A comprehensive analysis of supermere, exomere, and extracellular vesicle isolation and cargo in colorectal cancer

SUMMARY

Biofluids contain a heterogeneous mixture of extracellular vesicles and non-vesicular nanoparticles (including exomeres and supermeres) that transport a diverse array of proteins, RNA, and lipids. Our previous efforts to characterize the contents of these carriers in colorectal cancer relied on 2D culture systems requiring large-scale setups and time-consuming ultracentrifugation-based isolation. To streamline this process, we have combined 3D hollow-fiber bioreactor production and fast-protein liquid chromatography-based size-exclusion chromatography. Here, we compare the impact of culture methods and purification strategies on small extracellular vesicle, exomere, and supermere cargo. Proteomic analyses show consistently distinct profiles for extracellular vesicles, exomeres, and supermeres regardless of culture conditions or isolation method. In contrast, these two variables influence small RNAs, their base modifications, and lipidomic profiles. We present an online tool to query these and future secretome datasets (https://superomics.shinyapps.io/browse).

Graphical Abstract

In brief

Supermeres are the newest addition to the world of secreted extracellular vesicles and nanoparticles. Here, Tutanov, Massick, et al. present new methods for producing and isolating supermeres and comprehensively compare their protein, RNA, and lipid composition to that of extracellular vesicles and exomeres, with multiomics data available via a companion web application.

INTRODUCTION

Extracellular vesicles (EVs) and non-vesicular extracellular nanoparticles (NVEPs) are increasingly recognized to be important mediators of cell-cell communication. EVs are particles delimited by a lipid bilayer that are secreted from all cell types and transport bioactive signaling payloads, including proteins, small RNAs (sRNAs), and lipids.¹ Small EVs (sEVs, <200 nm) are either released by multivesicular endosomes fusing with the cell membrane to release their intraluminal vesicles (exosomes) or by direct budding from the cell membrane (small ectosomes).^1–3 EV cargo may serve as biomarkers and therapeutic targets, but a translational roadblock is the intrinsic heterogeneity of EVs and variable methods for their isolation.^1,4 Exomeres and supermeres are two more recently identified amembranous NVEPs that also appear to be released by all cell types and contain distinct cargo.⁵ Exomeres (~28–50 nm) transfer unique cargo classes^5–7 and can functionally alter recipient cell glycan chains via transfer of beta-galactoside alpha-2,6-sialyltransferase 1 (ST6GAL1) and activate epidermal growth factor receptor (EGFR) signaling via the cognate ligand, amphiregulin (AREG).⁶ While exomeres were discovered by David Lyden’s group using asymmetric flow-field flow fractionation,⁷ we found that exomeres could be efficiently isolated by a simplified method of high-speed ultracentrifugation (UC) of the supernatant from the EV pellet.⁶ We discovered supermeres (~22–32 nm) by performing high-speed UC of the supernatant of the exomere pellet.⁵ Supermeres transport distinct proteins and sRNAs, along with lipids as described in this work, and are capable of transferring drug resistance to recipient cells.⁵ Compared to EVs, considerably less is known about NVEP biology,⁸ but identification of specific cargo may provide important clues into NVEP biogenesis and function.^8,9

An additional challenge to the study of EVs and NVEPs is how laborious and time-consuming UC-based methods are for obtaining sufficient purity and yield for full-scale characterization and functional studies. Indeed, while differential UC remains a recognized gold standard method for EV isolation, it is known to result in NVEP co-isolation if done without proper downstream purification steps like density gradient ultracentrifugation (DGUC). Additionally, UC-based exomere and supermere isolation is approximately a 3- to 4-day process, requiring rotors capable of reaching high centrifugal forces.

This study addresses these limitations by providing scalable options for culture conditions and purification strategies for EVs and NVEPs and rigorously comparing carrier composition using multiomics approaches. For the present study, we have chosen to use DiFi, a human microsatellite stable (MSS) colorectal cancer (CRC) cell line that expresses high levels of EGFR.¹⁰ DiFi cells have been studied extensively in our prior EV/NVEP-related work⁵ and were selected as a model cell line to examine emerging technologies by the Extracellular RNA Communication Consortium (ERCC).^11–13 We have extended our multiomics analysis presented here to include additional proteomic data of another human CRC cell line, CC-CR.^14,15 Culturing cells in hollow fiber bioreactors can increase yields of EVs and EV activity compared to 2D-produced material.^16–20 However, it is unknown how bioreactor culturing impacts NVEP content and yields. For more rapid purification of exomeres and supermeres directly from concentrated medium, we used fast protein liquid chromatography (FPLC)-based size-exclusion chromatography (SEC) as an alternative to UC.²¹

Herein, we show that 2D and 3D culture conditions maintain the key protein constituents in EVs, exomeres, and supermeres; however, sRNA cargo, RNA post-transcriptional modifications, and bioactive lipids were strongly affected by 3D culturing and FPLC-SEC isolation. To facilitate meaningful access to the extensive multiomics data generated in this study and enable readers to explore specific protein, sRNA, and lipid cargo beyond those presented here, we developed an online discovery tool (https://superomics.shinyapps.io/browse/).

RESULTS

3D culturing increases particle yield, and FPLC-SEC provides a time-effective alternative for isolation of supermeres and exomeres

Our previous characterization of EVs, exomeres, and supermeres was performed with fractions isolated by UC of 2D cultured medium. Here, in addition to UC isolation of sEVs, exomeres, and supermeres from 2D cultures, we extended our analysis to incorporate a 3D cultured source of the material. We also applied a novel FPLC-SEC-based method for exomere and supermere isolation, along with cushion-DGUC (C-DGUC) purification of the sEV fraction, and analyzed these carriers’ proteins (immunoblotting, liquid chromatography-tandem mass spectrometry [LC-MS/MS]), small RNAs (sRNAs) such as AlkB-mediated RNA (de)methylase-based sRNA sequencing (ARM-seq), post-transcriptional modifications (epitranscriptomics), and bioactive lipids (lipidomics) across matched NVEPs and EV-containing fractions.^{5,6,16,22–26}

Hollow-fiber bioreactors, which have been reported to increase the secretion of protein factors from cells, were used to generate 3D conditioned medium.^17–20 While hollow fiber bioreactors have been shown to have increased yields of EVs and increased specific activity, little is known about the impact of 3D culture conditions on NVEPs. Here, we observed that NVEP and EV protein yields were increased by 20-fold in 3D culture-produced material on a per volume basis, compared to 2D cultures (Table S1).

We proceeded to develop a rapid non-aggregating FPLC approach based on SEC to isolate exomeres and supermeres as an alternative to the previously reported UC method. To establish the NVEP-specific fraction ranges on the FPLC system, UC-purified supermeres and exomeres from 2D cultures were injected and analyzed by FPLC-SEC. The distribution of total protein across FPLC-SEC fractions showed that UC-purified supermeres eluted at fractions 19–24 (Figure 1A), while UC-purified exomeres were observed to be composed of three major species that are consistent with sEVs, exomeres, and supermeres (Figure 1A).

Figure 1. Characterization of EV and NVEP samples.

(A–F) Characterization of FPLC-SEC purification of exomeres and supermeres.

(A) Protein profile of FPLC-SEC fractionated UC-purified exomeres (green) and supermeres (blue).

(B–F) Protein (B), lipid (C), rhodamine-PE (D), RNA (E), and TGFBi (F) profiles of FPLC-SEC fractionated DiFi cell concentrated media.

(G and I) EV marker validation by immunoblot of EV and NVEP samples from wild-type DiFi cells (G) and DiFi cells expressing TGFBi-NG-His (I).

(H) Electron micrographs of EV and NVEP samples. Fields are at a direct magnification of 30,000×, dotted boxes depict additional 2× magnification. Scale bar, 100 nm.

(J) Nanoparticle tracking analysis of C-DGUC-purified media from DiFi cells expressing TGFBi-NG-His. (E and J) Data are represented as mean ± SEM. See also Figures S1 and S10 and Table S1.

To determine if UC-purified supermeres show similar fractionation profiles to FPLC-purified supermeres under 3D culture conditions, concentrated medium from 3D cultures was separated by FPLC-based SEC. The distribution of total protein showed a clear, dominant supermere species secreted from these cells within fractions 19–24 (Figure 1B). To determine if 3D culture-produced supermeres harbor phospholipids, the distribution of phosphatiylcholine (PC) across 3D medium fractions was quantified by colorimetric assays. Negligible PC levels were detected in fractions corresponding to supermeres using this approach (Figure 1C). Supermeres were also not found to co-fractionate with cholesterol or triglyceride (TG) by FPLC-SEC, suggesting that these particles do not transport cholesterol or neutral lipids (Figure 1C). To determine if supermeres isolated from 3D cultures bind and associate with phospholipids, concentrated medium was spiked with fluorescence-labeled phosphatidylethanolamine (PE, PE-Larissa rhodamine B), and samples were dialyzed to remove unbound fluorescence-labeled PE prior to FPLC-SEC separation. Under these conditions, we failed to observe fluorescence in supermere-associated fractions, suggesting that supermeres do not have phospholipid-binding capacity or a monolayer/bilayer membrane (Figure 1D). We did observe a small signal of fluorescence PE associated with potentially larger carrier species, likely corresponding to sEVs in the medium (fraction 10) (Figure 1D). As NVEPs likely do not share the same phospholipid-binding capacity that is observed with EVs and are at much greater concentrations than EVs, tagged phospholipids provide an innovative approach to separate EVs from NVEPs in complex biofluids. More broadly, our results show that supermeres can be quickly isolated from 3D culture medium by FPLC-SEC for compositional and functional analyses.

We have previously reported that 2D culture-produced, UC-purified supermeres are enriched with RNA cargo when compared to EVs and exomeres.⁵ To examine RNA cargo from 3D culture-produced supermeres, total RNA was isolated and assessed by pre-labeling concentrated DiFi medium with SYTO RNA Select, a lipid-penetrating, RNA-specific dye that fluoresces when bound to RNA.²⁷ We found that 3D culture-produced supermeres harbor robust levels of total RNA, consistent with previous results using 2D culture conditions (Figure 1E).⁵ From our previous proteomic analyses of DiFi-secreted supermeres, one of the most abundant proteins identified was transforming growth factor-beta-induced/BIG-H3 (TGFBi/BIGH3).⁵ To determine if supermeres isolated from 3D culture conditions also carry TGFBi, ELISA and immunoblotting were used to quantify TGFBi protein levels across FPLC-SEC fractions. We found that TGFBi was stably associated with supermeres isolated from DiFi cells cultured in 3D conditions (Figures 1F, S1A, and S1D). Across the FPLC-SEC fractions, we detected TGFBi protein overlapping with the EV and exomere peaks, but the highest levels were in supermere-containing fractions, including stable multimers and potentially cleaved smaller isoforms of TGFBi (Figures 1F and S1A). Based on these findings, we introduced a lentiviral TGFBi expression construct with a His tag into both DiFi and CC-CR cells for nickel column affinity purification of TGFBi-containing supermeres (Figure S2B).¹³ Immunoblots for the supermere markers extracellular matrix protein 1 (ECM-1) and the cleaved extracellular domain of discoidin domain receptor 1 (DDR1; DDR1-ECD) also showed enrichment in FPLC-SEC supermere fractions (Figures S1B–S1D). Full-length, uncleaved DDR1 that migrates differently from DDR1-ECD was present in EV pellet fractions (Figure S1D). These results suggest that 3D culturing methods combined with FPLC-SEC purification allow for isolation of robust levels of TGFBi⁺, RNA-rich, phospholipid-poor supermeres that are suitable for downstream experimentation without the need for time-consuming UC purification steps.

DGUC isolation of sEV presents more homogenous fractions with less NVEP contamination compared to traditional UC

Small EVs were isolated from medium produced from 3D growth conditions using C-DGUC fractionation for comparison to 3D-derived EV pellets (EVp) and previous DGUC results for 2D-derived DiFi samples.⁵ EV-containing medium was layered above a 2-mL iodixanol cushion and ultracentrifuged to concentrate EVs at the medium-iodixanol interface. EV-containing fractions were then transferred to the bottom of a flotation iodixanol step-gradient to isolate pure EVs.^24,28 Using this method, we found that EVs fractionated into two peaks, corresponding to fractions 6 and 7 (classic sEVs [EV]), and fractions 9 and 10 (heavy EVs [hEV]) of the 12-fraction gradient (Figures 1G–1J).²⁴ Heavy EVs appear to represent a 3D bioreactor-associated vesicle species, which is not typically observed in 2D cultures.⁵ Non-vesicular dense material did not float but instead remained at the bottom of the gradient in fraction 12 (NV). Characterization of EVs and NVEPs was carried out according to minimal information for studies of extracellular vesicles (MISEV) standards.² Specific EV markers (flotillin-1 [FLOT1], tumor susceptibility gene 101 protein [TSG101], programmed cell death 6-interacting protein [ALIX], and heat shock protein [HSP70]) were enriched in gradient-purified sEVs, hEVs, and EVp from 2D- and 3D-cultured samples and were absent or reduced in NVEP samples (Figure 1G). To assess the purity, shape, and electron density of NVEPs and EVs, isolated particles and vesicles were imaged by transmission electron microscopy (Figures 1H and S10). Notably, 3D culture-produced supermeres appeared to be uniform and highly pure, particularly from FPLC-SEC-isolated material (Figure 1H). Exomeres encompassed a more heterogeneous particle mix, while EVs displayed their classic shape and size (Figure 1H).

As an additional control, conditioned medium from DiFi cells expressing neon green-tagged TGFBi was fractionated by C-DGUC as above, analyzed by immunoblot for EV markers, including CD63, and assessed by nanoparticle tracking analysis (NTA) to establish EV concentrations (Figures 1I and 1J). Fractions 6 and 7 as well as 9 and 10 had CD63 as a cargo, representing light and heavy EVs. Fraction 12 was missing EV markers and represents non-vesicular material with some exomeres but without substantial supermere content.⁵

NVEP protein cargo is consistent across methods

We first analyzed the protein cargo based on culture conditions and an isolation method. LC-MS/MS proteomics was performed on purified sEVs, exomeres, and supermeres with data reported as normalized spectral counts (Table S2). Hierarchical clustering of 3D culture-produced material and principal-component analysis (PCA) of both 2D- and 3D-cultured samples showed that supermeres, exomeres, and EVs were distinct entities when based on protein composition (Figures 2 and S2). Furthermore, C-DGUC-purified samples were clearly distinct from all other carriers, independent of culturing (3D or 2D) or processing (UC or FPLC) conditions (Figures 2A and 2B). Of note, supermere and gradient-purified EV protein cargo were the most distinct from one another. Notably, EVp samples seemed to contain more NVEP proteins in comparison to gradient-purified EVs, possibly due to the high centrifugal forces inherent to the method (UC) and the generation of aggregates in the heterogeneous pellets. The non-vesicular fractions from density-gradient EV purification were clearly distinct from other fractions, as described above. These results also agree with previous findings from multiple labs that EV pellets contain non-vesicular material.^5–7,29,30 Moreover, each fraction displayed a distinct profile of RNA-binding proteins and signaling factors (Figure S4A; Tables S3 and S4). As expected, EVs were enriched with marker proteins associated with EV biogenesis, such as ALIX and syntenin-1 (SDCBP), along with a number of full-length transmembrane proteins (Figures 2A–2D, S2D, S3C, S4B, and S4C). In contrast, supermeres and exomeres were enriched with cleaved extracellular domains of transmembrane proteins, including hepatocyte growth factor receptor (MET) and EGFR (Figures 2E, 2F, S3A, S3B, S4B, and S4C), as previously observed.⁵

Figure 2. Proteomic analysis of NVEPs and EVs.

(A) Heatmap of the top 20 most abundant proteins from 3D-derived EV and NVEP samples; log2-transformed median normalized spectral counts.

(B) PCA of the proteins identified in EV and NVEPs samples. Samples EVp 2D/3D, exomere 2D/3D, and supermere 2D/3D were purified with UC from 2D and 3D, respectively; sEV, hEV, and NV were purified by C-DGUC; samples marked FPLC purified by FPLC-SEC; and samples marked FPLC/His purified with a nickel column after FPLC-SEC.

(C) UpSet plot of supermere samples’ proteomics results.

(D–F) Venn diagrams of the top 100 most abundant proteins from EV (D), exomere (E), and supermere (F) samples. Proteins comprising the overlap between all samples for a given carrier are listed.

(G) FAVS analysis of 2D UC supermeres using TGFBi and ENO1. See also Figures S2–S4 and Tables S2–S4.

Since TGFBi is one of the most abundant supermere proteins, we used TGFBi and alpha-enolase (ENO1), another supermere marker protein, for flow-cytometric fluorescence-activated vesicle sorting (FAVS) analysis of 3D UC supermeres. FAVS analysis showed 53% and 39% of supermeres were positive for TGFBi and ENO1, respectively, with 33% identified as double-positive and about 34% identified as double-negative, high-lighting the heterogeneity of different supermere complexes on an individual particle basis (Figure 2G). Importantly, the majority of the supermere proteins identified by LC-MS/MS (576 proteins, including TGFBi, ENO1, ECM1, and DDR1) were found to be present in each supermere sample, regardless of growth condition, purification strategy, or cell line of origin (Figures 2C and 6A). This overlapping subset was found to be enriched for RNA-binding proteins with 24% (137/576) of the proteins previously identified by the Gene Ontology consortium as such (Table S4). Interestingly, according to PCA analysis, 3D supermeres from different cell lines were found to be closer to each other than to 2D DiFi supermeres (Figure 2B). Indeed, both DiFi and CC-CR supermeres displayed similar profiles of the most abundant proteins with a few differences, which may be due to inherent differences between cell lines.

Figure 6. Superomics web portal.

(A) Web portal encompassing multiomics results generated in this study.

(A–D) Examples of querying proteomics (A), sRNA-seq (B), and lipidomics (C, D) data for samples grouped by EV/NVEP type (A–C) or individual (D) samples, with faceting for isolation method (B), growth conditions (D), or both (A, C).

sRNA profiles are method dependent

Next, we investigated the impact of specific growth and isolation methods on sRNA profiles, including 3D culturing and FPLC-SEC isolation compared to 2D culturing and UC isolation approaches (Figure 3A). We previously used conventional sRNA sequencing (sRNA-seq) to profile supermere-associated sRNAs isolated by UC from medium post-2D culturing of DiFi cells and found abundant reads corresponding to fragments of transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs), two classes of RNA transcripts known to harbor extensive post-transcriptional nucleotide modifications.^26,31–33 Conventional sequencing methods fail to capture modified (e.g., methylated) sRNAs due to modification-related inhibition of 1^st strand cDNA synthesis.³⁴ To overcome this barrier, ARM-seq approaches were used to remove the following specific methylations prior to first-strand synthesis: 3-methylcytidine (m³ C), 1-methyladenosine (m¹ A), 1-methylguanosine (m¹ G), and 2,2-dimethylguanosine (m²₂ G).³⁴ Using this approach, supermeres purified from 2D and 3D culture conditions were found to have a higher percentage (proportion) of host (human) sRNAs compared to other types of RNAs. We observed considerable differences between methods and particle classes based on the proportion of assigned reads; however, supermeres consistently showed the highest levels of human (host) RNA levels based on normalized read counts, while EVs purified by C-DGUC showed the least amount of host RNAs (Figures 3A and S5A, SB). EVps from UC pelleting also showed high levels of host RNAs for both 2D and 3D growth conditions, but those findings could be due to the contribution of non-EV carriers contaminating this heterogeneous pellet (Figures 3A and S5B). The sRNA profiles of supermeres were similar to exomeres, yet distinct from EVs. For example, EVs were found to carry abundant microRNAs (miRNAs), tRNA-derived sRNAs (tDRs), and rRNA-derived sRNAs (rDRs), regardless of the purification method. On the contrary, NVEP sRNA profiles were greatly affected by culturing and isolation methods. Supermeres isolated from 3D culturing conditions were observed to harbor abundant Y RNA-derived sRNAs (yDR) and rDRs based on either UC or FPLC-SEC methods; however, FPLC-SEC isolation was associated with elevated levels of rDRs (Figure 3B). Supermeres from 2D culture conditions were found to carry elevated levels of miRNAs and snRNA-derived sRNAs (snDR) compared to supermeres isolated post-3D culturing (Figure 3B).

Figure 3. Analysis of small RNA content of NVEPs and EVs.

(A) Proportion of mapped reads in different sequencing categories.

(B) Proportion of different RNA species. Shown are exomeres (Exo), supermeres (Sup), EVp and C-DGUC-purified EVs, hEVs, and NVs from 3D material.

(C) PCA of miRNA composition in different samples. EV, hEV, and NV purified by C-DGUC; samples marked FPLC purified by FPLC-SEC; and all others purified by ultracentrifugation. The line separates 2D and 3D-derived samples.

(D and E) Volcano plots comparing miRNA (D) and yDR (E) in 3D- versus 2D-derived UC supermeres.

(F–I) Volcano plots comparing miRNA (F), yDR (G), tDR (H), and rDR (I) in FPLC- versus UC-isolated supermeres. Gray dot, non-significant; green, log2FC significant; red, log10 adjusted p value (adjp) and log2FC significant; FC, fold change, p < 0.05. See also Figures S5 and S6 and Table S5.

For miRNAs, PCA showed distinct profile separation of supermeres, exomeres, and EVs across culture conditions (Figure 3C). In addition, EV profiles were largely defined by principal component 2 (PC2) toward the top of the plot. We identified many significant (padj < 0.05) differentially expressed (fold change [FC] > 1.5) sRNAs across the different methods and particle sub-classes (Table S5; Figure S6). Volcano plots comparing miRNAs from UC-purified supermeres showed a decrease in miRNAs when using 3D cell culture conditions compared to 2D culture (Table S5; Figure S6), consistent with the proportion of the miRNAs in the various particles being substantially less when isolated from 3D cultures than from 2D cultures (Figures 3B and 3D). Conversely, 3D culture-produced particles appeared to replace those lost miRNAs with other sRNAs, including more yDRs than 2D culture-produced particles (Figure 3B and 3E). This pattern was also observed for exomeres (Figure 3B). These results demonstrate that the sRNA composition of NVEPs is significantly influenced by 3D growth conditions, in contrast to EV profiles, which appear to be more stable across conditions.

Overall, the isolation of 3D culture-produced supermeres by FPLC-SEC was found to alter specific classes of sRNAs, including increased rDR reads compared to UC approaches (Figures 3B, 3G–3I). It is possible that FPLC-SEC is a gentler approach than UC for supermere isolation, which may lead to increased retention of sRNA cargo for supermeres when compared to UC. The top-most abundant miRNAs enriched for 3D culture-derived, FPLC-SEC-produced supermeres were miR-1246, miR-92a-3p, and miR-22–3p (Table S5); however, miR-1246 and miR-22–3p were also enriched in 2D culture-derived, UC-produced supermeres within this dataset and in our previous studies,⁵ suggesting that specific miRNAs on supermeres may be less sensitive than others to culture conditions or isolation methods.

3D bioreactors impact the identification of nucleoside modifications

sRNAs-derived from tRNAs and rRNAs should harbor the most post-transcriptional RNA modifications across all RNA classes based on their parental transcripts; however, miRNAs and other classes of sRNAs have also been reported to be modified as well.^{32,33,35–39} To quantify the impact of 3D culturing and FPLC-SEC isolation methods on the modification levels for all sRNAs associated with supermeres, exomeres, and EVs, total RNA was isolated and digested into single nucleosides, and modifications were quantified by LC-MS/MS with standards for 18 distinct modified nucleosides and the four main nucleosides. For supermeres, guanosine (G) levels were significantly decreased in comparison to other particles (Figure 4A), whereas cytidine (C), adenosine (A), and uridine (U) levels were significantly increased in 3D culture-produced supermeres purified by UC when compared to other particles. In contrast, supermeres purified by FPLC showed a higher proportion of unmodified G levels and less C, A, or U compared to 2D culture-produced, UC-purified supermeres (Figures 4B–4D). Therefore, G-containing sRNAs from 3D cultures are highly sensitive to carrier isolation methods. Strikingly, many of the detected modifications were found to be significantly increased in UC-purified supermeres from 3D culture conditions compared to supermeres isolated under 2D growth conditions, including ino-sine (I), m²A, m⁵C, m⁵U, dihydrouridine, m²₂G, m²G, and m⁷G modifications (Figures 4E–4L). For most modifications, we did not observe a significant difference in levels between supermeres isolated by UC or FPLC-SEC, with the exception of m⁵C and m²G (Figures 4G and 4K). Exomeres were similar to supermeres in this regard, as A, U, and five modifications (I, m⁵C, m²₂G, m²G, and m¹A) were observed to be significantly increased in exomeres isolated under 3D compared to 2D growth conditions (Figures 4C–4E, 4G, 4J, and 4K; S7A). UC-produced EV pellets were found to have higher levels of G and reduced levels of C, A, U, I, and m²A when comparing 2D to 3D growth conditions (Figures S8A–S8O). On the contrary, UC-pelleted EVs from 2D cultures showed significantly increased levels of m⁵C, m²G, m¹I, and pseudouridine (Ψ) when compared to UC-pelleted EVs from 3D bioreactor cultures (Figures S8G, S8K, S8M, and S8N). Despite some overlap, these data suggest that EVs and NVEPs transport RNAs with distinct profiles of base modifications. These observations could reflect differences in class composition or differential effects of epitranscriptomics writers and erasers under 3D and 2D culture conditions.

Figure 4. Epitranscriptomics of NVEPs and EVs.

(A–L) Ribonucleotide and modified base distribution in exomeres and supermeres for 2D and 3D fractions using UC and FPLC. Distribution of G, C, A, U, I (A–E) and modified (F–L) bases.

Data are represented as violin plots with mean, *p < 0.05; **p < 0.01; and ***p < 0.001. See also Figures S7 and S8 and Table S7.

Lipid composition is sensitive to culture condition and isolation method

Lipid profiles were quantified for NVEPs and EVs by non-targeted LC-MS/MS with internal standards, and the data were analyzed by MSDial software.⁴⁰ EVs were observed to carry substantially more lipids than NVEPs at the category (e.g., glycerophospholipids), main class (e.g., glycerophosphocholines), sub-class (e.g., PC), and species (e.g., PC 34:1) levels (Figures 5A–5D). All carriers showed similar levels of glycerolipids, despite colorimetric assays showing little to no TG in supermeres or DiFi cell material concentrated from 3D media (Figure 1C). EVs were found to be enriched with glycerophospholipids and sphingolipids (Figures 5A–5D). PCA showed a separation between supermeres and EVs or exomeres (Figure 5E).

Figure 5. Lipidomics of NVEPs and EVs.

(A–D) Lipid concentrations (top) and proportions (bottom) for top 10 categories (A), main classes (B), sub-classes (C), and species (D).

(E) PCA of the lipid species identified in EV and NVEP samples.

(F–I) Volcano plots comparing lipid species (F, H) and subclasses (G, I) in 3D- versus 2D-derived (F, G) or UC- versus FPLC-SEC-isolated (H, I) supermeres. Gray dot, non-significant; green, log2FC significant; red, log10 p value and log2FC significant; FC, fold change, p < 0.05. See also Figure S9 and Table S6.

As shown above, 3D culture-derived supermeres and exomeres failed to detectably bind to fluorescence-labeled PE spiked into media samples prior to FPLC-SEC analyses. This is consistent with supermeres and exomeres lacking a phospholipid bilayer or monolayer. By mass spectrometry, supermeres were found to contain little to no detectable levels of phospholipids, further supporting the lack of a phospholipid bilayer or monolayer shell. Therefore, we sought to define the non-phospholipid composition of supermeres and determine the impact of culture conditions and isolation methods on non-phospholipid cargo. Using LC-MS/MS, we found that supermeres carry specific bioactive lipids that could mediate cell signaling and other cellular responses in recipient cells (Table S6). To determine the impact of 3D versus 2D culturing on supermere-associated lipid cargo, we compared UC-purified carriers at each categorical level and found that 133 species, seven sub-classes, two main classes, and one category of lipids were significantly altered for UC-purified supermeres when comparing 3D and 2D culture conditions (Table S6). UC-purified supermeres from 3D culture conditions have increased phosphoinositol-ceramides and cardiolipins compared to 2D culture-derived particles (Figures 5F and 5G).

To determine the impact of FPLC-SEC, as compared to UC isolation, NVEPs and EVs were compared at each lipid level. We observed similar overall lipid profiles for exomeres between 3D and 2D culture conditions (Figures 5A–5C; S9A), with limited overall specific changes in ceramides. 3D conditions were found to alter EV-specific lipid species, including ceramides, glycerophospholipids, and fatty acyls (Figures S9B–S9C). For supermeres, we observed 105 lipid species and seven sub-class groupings that were significantly altered between FPLC-SEC or UC separation of 3D media (Figures 5H and 5I). Strikingly, FPLC-SEC increased the detection of many lipid species for supermeres, including ornithines, diacylglycerols, and cardiolipins (Figures 5H and 5I). The impact of using the FPLC-SEC isolation method instead of UC isolation for exomeres resulted in similar changes as those observed with supermeres (Figures 5A–5D). Collectively, 3D culture conditions led to increased detection of ceramides and bioactive lipids for NVEPs compared to 2D cell culture, and isolation of NVEPs by FPLC-SEC allowed for retention of specific lipids that appear to be lost during UC isolation.

Multiomics data are accessible through an interactive web application

To facilitate analysis of the proteomic, sRNA-seq, and lipidomic data generated in this and future studies, we developed an interactive web application (https://superomics.shinyapps.io/browse/) (Figure 6A). The application allows users to create dot plots (displaying levels of analytes in each individual replicate from this study) and pirate plots (displaying grouped results per EV/NVEP type) for any individual protein, sRNA, or lipid on a per-sample or per-particle basis, with optional subsetting of the plots by growth condition (3D versus 2D culture conditions) or isolation method (e.g., FPLC-SEC versus UC) (Figures 6B–6D). It also supports detailed filtering of the results and inclusion of some of our previously published EV and NVEP omics datasets,⁵ presenting an easily accessible and extensive snapshot of these extracellular carriers.

DISCUSSION

Producing a sufficient amount of purified NVEPs and EVs for biomolecular characterization and functional testing often requires large quantities of starting material. Based on the literature, 3D hollow-fiber bioreactors yield highly concentrated conditioned media, containing large amounts of secreted EVs that have high specific activity in various assays.^41–46 In this study, we have found that 3D culturing also allows for increased yields of exomeres and supermeres compared to large-scale 2D tissue culture preparations (Table S1). Other challenges of working with EVs and NVEPs are that traditional methods are time-consuming and typically either require high physical forces during UC or co-isolate NVEPs with the EV samples. To overcome these challenges, we present a fast and non-aggregating alternative strategy by using FPLC-SEC for isolation of NVEPs and employ DGUC with a cushion addition to provide further refinement of EV isolation.^22,24,28,47 Additionally, our proteomic analysis of supermeres involves isolation from both DiFi and CC-CR cells with either just FPLC-SEC or with further affinity purification using His-tagged TGFBi that is abundant in supermeres (Figures S2A and S2B).⁵ Using these methods, we expanded on our previous analysis of protein and sRNA cargo present within EVs, exomeres, and supermeres,⁵ and comprehensively evaluated the changes conferred by 3D growth conditions and FPLC-SEC purification.

In general, we have found that different macromolecules are differentially impacted by culture conditions and methods of isolation. Protein composition was found to be similar within a particle type, regardless of growth conditions or purification strategies. In contrast, sRNAs, nucleoside modifications, and bioactive lipids were found to be greatly dependent on growth conditions and isolation methods. Besides the obvious differences in growing cells in 2D plastic flasks versus 3D bioreactor cartridges, 2D cultures typically involve growing cells in serum-containing media until just prior to particle isolation, whereas cells grown in 3D bioreactors are cultured in chemically defined serum-free media. In addition, cells in 3D bioreactors are grown at a very high density for long periods of time. These differences likely underlie the observed differences in the content and diversity of cargo for secreted EVs and NVEPs.²⁴

For protein composition, we compared EVs isolated by two common methods, UC pelleting and C-DGUC. Both UC-isolated EVps and C-DGUC-purified sEVs were observed to harbor common EV protein markers, including tetraspanins, FLOT1, TSG101, and HSP70 (Figures 1G and 2A).⁴⁸ Similarly, in the proteomics data, EV preparations were enriched with typical EV markers and transmembrane proteins, e.g., integrins and EGFR. However, in concordance with previously published studies,^5,16,22 the use of UC alone results in crude EV pellets that likely represent a heterogeneous mixture of both sEVs and NVEPs, while the subsequent density gradient purification allows for the isolation of vesicles with greater uniformity and purity (Figures 2A and 2B). Analysis of the proteins that are most abundant in specific fractions showed that EVp fractions shared these proteins with either EVs (ITGAV, TSPAN8, and LGALS3BP) or NVEPs (FASN), while sEV and hEV fractions were differentially enriched for some of the marker EV proteins (CD63 and CD81) (Figure S3). Indeed, although culture conditions certainly affect specific cargo composition, the greatest difference was found between C-DGUC purification and other approaches (Figure 2B). This is not surprising as the standard UC approach for pelleting EVs includes high centrifugal forces over long times that might lead to pelleting of heterogeneous material, while DGUC allows for a more robust purification of sEVs, free of non-vesicular contaminants.^22,24 Regardless, comparing the top 20 most abundant proteins in each sEV and EVp dataset, nine proteins were shared between them, including EV hallmark proteins recommended by MISEV, such as ALIX, ITGA6, and ITGB4 (Figures 2D and S2C).

In contrast to EVs, the protein cargo of exomeres and supermeres appeared to be more consistent across purification methods (Figures 2A, 2E, 2F, and S2). Proteins that were most abundant among the specific carriers included previously described marker proteins such as ENO2 and PCSK9 for supermeres and VCP and FASN for exomeres, as well as several new candidate marker proteins for exomeres (HUWE1 and EIF3) and supermeres (ECM1, DDR1, and LCP1) (Figures S1, S3, and 6A). Metabolic enzymes were found to be specifically associated with supermeres and exomeres, consistent with our previous characterization (Figures 2A–2F, S2C, and S2D).^5,49 Interestingly, many of these metabolic enzymes have also been reported to be RNA-binding proteins (Figures S4A).⁵⁰ Exomeres and supermeres were also found to be enriched for cleaved extracellular domains of transmembrane proteins (Figures S4B and S4C), as previously reported.⁵ These shed ectodomains that are likely to confer distinct functions in contrast to their transmembrane forms, as has been described for processed MET domains,⁵¹ while protein cleavage may represent an important feature reflective of the tumor microenvironment.⁵² NVEPs were also found to be enriched with other enzymes and secreted matrix-associated proteins such as ECM-1 and TGFBi (Figures 1F, 2A, 2E, 2F, S1A, S1B, and S2). In this study, we used TGFBi and ENO1 as supermere marker proteins⁵ for flow-cytometric FAVS analysis and found that 53% and 39% of supermeres are positive for TGFBi and ENO1, respectively, with 33% being double-positive (Figure 2G). This underscores the heterogeneity of different supermere complexes when analyzed on an individual particle basis. Based on these results, we decided to isolate TGFBi-containing supermeres by introducing a TGFBi expression construct with a His tag for nickel column affinity purification of DiFi and CC-CR cells.

As expected, there were some differences in supermere proteomic composition from different cell lines and isolation methods. For instance, CC-CR supermeres contained complement components C4A and C4B, which have been linked to the regulation of ECM and immune response.⁵³ The presence of complement proteins in supermeres isolated from an MSI-H CRC line, CC-CR, might have ties to the immune responsiveness to immunotherapy.⁵⁴ In contrast, there was an enrichment of three cargoes (TGFBi, DPEP1, and DDR1) in the MSS DiFi cell fractions in comparison to CC-CR. These data are consistent with our discovery that the genes encoding these proteins are enriched in MSS CRCs at the tissue level and might reflect potential biomarkers for distinguishing MSS versus MSI-H CRC.⁵⁵ We have previously shown that these genes predict an immune-excluded phenotype, which we define as a paucity of CD8⁺ T cells in the tumor microenvironment.⁵⁵ A future direction of this work could be using the levels of these proteins in EVs and NVEPs to determine the likelihood of responsiveness to immune checkpoint inhibition. Additionally, some proteins like MAPK1, EXOSC9, and AP1M1 were found to be enriched in supermeres for both cell lines with affinity purification compared to just FPLC-SEC (Figures S2 and S3; Table S2). Most of these proteins, which appear to be NVEP exclusive in DiFi and CC-CR, have been previously reported in exosomes. Whether this is due to the heterogeneity of sEVs, a case of cargo reassigned to their proper carriers, or a result of interactions between EVs and NVEPs remains to be seen.

The majority of supermere proteins (576) reported in this study were shared between all supermere samples regardless of growth conditions, isolation method, or cell line of origin (Figures 2B, 2C, and S2A). Moreover, these shared supermere proteins were enriched for RNA-binding proteins, with 24% of them previously identified as RNA-binding (Table S4). RNA is a major cargo component in supermeres, so it is not surprising that a plethora of RNA-binding proteins were observed (Figure S4A; Tables S3, S4), including translation-associated proteins as well as a group of RNA-binding proteins that may play a role in shielding extracellular RNA from degradation¹¹ (Tables S3 and S4). As supermeres contain the majority of secreted RNA,⁵ core RNA-binding protein constituents would be expected. It could be that proteins are held within supermeres through their RNA interactions and not by protein-protein interactions. This might lead to some of the heterogeneity observed for supermeres (Figure 2G) as RNA processing in individual particles could change what proteins are carried.

For sRNA composition, we observed major differences in profiles from carriers purified by the different methods and culture conditions, particularly for yDRs in NVEPs (Figures 3B and 3E). The increase in yDRs under 3D culture conditions, compared to 2D, seems to be linked to a decrease in levels of miRNAs and snDRs. yDRs were first identified for their association with ribonucleoprotein (RNP) autoantigens in patients with systemic lupus erythematosus⁵⁶ and include the Y RNA-binding protein RO60, which we found to be increased when comparing 3D and 2D culture conditions (Table S2). How or why abundant nuclear RNPs escape cells and elicit immune responses is not known. Most early studies proposed that full-length Y RNAs act as scaffolds during the assembly of larger RNP complexes.⁵⁷ More recent studies have found complexes with misfolded RNA, which might suggest a need to remove defective Y-RNA-RO60 RNPs, perhaps by supermere-mediated export.^58,59 Y RNAs have also been implicated in DNA replication, origin binding, and proliferation.⁶⁰ As with TGFBi, Y RNAs and their fragments (yDR) may have predictive value as biomarkers in various cancers.⁶¹

miRNAs are perhaps the most well-characterized sRNAs in extracellular biofluids and carriers, largely because of their small size and the ability to be easily monitored for effects in recipient cells. miRNAs comprise the largest class of sRNAs in EVs, supermeres, and exomeres purified from DiFi cells cultured under 2D conditions. They remain the largest category of sRNA in EVs purified from 3D culture conditions, but are dramatically reduced in supermeres and exomeres isolated from 3D cultures, which are replaced by yDRs and rDRs (Figures 3B and 6B). These observations have important implications for defining the mechanisms of miRNA export to NVEPs. It is well known that miRNA sorting to EVs is selective, as EV-miRNA content does not simply reflect cellular levels.^62,63 Nevertheless, the biogenesis and cargo-sorting mechanisms for exomeres and supermeres are unknown, although the presence of metabolic proteins in NVEPs might suggest they are produced in response to metabolic changes. The mechanisms that drive selective miRNA export under 2D conditions may differ from 3D conditions; serum-free defined medium may contribute to these differences, as is evidenced by reduced miRNAs in 3D cultured EVps and NVEPs (Figures 3B and 3D).

Modifications to carrier RNAs likely confer functional changes to sRNA fragments. Modified nucleotides have been reported to alter interactions with RNA-binding proteins and regulate cellular RNA folding, splicing, maturation, and stability.^32,64,65 Little is understood about the roles of RNA modifications on cell-free sRNAs. Many cell-free sRNAs likely serve as agonists and ligands for RNA-sensing TLR7/8 on immune cells.^65–71 Some modifications, namely methylations residing on carrier sRNAs, have been reported to antagonize TLR7/8 signaling and attenuate sRNA inflammatory signaling.^72–75 For example, m⁶A, m⁵U, and m⁵C have been demonstrated to inhibit RNA signaling through TLR7/8⁷⁶. In both NVEP classes, m⁵C was significantly increased in particles isolated from 3D cultures compared to 2D for UC isolation; however, isolation of 3D particles by FPLC-SEC resulted in decreased levels of this modification on supermeres (Figure 4G). Conversely, EV pellets from 3D culture conditions harbored significantly lower m⁵C levels than EV pellets from 2D culture conditions (Figure S8G). Therefore, the choice of culture conditions differentially affects specific carrier nucleoside modifications and suggests that 3D culturing conditions for specific carriers may generate particles that confer less inflammatory pressure on immune cells upon uptake and catabolism. Although the biological functions of these modifications are unknown, they likely contribute to the overall immunogenicity of supermeres. Different strategies for culturing and purification of NVEPs may produce carriers with variable immunogenicity due to altered modifications on sRNA cargo.

It appears that NVEPs do not transport phospholipids (Figures 5A–5D). Nevertheless, we identified other potential bioactive lipids on NVEPs and found that these lipids are dependent on culture conditions and purification strategies (Figure 5). We also noted differences between growth conditions for bioactive lipid content on EVs. The delivery of bioactive lipids and subsequent activation of signaling cascades could provide an explanation for how limited numbers of EVs or NVEPs could amplify gene expression changes in recipient cells.^6,76–78 Overall, the immunogenicity of EVs and NVEPs is likely conferred by the diversity and composition of the payload delivered to recipient cells, including bioactive lipids, sRNAs, and their modifications. We report here that the composition of these stimuli on EVs and NVEPs is sensitive to culture conditions and isolation methods and likely contributes to the overall biological functions of these purified carriers.

To make the multiomics data obtained from these studies more accessible to readers interested in specific analytes, we developed an interactive web application (Figure 6). This platform represents the largest publicly available collection of multiomics data on EVs, exomeres, and supermeres isolated from the same samples under different conditions, offering valuable insights into the cargo and potential functional roles of these extracellular carriers, and enabling easy querying of specific targets to guide decisions between 3D versus 2D culture conditions and UC versus FPLC-SEC isolation methods.

In summary, we conducted a comprehensive comparison of proteins, sRNAs, nucleoside modifications, and lipids associated with EVs, exomeres, and supermeres within the CRC context. We show that the choice of 2D versus 3D culture conditions and purification method affects cargo content with important implications related to the selection of particles for experiments and characterization. The use of 3D bioreactor culture conditions and FPLC-SEC provides significant advantages for the rapid and non-aggregation-based isolation of extracellular carriers. Further refinement of FPLC-SEC methods and the establishment of new purification modalities will help uncover the degree to which NVEPs and supermeres are themselves heterogeneous and to what extent lower-molecular-weight fractions might contain free protein. We also confirm that exomeres and supermeres exist as distinct NVEP entities, reminiscent of subgroups within the EV class that have been highlighted by researchers as having specific activity independent of other subgroups.^79,80 Whether supermeres are part of a heterogeneous mix of similar-sized nanoparticles has been the subject of some debate.^79,80 The ongoing structural studies will help to address this question. While beyond the scope of this study, the multiomics data presented here offer a valuable source of insights to begin to approach some of the many still open questions, such as the interplay of EVs and NVEPs, supermere biogenesis, biological functions, and their role in cancer.

Limitations of the study: This study has several limitations. First, only two CRC cell lines are presented here as sources of EVs and NVEPs, due to the depth of the downstream analyses performed. Second, the sub-50 nm size of supermeres and exomeres makes them extremely difficult to quantify with current technologies such as NTA; protein concentrations were used to estimate particle yield. Third, sEV fractions from FPLC-SEC were not included in the downstream multiomics, thus limiting 3D FPLC-SEC datasets to supermeres and exomeres.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Robert J. Coffey (robert.coffey@vumc.org).

Materials availability

All unique/stable reagents generated in this study are available from the lead contact upon request, but we may require payment and/or a completed materials transfer agreement if there is potential for commercial application.

Data and code availability

• Multiomics results generated in this study are available as a standalone website (https://superomics.shinyapps.io/browse/). The raw data have been deposited in the publicly available databases as listed in the key resources table. Proteomics is accessible from ProteomeXchange (PXD066872), sRNAseq from the Gene Expression Omnibus (GEO: GSE304273), and Lipidomics from MetaboLights (MTBLS12859).

• All original code has been deposited at Zenodo as listed in the key resources table.

• Any additional information required to reanalyze the data reported in this work is available from the lead contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-TGFBi (“TGFBI/BIGH3 Monoclonal Antibody (3E11D11))”	Proteintech	RRID:AB_10896828; Cat#60007–1-IG; 3E11D11
anti-ENO1 (“Anti-ENO1 + ENO2 + ENO3 antibody [EPR10863(B)])	Abcam	Cat#ab155102
anti-rabbit IgG (“Invitrogen™ F(ab’)2-Goat anti-Rabbit IgG (H + L) Cross-Adsorbed Secondary Antibody, PE”)	Fisher Scientific	Cat#PI31864
anti-ALIX (“Alix (3A9) Mouse mAb #2171)	Cell Signaling	RRID:AB_2299455;Cat#2171
anti-CD63 (“Anti-CD63 antibody [EPR5702] - Late Endosome Marker”)	Abcam	RRID:AB_2800495; Cat#ab134045
anti-DDR1 (“Human DDR1 Antibody AF2396”)	R&D Systems	RRID:AB_2092092; Cat#AF2396
anti-ECM-1 (“ECM1 antibody (11521–1-AP)”)	Proteintech	RRID:AB_2261964; Cat#11521–1-AP
anti-FLOT1 (“Purified Mouse Anti-Flotillin-1”)	BD Biosciences	RRID:AB_398139; Cat#610820
anti-GM130 (“BD Transduction Laboratories™ Purified Mouse Anti-GM130”)	BD Bisosciences	RRID:AB_398141; Cat#610822
anti-TGFBI (“Anti-TGFBI antibody [EPR12078(B)] (ab170874)”)	Abcam	RRID:AB_2895231; Cat#ab170874
anti-TSG101 (“Anti-TSG101 antibody (ab30871)”)	Abcam	RRID:AB_2208084; Cat#ab30871
anti-mouse IgG HRP (“Anti-Mouse IgG (H + L), HRP Conjugate”)	Promega	RRID:AB_430834; Cat#W4021
anti-rabbit IgG HRP (“Anti-Rabbit IgG (H + L), HRP Conjugate”)	Promega	RRID:AB_430833; Cat#W4011
Donkey anti-goat IgG (“IRDye® 680RD Donkey anti-Goat IgG (H + L)”)	LI-COR	RRID:AB_10956736; Cat#926–68074
Bacterial and virus strains
Human TGFBi-NG-His Lentivural construct	VectorBuilder	Vector ID: VB220509–1203dhk Codon optomized for human Available from Robert Coffey
BL21(DE3) competent E. coli cells	New England Biolabs Inc	C2527H
Chemicals, peptides, and recombinant proteins
Cholesterol	Pointe Scientific	Cat#7509, 7510
Pointe Scientific's Triglycerides Liquid Reagents	Pointe Scientific	Cat#23-666-412
Critical commercial assays
Phospholipids C kit	FujiFilm	Cat#NC9993780
Pierce™ BCA Protein Assay Kit	Thermo Scientific	Cat#23225
SYTO ® RNASelect™	Thermo Scientific	Cat#S32703
RNeasy Mini Kit	Qiagen	Cat#74104
NEXTFLEX® Small RNA-Seq Kit v3	Bioo Scientific Corp	Cat#NOVA-5132–05
Nucleoside Digestion Mix	New Enlgand Biolabs	Cat#M0649S
Deposited data
sRNAseq	This paper	GEO: GSE304273
ProteomXchange	This paper	PXD066872
MetaboLights	This paper	MTBLS12859
Experimental models: Cell lines
Human: DiFi	From Bruce Boman and Robert Coffey	From refs (10) and (22) rectal tumor cell line from an FAP patient
Human: CC-CR	From Robert Coffey	Derived from the colon tumor cell line HCA7 in (23)
Human: DiFi overexpressing TGFBi neon green	This paper	N\A
Human: CC-CR overexpressing TGFBi neon green	This paper	N\A
Recombinant DNA
Plasmid: TGFBi-neon green TEV protease-His tagged construct	VectorBuilder	Vector ID: VB220509–1203dhk Codon optomized for human Available from Robert Coffey
Plasmid: AlkB wild-type protein	Addgene	pET30a-AlkB ΔN11
Plasmid: D135S mutant protein	Addgene	pET30a-AlkB-D135S
Plasmid: D135T mutant protein	Addgene	pET28a(+)-AlkB-D135T
Software and algorithms
Scaffold version 5.1.0	Proteome Software	https://www.proteomesoftware.com/products/scaffold-5
TIGER	Allen et al.81	https://doi.org/10.1080/20013078.2018.1506198
MS-DIAL	Tsugawa et al.82	https://doi.org/10.1038/s41587-020-0531-2
AB SCIEX Analyst Software version 1.6.2	Applied Biosystems	https://sciex.com/products/software/analyst-software
R version 4.4.1	R core team	https://www.r-project.org/
“limma” R package v. 3.59.1	Bioconductor	https://bioinf.wehi.edu.au/limma/
Cutadapt version 4.5	National Bioinformatics Infrastructure Sweden	https://doi.org/10.14806/ej.17.1.200
FastQC version 0.12.1	Babraham Bioinformatics	https://doi.org/10.12688/f1000research.15931.2
Bowtie version v1.3.1	John Hopkins University	https://doi.org/10.1186/gb-2009-10-3-r25
Superomics	This paper	https://doi.org/10.5281/zenodo.16649130
DESeq2 version 1.42.1	Bioconductor	https://bioconductor.org/packages/devel/bioc/html/DESeq2.html

STAR★METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Cells

DiFi cells were generated from a rectal carcinoma from a familial adenomatous polyposis patient with a germline mutation in APC.^10,83 CC-CR cells are clones derived from HCA-7, a human microsatellite instability-high (MSI-H) CRC cell line.⁸⁴

Cell culture

2D: DiFi and CC-CR cells were cultured in DMEM medium supplemented with 10% bovine growth serum, 1% glutamine, 1% non-essential amino acids, and 1% penicillin–streptomycin at 37 °C in a 5% CO₂ humidified incubator. All cell culture media was purchased from Corning Cellgro and all cell culture supplements were from Hyclone, unless stated otherwise. Cells were cultured until 80% confluent, then washed 3X with PBS and cultured in serum-free medium for 48 h prior to collection of EVs and NVEPs.

A total of three separate 2D pools of DiFi medium (1,360 mL) were harvested from 68 plates with 20 mL of medium isolated when culturing approximately 5.1 × 10⁹ cells. Medium was concentrated and used to isolate UC-purified EV pellets (EVp), exomeres, and supermeres, as described.^5,16

3D: Hollow fiber bioreactors were loaded with 1–5 × 10⁸ DiFi or CC-CR cells, allowed to reach an equilibrium growth phase based on glucose utilization (using 1–2 mg of glucose per day), and medium was harvested and processed as described,^11,12 except for the following modifications. With each harvest, 20 mL of conditioned medium was centrifuged sequentially at 250g and 2,500g for 10 min each (Thermoscientfic Xpro series centrifuge) to remove cellular debris. Medium was then gravity filtered through a Millex 0.22-μm pore syringe filter (Millipore Sigma, Burlington, MA).

Four separate pools of approximately 400 mL of conditioned medium were generated and distributed concurrently. Each pool was divided into 60 mL aliquots for purification by UC, FPLC, or cushion-DGUC (C-DGUC), with the remainder kept in reserve for quality control (QC) or other experiments.

METHOD DETAILS

UC-based isolation of EVs and NVEPs

UC-based EV and NVEP isolation was performed as previously described.^5,16 Specifically, 2D serum-free conditioned medium was centrifuged for 15 min at 1,000g to remove cellular debris and the resulting supernatant was then filtered through a 0.22 μm polyethersulfone filter (Nalgene). The filtrate from 2D conditioned media was concentrated using a centrifugal concentrator with a 100,000 Dalton molecular weight cutoff (Millipore). The concentrate or 3D conditioned media was then subjected to UC at 167,000g for 4 h in an SW32 Ti swinging-bucket rotor (Beckman Coulter) and the resulting sEV pellet was resuspended in PBS containing 25 mM HEPES (pH 7.2) and washed by centrifuging again at 167,000g for 4 h. The washed pellet was designated as EVp.

To isolate exomeres, the supernatant collected from the 4 h UC was ultracentrifuged at 167,000g for 16 h. The resulting pellet was resuspended in PBS containing 25 mM HEPES (pH 7.2) and washed by centrifuging again at 167,000g for 16 h. The washed pellet was designated as exomeres.

To isolate supermeres, the supernatant from the pelleting of exomeres was subjected to UC at 367,000g using a Beckman Coulter SW55 Ti rotor (k factor of 48, Beckman Coulter) for 16 h. The resulting pellet was resuspended in PBS containing 25 mM HEPES (pH 7.2) and was designated supermeres.

Density gradient ultracentrifugation

The isolation of sEVs, heavy EVs, and non-vesicular fractions using C-DGUC has been previously described.²⁴ In brief, the 60 mL 3D-sourced aliquots were spun at 10,000g for 30 min in a Ti 45 (Beckman Coulter) at 4°C. The supernatant from 10,000g spin was concentrated with Centricon Plus-70 centrifugal filter (cat no. UFC710008, Millipore) at 4°C to 30 mL. The concentrate was overlaid onto a 2 mL cushion of 60% iodixanol (OptiPrep) and centrifuged at 100,000g for 4 h in an SW32 rotor (Beckman Coulter) at 4°C. For the preparation of the density gradient, the 40% iodixanol layer was made by transferring the bottom 3 mL of the cushion (which includes 1 mL of collected EVs and 2 mL of 60% iodixanol) into a new tube. The 20%, 10%, and 5% iodixanol dilutions were prepared by diluting OptiPrep with 0.25 M sucrose/10 mM Tris, pH 7.5. Three mL of each iodixanol dilution were subsequently layered on top of the 40% layer. The density gradients were spun at 100,000g for 18 h in an SW40 rotor (Beckman Coulter) at 4°C. After this centrifugation step, 1 mL density gradient fractions were collected. Fractions 6, 7, 9. 10, and 12 were diluted in 2.5 mL PBS and centrifuged at 100,000g for 3h in a TLA-110 rotor at 4°C. To quantitate the size and concentration of these fractions, nanoparticle tracking analysis was performed using a Particle Metrix ZetaView PMX 110. The final pellets were resuspended in 100 μL of PBS for C-DGUC. To quantitate the size and concentration of these fractions, nanoparticle tracking analysis (NTA) was performed using a Particle Metrix ZetaView PMX 110.

Size-exclusion chromatography

Size-exclusion chromatography (SEC) was performed on an AKTA Purifier fast protein liquid chromatography (FPLC) system (GE) with two inline Superose 6 Increase resin columns (Cytiva). Samples were filtered offline using 0.22 μm spin-filters and 0.5 mL of filtered sample was injected into the FPLC. Samples were run at a flow rate of 0.3 mL per min and fractionated into 1.5 mL wells of 96-deep well plates. Colorimetric assays for phosphatidylcholine (PC) (FujiFilm, NC9993780), total cholesterol (Pointe Scientific, 7510, 7509), total protein by BCA (Thermo Scientific, cat. 23225), and triglycerides (Pointe Scientific, 23-666-412) were performed, per the manufacturer’s instructions. Total RNAs were quantified in each fraction by SYTO RNASelect (Thermo Scientific, S32703) and a sRNA standard curve, as described previously.⁸⁵ The supermere samples were a result of combining fractions 19–24, and exomeres samples were pooled from fractions 14–17.

Purification of TGFBi-associated supermeres

The TGFBi-neon green-TEV protease site-His tagged (TGFBi-NG-His) construct was generated via VectorBuilder (Figure S2B) and infected via lentivirus into DiFi or CC-CR cells, as described previously.¹³ In brief, human codon optimized TGFBI (nm_000358.3) was synthesized by VectorBuilder (Chicago, Il) with a 3x GGGS linker C-terminal neon green fluorescent tag followed by 3x-Flag-tag, TEV protease site, HIS-tag, MYC-tag cassette in a pLV[Exp]-Puro-CB retroviral vector. Virus produced from this construct from VectorBuilder was used to transduce this construct into DiFi and CC-CR colorectal cancer cell lines. Puromycin-selected cells were also flow sorted for neon green fluorescence to isolate high expressing cells. Transfected cells were grown in a hollow fiber bioreactor as described above. Bioreactor-produced medium from DiFi and CC-CR cells expressing TGFBi-NG-His was processed for isolating supermeres by FPLC SEC as described above. Fractions containing SEC-purified supermeres were concentrated utilizing 10 kDa MWC Ultra Centrifugal Filters (Amicon) at 2,500g in a tabletop swinging bucket centrifuge (Beckman Coulter) at 4° C. Concentrated pooled fractions were mixed with 1 mL fresh HisPur Ni-NTA resin (Thermo Scientific) that was prepared in PBS in accordance with manufacturer’s guidelines. The slurry was incubated overnight at 4 °C with end-over-end rotation using a tabletop rocker. Resin and immobilized TGFBi-NG-His supermeres were pelleted via centrifugation at 200 rpm for 1 min and the supernatant was decanted. The affinity resin was then washed with 10 mL of PBS for 10 min using a tabletop rotator and the resulting supernatant was decanted. TGFBi-NG-His supermeres were eluted from the affinity resin with 5 mL of buffer (20 mM Tris-HCL, 500 mM NaCl, 1M imidazole, pH 7.9). Affinity-purified TGFBi-NG-His supermeres were then sequentially dialyzed three times with 4 h exchanges of 4 L PBS at 4 °C. Next, the His affinity tag was removed from the TGFBi-NG-His supermeres by cleavage with the TEV protease separating the mNeonGreen fusion protein and the His-tag. To do so, dithiothreitol (DTT) was added to dialyzed TGFBi-NG-His supermere samples to a final concentration of 3 mM and TEV protease was added to supermere samples at a total protein concentration ratio of 10:1 (supermere/enzyme; w/w), and incubated with gentle rocking at room temperature for 3 h, followed by an overnight incubation at 4°C. Samples were dialyzed against PBS, as above, to remove DTT from the samples prior to subsequent Ni-NTA removal of the His tag. Dialyzed samples were applied to freshly equilibrated Ni-NTA resin and the unbound fraction containing TGFBi-NG-His-tagged supermeres was collected for downstream applications.

FAVS analysis

For Fluorescence-activated vesicle sorting (FAVS), the samples were prepared and analyzed as described previously.^5,86 Briefly, UC-purified intravenous immunoglobulin (IVIG)-blocked supermeres from 3D DiFi medium were single or double-stained for TGFBi and/or ENO1 by first staining for TGFBi with CoraLite-488-conjugated anti-TGFBi (diluted 1:500 mouse MoAB clone: 3E11D11, Proteintech) for 1 h at 4 °C. One portion of TGFBi-labeled and one portion of unlabeled supermeres were stained with unconjugated anti-ENO1 (diluted 1:500 rabbit recombinant MoAB clone: EPR10863 (B), Abcam) incubated for 1 h at 4 °C. In single-stained ENO1 and double-stained TGFBi, ENO1 samples PE-conjugated F(ab’)2-goat anti-rabbit IgG(H + L) cross-adsorbed polyclonal secondary (diluted 1:1000, PI31864, Fisher Scientific) was added and supermeres were stained overnight at 4 °C. Single-labeled TGFBi or ENO1 or dual-labeled supermeres were then washed, diluted, and analyzed as previously described.^5,86

Electron microscopy

Negatively stained grids for transmission electron microscopy (TEM) were prepared similarly as previously described²⁵. Exomere and supermere samples, as well as selected density gradient fractions were carefully resuspended and adhered to freshly glow-discharged 300 mesh carbon grids (Electron Microscopy Sciences) by floating the grids on top of 5 μL of samples for 20 s. Samples were washed by touching the grid to two 40 μL drops of ddH₂O followed by incubation with 5 μL of 2% uranyl acetate for 20 s. The negative stain was blotted by pressing the side of the TEM grid against a freshly torn piece of #1 Whatman filter paper until the grid was dry. TEM imaging was performed on a JEOL 2100 Plus operating at 200 keV using an AMT nanosprint CMOS camera. Random fields of EVs were imaged over the entire grids.

Immunoblotting

Protein quantifications of EV and NVEP samples were done using a Direct Detect system or by SDS-PAGE/SYPRO Ruby protein staining densitometry analysis, following a previously described protocol.^87,88 Then, 4X loading buffer (Bio-Rad, 1610747) or 2x SDS-PAGE sample buffer (4x SDS sample buffer: 0.2M Tris, pH 6.8, 8% SDS, 40% glycerol, 0.16M DTT and 0.29 M bromophenol blue in water) was added to samples, which were then heated at 100 °C for 5 min and separated on 9% SDS–PAGE bis–Tris gels (Life Technologies) or 10% Mini-PROTEAN TGX Precast Protein Gels (Bio-Rad, 4561033), depending on the primary antibody used, before being transferred to nitrocellulose membranes (Amersham Protan, 10600001) overnight at 4 °C. Membranes were stained with Ponceau S solution (Sigma, P7170) before blocking. The membranes were blocked at RT for 3 h in Odyssey TBS blocking buffer (LI-COR, 927–60001) or 1 h in 10% 5% non-fat dry milk TBS-T with 0.1% Tween 20, depending on the primary antibody used. The membranes were incubated with primary antibodies overnight at 4 °C. After washing with TBS-T membranes were incubated with secondary antibodies for 1 or 2 h at RT depending on the secondary antibody used and washed with TBS-T. To capture results, the immunoblots were developed using chemiluminescence (Western Lightning Plus-ECL, PerkinElmer) for an Odyssey XF Imaging System, or incubated with either ECL reagent (Thermo Fisher Scientific, 32106) or 25% Femto ECL (Thermo Fisher Scientific, 34095) for 2 min before being scanned using an iBright Imager (Thermo Fisher Scientific). The Protein simple Jess system (Bio-techne, Minneapolis, MN) was also used according to manufacturer’s instructions for immunoblotting.

The primary antibodies against the following proteins were used: ALIX (Cell Signaling, 2171), CD63 (Abcam, ab134045), DDR1 (R&D Systems, AF2396), ECM-1 (Proteintech, 11521–1-AP), FLOT1 (BD BioSciences, 610820), GM130 (BD Biosciences, 610822), Hsp70 (Santa Cruz, sc-66048), TGFBi (Abcam, ab170874), TSG101 (Abcam, ab30871). The following secondary antibodies were used: HRP-conjugated secondary antibodies (Promega, anti-mouse IgG HRP cat W4021 and anti-rabbit IgG HRP, W4011), Odyssey anti-goat IRDye 680RD antibody (LI-COR, 926–68074). Primary antibodies were diluted 1:5000 for all blots except for FPLC-SEC fraction (1:1000) and Jess system (1:20) blots.

Mass spectrometry proteomics

Samples were brought to a final concentration of 5% SDS, reduced with 10mM TCEP (tris(2-Carboxyethyl)phosphine) at 55 °C for 15 min, and alkylated with iodoacetamide (20 mM) for 30 min in the dark at room temperature. Protein samples were then prepared by S-Trap (ProtiFi) digestion. Aqueous phosphoric acid was added to a final concentration of 2.5%, followed by addition of 90% methanol containing 100 mM TEAB (Tetraethylammonium borohydride) at 6X the volume. The samples were loaded onto S-Trap micro columns and centrifuged at 4,000g. The columns were washed 4X with 150 μL 90% methanol containing 100 mM TEAB. Proteins were digested with trypsin (Promega) at a 1:10 enzyme to protein ratio in 50 mM TEAB, pH 8.0, for 1 h at 47 °C. Peptides were eluted by sequential addition of 40 μL each of 50 mM TEAB, 0.2% formic acid, and 0.2% formic acid in 50% acetonitrile. Eluted peptides were dried in a speed-vac concentrator and resuspended in aqueous 0.2% formic acid for analysis by LC-coupled tandem mass spectrometry (LC-MS/MS). A reverse phase capillary analytical column (100 μm I.D.) was packed with 20 cm of C18 material (Jupiter, 3 μm beads, 300Å; Phenomenex) directly into a laser-pulled emitter tip. Peptides were loaded on the column using a Dionex Ultimate 3000 nanoLC and autosampler. The mobile phase solvents consisted of 0.1% formic acid, 99.9% water (solvent A) and 0.1% formic acid, 99.9% acetonitrile (solvent B). Peptides were gradient-eluted at a flow rate of 350 nL/min, using a 90-min gradient. The gradient consisted of the following: 0–72 min, 2–38% B; 72–78 min, 38–90% B; 78–80 min, 90% B; 80–81 min, 90–2% B; 81–90 min (column re-equilibration), 2% B. Peptides were analyzed using a data-dependent method on an Orbitrap Exploris 240 mass spectrometer (Thermo Scientific), equipped with a nanoelectrospray ionization source. The instrument method consisted of MS1 (AGC target value of 3 × 10⁶), followed by up to 20 MS/MS scans (AGC target of 1×10⁵) of the most abundant ions per MS scan. The intensity threshold for MS/MS was set to 1 × 10⁴, HCD collision energy was 30 nce, and dynamic exclusion (15 s) was enabled. Proteomics samples generated for TGFBi-containing supermere comparisons were analyzed on a Q Exactive Plus mass spectrometer using a top 20 data-dependent method. The MS/MS AGC target was set to 5 × 10⁴ and HCD collision energy was set to 27 nce. Tandem mass spectra were searched with Sequest (Thermo Fisher Scientific) against a Homo sapiens database created from the UniprotKB protein database (www.uniprot.org). Variable modifications included +15.9949 on Met (oxidation) and +57.0214 on Cys (carbamidomethylation). Search results were assembled in Scaffold 5.1.0 (Proteome Software) using a minimum filtering criteria of 95% peptide probability and 99% protein probability. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE⁸⁹ partner repository with the dataset identifier PXD066872.

ARM-seq

Mutant AlkB enzyme mix was generated through purification of AlkB wild-type protein (pET30a-AlkB ΔN11) (Addgene), D135S mutant protein (pET30a-AlkB-D135S), and D135T mutant protein (pET28a(+)-AlkB-D135T) (Addgene), which were expressed in BL21(DE3) competent E. coli cells (New England Biolabs Inc). Protein expression was induced by 1 mM isopropyl β-d-1-thiogalac-topyranoside with 5 μM iron sulfate for 4 h at 30 °C. Cells were centrifuged and resuspended in Buffer A (10 mM Tris-HCL pH 7.4, 300 mM NaCl, 2 mM CaCl2, 10 mM MgCl2, 2 mM 2-mercaptoethanol, 5% glycerol) containing protease inhibitors. Samples were lysed by sonication. Supernatants were loaded onto a Ni-NTA column (Qiagen) pre-equilibrated with Buffer A and eluted using 250 μM of imidazole. Fractions containing purified AlkB proteins were pooled and dialyzed against storage buffer (30 mM Tris pH 8, 2 mM 2-mercaptoethanol), then supplemented with 30% glycerol and snap frozen. RNA demethylation was performed in a 50 μL reaction volume containing 1 μM purified AlkB (wt, D135S, and D135T) enzymes, 25 mM MES buffer pH 6.0, 2 mM MgCl2, 270 mM KCl, 2 mM L-ascorbic acid, 300 μM α-ketoglutarate, 283 μM (NH4)2Fe(SO4)2, and 1 U/mL SUPERaseIn RNase Inhibitor. Reactions were carried out at 25 °C for 2 h then stopped by the addition of EDTA. RNA was re-purified using RNeasy Mini Kit (Qiagen, 74104) and used as input for sRNA sequencing library preparation. Libraries were prepared using NEXTFlex Small RNA-seq kit v3 (PerkinElmer) and size-selected using Pippin Prep (Sage). Libraries were quantified with Qubit (Thermo Fisher) and assessed using Bioanalyzer high-sensitivity DNA chips (Agilent). Paired-end sequencing (PE-150) using equimolar concentrations of multiplexed libraries were completed using NovaSeq6000 (S4, Illumina) by the Vanderbilt Technologies for Advanced Genomics Core. Data were analyzed using the TIGER platform.⁹⁰ The data have been deposited in NCBI’s Gene Expression Omnibus⁹¹ and are accessible through GEO Series accession number GSE304273 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE304273).

Mass spectrometry lipidomics

Supermeres, exomeres, and sEVs were subjected to liquid chromatography-tandem mass spectrometry (LC-MS/MS) approaches in collaboration with the Mass Spectrometry Shared Resource at Vanderbilt University Medical Center. Lipids were initially extracted by Bligh-Dyer⁹² lipid isolation methods and injected into a Vanquish Horizon UPLC system (Thermo) system prior to mass spectrometry analyses using a QExactive hybrid quadrupole/Orbitrap high-resolution mass spectrometer (Thermo). MS Dial(1, 2) software was used for downstream low-level analyses. Equivolume pooling of all samples was used as an internal quality control (QC) and sampled at the start, middle, and end of runs to assess drift. Solvent negative controls, aqueous internal standards (AIS; Equisplash lipidomix, Avanti), and blanks were also applied. The mass spectrometry lipidomics data have been deposited to the MetaboLights⁹³ database with the dataset identifier MTBLS12859.

Mass spectrometry epitranscriptomics

Total RNA was extracted from equal quantities of sEVs using the RNeasy Mini Kit (Qiagen), following the manufacturer’s protocol. RNA samples and ribonucleoside standards were digested into single nucleosides enzymatically using the Nucleoside Digestion Mix (BioLabs). Before analysis, RNA nucleosides were spiked with 1 nM of an internal standard, [15N]-dA. The samples were then analyzed using LC-MS/MS on a Shimadzu Nexera system paired with a QTRAP 6500 (Applied Biosystems) with an electrospray ionization source. All separations were performed on a Hypersil GOLD aQ C18 column (100 mm × 2.1 mm, 175 Å pore size, 1.9 μm particle size) (ThermoFisher) with a flow rate of 0.4 mL/min using 0.1% formic acid (solvent A) and acetonitrile with 0.1% formic acid (solvent B). The gradient for each sample was: 0–6 min, 0% B; 6–7.5 min, 1% B; 7.5–9.5 min, 6% B; 9.5–10.5 min, 10.5–12 min, 50% B; 12–14 min, 75% B; 14–16 min, 75% B. Data acquisition and processing were conducted using AB SCIEX Analyst Software 1.6.2 (Applied Biosystems). The modification levels were compared across samples by calculating the percentage ratio of modified nucleosides to total nucleosides. The retention times (rt), m/z values for precursor (Q1) and product ions (Q3), collision energy (CE), declustering potential (DP), and dwell time (dt) for standards are detailed in Table S7.

QUANTIFICATION AND STATISTICAL ANALYSIS

Bioinformatics analysis of the omics results

Proteomics

Proteins with an average spectral count of ≥1 in each group were considered detectable. Spectral counts of proteins were normalized to the total spectral counts and log₂-transformed. Principal component analysis was performed on the log-transformed normalized counts using the prcomp function in R. Differential analysis between groups was performed using limma (v 3.59.1).⁹⁴ The Benjamini-Hochberg correction was used to adjust the p-value for multiple testing. Proteins with absolute fold change ≥ 2 and an FDR ≤ 0.05 were considered significantly different. For heatmaps, proteins with the highest average abundance per group were selected, and the log-transformed normalized counts were visualized using the pheatmap R package. Gene Ontology term “GO:0003723” was used to determine RNA-binding proteins.

Arm-seq

Bioinformatics processing of sRNA sequencing data was performed with the TIGER (“Tools for Integrative Genome analysis of Extracellular sRNAs”) analysis pipeline.⁹⁰ Briefly, Cutadapt (v4.5)⁹⁵ was used to trim 3′ adapters of raw reads. All adapter-trimmed reads with less than 16 nucleotides were discarded. Quality control on both raw reads and trimmed reads was performed using FastQC (v0.12.1) (www.bioinformatics.babraham.ac.uk/projects/fastqc). The trimmed reads were mapped to the GENCODE GRCh38.p13 genome, in addition to rRNA and tRNA reference sequences, by Bowtie1 (v1.3.1),⁹⁶ allowing only one mismatch. Significantly differential expressed small RNAs with absolute fold change ≥ 1.5 and p-value ≤ 0.05 were detected by DESeq2 (v1.42.1),⁹⁷ using total mappable reads as the normalization factor. After mapping, those remaining unmapped reads were then aligned in parallel to exogenous structural RNA databases and curated microbial genome databases allowing no mismatches. Reads that failed to align to any strategy were categorized as ‘unknown’. Principal component analysis (PCA) was performed across all samples to capture the largest sources of variance. As a result, differences between the two specific groups of interest may not be well represented in the PCA plot, as they account for a smaller portion of the overall variance, as in Figure 3C. However, when PCA is performed using only the two groups of interest, their differences become more pronounced.

Lipidomics

Lipid species with a) blank/QC values > 10%, b) relative standard deviation (RSD) > 20% for QC samples, or c) nomenclature issues (unknowns, w/0 ms2, and RIKEN) were removed. Data were normalized using the LOWESS method within MSDial. Positive and negative ion mode data were combined. In instances where lipids were identified in positive and negative ion mode, the lipid with the highest value was kept. Lipids were annotated according to the lipid class nomenclature provided by MSDial.⁴⁰ Differential analysis between groups was performed using Limma (v 3.59.1).⁹⁴ The Benjamini-Hochberg correction was used to adjust p-values for multiple testing. Lipids with absolute fold change ≥ 1.5 and FDR ≤ 0.05 were considered significantly different.

Statistics

Student’s T-tests (two-way) were used to compare between two groups. One-way analysis of variance (ANOVA) was used to compare between >2 groups. P-values <0.05 were considered significant: *p < 0.05; **p < 0.01; and ***p < 0.001.

Additional resources

Web application

The web application was developed in RStudio using the ‘Shiny’ package and, along with the data generated from proteomics, lipidomics, and sRNA-seq analyses used for plotting, is accessible online as a webpage (https://superomics.shinyapps.io/browse/) and via a GitHub repository.⁹⁸

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.celrep.2025.116287.

Highlights.

• 3D cultures allow for up to 20-fold increased yields of exomeres and supermeres versus 2D

• FPLC-SEC is a time-efficient, gentle alternative for exomere and supermere isolation

• Protein composition mainly depends on carrier type, not culturing or isolation method

• RNA/lipid profiles of EVs and NVEPs are influenced by culturing and isolation methods