organs were harvested and imaged ex vivo in the time intervals of 48 and 72 hr to minimize signal interference (Fig. 6 B and C). To rule out of the signal originating from the blood in the organs or from the free dye, we normalized exosome tracking signal with the negative control signal to a rm the in vivo tacking.
In fact, the negative control images did not show much detectable signals indicating no remaining free dye background signal during in vivo tracking of exosomes. By further observing the harvested organs, HTB9-derived exosomes exhibit different biodistribution pro le in lung, liver, kidney, spleen, heart, and brain, as compared to exosomes isolated from the non-malignant HEK293 cells. Exosomes prepared from the HTB9 tumor cells were more concentrated in the liver and spleen with gradually increased intensity from 48 hrs to 72 hrs post injection. In contrast, non-malignant HEK293-derived exosomes tend to spread from liver to lung and spleen after 48 hrs post injection. Although HTB9 exosome biodistribution pro le has not been reported elsewhere previously, the HEK293 exosomes biodistribution pro le is consistent with reported study in C57BL/6 mice 58 . Fig. 6C provides the repetitive and quantitative analysis of biodistribution pattern over time. The results potentially indicate the distinctive biodistribution pro le from cancer-associated exosomes which could be very important for understanding tumor cell-mediated communications within the microenvironment. Currently, substantial efforts have been made for using exosomes as therapeutic agents or delivery vehicle in vivo. Thus, being able to reproducibly prepare pure and homogenous exosomes is critical for maintaining consistent biodistribution patterns. During the entire in vivo study, we did not observe any adverse effect and NanoPoms prepared exosomes are well tolerated.
Recent and substantial researches have shown the heterogeneity of EVs 20,62-66 in terms of density, molecular cargos, and morphology, which are even released by a single cell type 15,16,67 . Our recent study also observed that molecular packaging of secreted EVs or exosomes is highly variable upon the change of cellular culture environment as well as surrounding community 68 . Thus, the more advanced analytical methods are urgently needed to be able to decipher such heterogeneity in precision. Additionally, for therapeutic delivery, the well-de ned molecular components from the homogenous exosome population is also critical to precisely maintain controllable biodistribution pattern and delivery behavior 69 . Due to the unique 3D nano pom poms structure and speci c marker de ned capture-release process, our developed isolation approach can prepare pure and homogenous exosome subpopulations which enrich tumor associated biomarkers. In our study, the NGS and ddPCR analysis demonstrated that DNAs isolated from NanoPoms prepared exosomes are enriched for tumor-associated DNA mutations which are highly relevant to the bladder cancer. This evidence further supports that speci c cancer-associated biomarker are enriched in exosome type urinary EVs and can serve as surrogates for tumor cells.
The miRNAs represent the most dynamic nucleic acid cargos in exosomes, which is relatively sensitive to external stimulus and changes. Thus, in order to gauge the impact of light release process on exosome isolation via NanoPoms approach, we compared miRNA pro les with or without light release process, which did not show statistically signi cant differences based on dendrogram clustering analysis ( Fig. 4A and Fig. S6). The light release process also is able to ensure the speci city via releasing captured exosomes only, to avoid non-speci c binders. This data supports the quality and integrity of NanoPoms prepared exosomes as a novel, rapid, and easy-to-use method. Currently, although urinary miRNA pro ling is highly essential for BC diagnosis, such study and relevant database have not been fully established yet. NanoPoms based exosome sample preparation could potentially speed up this research direction by offering much simple and speci c exosome preparation.
The urinary exosome cargos at the protein level from our study reveals the consistent expression of exosomal proteins CD9, CD63, and TSG101 from both patient urinary exosomes and cell lines using UC and NanoPoms preparations. In contrast, EDIL-3 levels have been observed much higher in BC patient urinary exosomes compared to healthy individuals which is consistent with reported literature 52 , indicating the high-quality preparation of exosome using NanoPoms approach (Fig. 5A). Further, the proteomic pro ling also supports that NanoPoms prepared urinary exosome proteins can be used to differentiate BC disease from healthy status (Fig. 5 B and C, and Table S4) with unique identi cation of pathogenesis relevant exosomes proteins, suggesting a promising avenue using NanoPoms prepared exosomes to develop non-invasive bladder cancer diagnosis.
Overall, we identi ed 10 more miRNAs and 10 more proteins which are uniquely and highly expressed only from the BC patient using NanoPoms preparation, which potentially expanded the landscape of targetable cancer biomarkers. In order to further prove the integrity and biological activity of NanoPoms prepared exosomes, in vivo biodistribution study exhibits distinctive distribution patterns between tumorassociated exosomes and non-malignant exosomes (Fig. 6). This result may indicate that different subtypes and sources of exosomes could have impact on the performance of drug delivery while using exosomes as the carrier. To date, the therapeutic potential of different subpopulations of exosomes is not well known. It has been discussed that possibly only a small fraction of the exosomes from a cell can mediate the therapeutic effects 69 . Thus, the reproducible isolation of speci c exosome subpopulations is essential to support the development of exosome-based therapeutic delivery. The speci c isolation and enrichment of exosome subtypes enabled by NanoPoms approach with marker de nition could open a new avenue for preparing pure and homogenous exosomes with improved therapeutic e cacy.

Declarations Data Availability
All data generated or analysed during this study are included in this published article (and its supplementary information les). The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.  (C) The immune gold nanoparticle staining TEM imaging of captured exosomes fully covering Nano pom-poms surface. Captured EVs are con rmed by antiCD63 gold nanoparticles. The insert shows the captured single exosomes in the size range of ~100 nm with three gold nanoparticles bound (~10 nm).
(D) Nanoparticle tracking analysis of NanoPoms isolated exosomes with much narrower size distribution in comparison with UC isolated EVs. (E) Nanoparticle tracking analysis of the size of NanoPoms isolated exosomes (n=4), compared with ExoEasy isolation (n=4), which showed reproducible and smaller size of exosomes from NanoPoms preparation. (F) SEM images showing the dense exosomes are captured covering the surface of Nano pom-poms, and can be completely released via on-demand photo-cleavage.
After release, intact exosomes can be harvested for downstream multi-omic analysis including next generation sequencing of DNAs, RNAs, western blotting and proteomic analysis, as well as in vivo study.   The NGS analysis of small RNAs from bladder cancer patient urinary exosomes. (A) The distribution of small RNA categories from both NanoPoms and UC prepared urinary exosomes in BC patients and healthy individuals. (B) Heatmap with dendrogram clustering analysis depicts the top 100 highly expressed miRNAs from urinary exosomes isolated from both BC patient and healthy individual using UC, NanoPoms, and NanoPoms without light release process. Red color indicates a higher expression z-score. Hierarchical clustering was performed, using the Spearman correlation method. NanoPoms isolation approach with or without light release processes have been clustered together due to higher similarities in their transcript expressions. (C) Volcano plot analysis depicts the most biologically signi cant urinary exosome miRANs with large fold changes identi ed by using NanoPoms preparation compared to UC preparation. Top 5 highly signi cant miRNAs are labelled in plot, which are from NanoPoms preparation.  In vivo biodistribution analysis of NanoPoms prepared exosomes. (A) Representative IVIS images at 24 hours, 48 hours, and 72 hours post-injection of live mice. The HTB9 tumor cell derived exosomes and non-malignant HEK cell derived exosomes with DiR labelling (2.0 ×109 particles/ml) were prepared by NanoPoms approach for intravenous tail injection into BALB/cJ mice. The buffer solution without exosomes was used as the negative control. (B) Representative IVIS images of harvested organs (lung, liver, kidney, spleen, heart, and brain) at 48 hours and 72 hours post injection from mice. (C) The uorescence signals normalized with negative control from IVIS images in each organ harvested at 48 hours and 72 hours post injection (n=2, mean ± SD).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.