CD151 Maintains Endolysosomal Protein Quality to Inhibit Vascular Inflammation

Data availability.

The data that support the findings of this study are available from the corresponding author upon reasonable request. Detailed descriptions for all methods are provided in the Supplemental Material.

Cd151 ablation in endothelia causes lung inflammation and potentiates inflammatory responses in various diseases.

To reveal the relationship of CD151 with diseases, we performed RNAseq with primary mouse lung endothelial cells (MLECs) isolated from Cd151^+/+ and Cd151^−/− mice²². The Overrepresented Diseases analysis on the RNAseq data showed that Cd151 ablation-mediated changes in gene expression associated CD151 to inflammation and inflammatory diseases (Figure S1a). In the lungs of human patients with chronic obstructive pulmonary diseases (COPD) or Covid-19, endothelial CD151 proteins were significantly decreased, compared to normal lungs (Figure S1b), and CD151 mRNA in entire lung tissue of Covid-19 patients also displayed a trend of reduction (Figure S1c). Hence, changes in endothelial CD151 expression are related to human inflammatory diseases. Moreover, inflammatory stimuli LPS, TNF-α, and IFN-γ reduced the EC surface and/or total cellular levels of CD151 (Figure S1d-e).

To assess CD151 role in vascular inflammation in vivo, we generated endothelial conditional Cd151 knockout mice (Cd151^iΔEC) (Figure S2a), given the abundance of CD151 in endothelia. With specific removal of CD151 from endothelia (Figure S2b-c), Cd151^iΔEC mice displayed disseminated pulmonary lesions at the microscopic scale, as shown by H&E staining (Figure 1a). The pathological changes appear to be progressive, given no obvious pulmonary abnormality in Cd151^iΔEC mice at 2- and 8-weeks post-Tamoxifen administration (Figure S3a). Miles assay reported much more pronounced leakage of lung vasculature in Cd151^iΔEC mice, compared to control (Ctrl.) mice, at 16 weeks, but not 2 weeks, after tamoxifen induction (Figure 1b), suggesting that Cd151 ablation doesn't cause lung vascular leakage immediately. Staining of murine IgG deposition in perfused tissues, an established assay evaluating vascular leakage in vivo^{23, 24}, further confirmed pulmonary vascular leakage in Cd151^iΔEC mice at 16 weeks after tamoxifen induction (Figure S3b). We also examined IgA, IgE, C3, and C1q and found no obvious accumulation of these proteins in the lungs of Cd151^iΔEC mice (Figures 1b and S3b-d), excluding the pulmonary inflammation phenotype as the result of Ig deposition and complement activation from autoimmune response.

To follow endothelial CD151 in vascular inflammation development, we included mTmG lineage tracing model for the phenotype analysis (Figure S4a)²⁵. Cd151 ablation damaged endothelium integrity, i.e., discontinuities of endothelia, as denoted by mEGFP in Cd151^iΔEC:mTmG and endomucin in Cd151^iΔEC mouse lungs, compared respectively with those in Ctrl.:mTmG and Ctrl. mouse lungs (Figures 1c-d and S4b, arrows). In addition, Cd151 ablation in endothelia caused evidently more intra- and peri-vascular cellularity (Figure 1d, arrowheads) and more infiltration of i) non-ECs or leukocytes (CD45⁺ cells) inside and around walls of small blood vessels, ii) macrophages (F4/80⁺ cells) in lung interstitial areas (Figure 1d), and iii) neutrophils, which are reflected by myeloperoxidase (MPO) activity (Figure 1e). Pulmonary accumulation of macrophages became detectable in Cd151^iΔEC mice at 8 weeks, not 2 weeks, after tamoxifen injection (Figure S4c). But the levels of VCAM1, a molecule important for leukocyte recruitment, exhibited no difference between Cd151^+/+ and Cd151^−/− MLECs (Figure S4d).

No noticeable abnormality of vasculatures was found in other organs such as liver (Figure S4e) in Cd151^iΔEC mice at the basal level. Overall, these observations underline the integrated vascular inflammatory responses in lung upon endothelial Cd151 removal.

We then expanded the study to inflammatory conditions or diseases. In heart, endothelial Cd151 removal caused substantially more infiltration of inflammatory cells in myocardium, compared to control mice, in Angiotensin II (Ang II)-induced chronic inflammation (Figure 1f). During progression of the sepsis induced by bacterial endotoxin lipopolysaccharide (LPS), Cd151^iΔEC mice displayed a significantly lower survival rate than control mice (Figures S4g and 1g). For LPS-induced acute respiratory distress syndrome (ARDS), Cd151^iΔEC mice exhibited substantially more severe lung injury than control mice (Figure 1h). Moreover, we examined CD151 role in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-induced lung injury, by challenging Cd151^+/+;hACE2-tg and Cd151^−/−;hACE2-tg mice with SARS-CoV-2. Cd151 ablation reduced animal survival and elevated weight loss only at the end (Figure 1i). More lung injury and inflammatory cell infiltration were observed in Cd151^−/−;hACE2-tg mice than in Cd151^+/+;hACE2-tg mice (Figure 1j). These results collectively indicate that endothelial Cd151 ablation exacerbates various cardiovascular inflammations.

CD151 removal elevates endothelial production of inflammatory molecules by facilitating exosomal release.

To explore how CD151 restrains inflammation, we compared RNA-seq data from Cd151^+/+ and Cd151^−/− MLECs and found that IL-1 and TNF pathways were the most affected signaling (Figure 2a). Analysis on cytokine secretomes of the MLEC culture supernatants revealed that IL-1β was unchanged, TNFα elevated, but angiopoietin-2 (ANGPT2) dramatically increased upon Cd151 ablation (Figures 2b and S5a). In blood plasma of Cd151^iΔEC mice, only ANGPT2, neither TNF-α nor IL-1β, was significantly increased, compared to those in control mice (Figures 2b and S5b). More ANGPT2 proteins were detected in the endothelium and interstitium of Cd151^iΔEC lung (Figure 2c). Since ANGPT2 is the only cytokine that was consistently altered in Cd151^−/− MLECs and Cd151^iΔEC mice, we focused on ANGPT2 in the following studies.

Interestingly, the exosomes released by Cd151^−/− MLECs carried more ANGPT2 proteins (Figure 2d). Moreover, Cd151^−/− MLECs released more but smaller extracellular vesicles (EVs) than Cd151^+/+ MLECs, and the changes in EVs from Cd151^−/− MLECs mainly resulted from exosomes (<200 nm), not from microvesicles (>200 nm) (Figures 2e and s5c). Exosomes from Cd151^−/− MLECs also displayed larger heterogeneity or irregularity in morphology (Figure 2e). CD151 knockdown in human dermal microvascular ECs (HMEC), bEnd.3 murine brain ECs, and HT1080 fibroblastoma cells reduced EV size but didn't alter EV number (Figure S5d).

Consistently, condition media of Cd151^−/− MLECs markedly disrupted the barrier function of Cd151^+/+ MLEC monolayer, and ANGPT2 largely pheno-copied effect of the condition media (Figure 2f), suggesting that endothelial barrier-disruptive soluble factor(s) from Cd151^−/− MLECs is attributed largely to ANGPT2.

As multi-vesicular bodies (MVBs) undergo exocytosis to release exosomes, we then examined MVBs and found substantial accumulation of ANGPT2 within MVBs in CD151-silenced HMECs, evidenced by elevated colocalization of ANGPT2 with lysobisphosphatidic acid (LBPA), a MVB marker (Figure 2g), indicating more ANGPT2 trafficking into MVBs.

We then applied exocytosis inhibitor Nexinhib 20 (Nex20) (5 μM)²⁶ to block exosome release. Nex20 abrogated or diminished Cd151 ablation-induced increases in i) EV production (Figure 2h), ii) exosomal release of ANGPT2 (Figure 2h), and iii) endothelial secretion of ANGPT2 (Figure 2i). In vivo, Nex20 reduced LPS-induced lung injury and abrogated the difference between control and Cd151^iΔEC mice (Figure 2j), indicating exocytosis inhibition rescues Cd151 removal-aggravated lung injury. Nex20 also diminished the difference in pulmonary infiltration of neutrophils, reflected by MPO activity, between the two groups (Figure 2k). Moreover, treatment of sepsis caused by LPS with Nex20 markedly improved survival rates of the mice and largely eliminated survival difference between control and Cd151^iΔEC groups (Figure 2l). Furthermore, ANGPT2-neutralizing antibody reduced LPS-induced lung injury in both control and Cd151^iΔEC groups and abolished the difference between the two groups (Figure 2m and S5e). Thus, Cd151 ablation upregulates exocytosis, leading to increased EV release of ANGPT2 and exacerbated inflammation.

CD151 removal facilitates exosomal release by upregulating IFITM3.

To identify molecular partners crucial for CD151-regulated vesicular trafficking, we searched the proteins that physically associated with CD151-integrin α3β1 complex using proteomics analysis. IFITM3, IFITM2, and Tspan14 were the novel candidates among high-score readouts, in addition to those earlier reported partners (Figure 3a). Because CD151 and IFITMs were both involved in endolysosomal trafficking^{11, 20, 27}, we focused on IFITMs. CD151-IFITM3 association was confirmed by co-immunoprecipitation under both 1% Brij98 and 1% Triton-X100 lysis conditions (Figure 3b) and by PLA (Figure 3c). CD151 palmitoylation-deficient mutant largely diminished physical association with IFITM3 (Figures 3d and S6a-b), indicating importance of CD151 lipidation for this association. The CD151-IFITM3 complex was further supported by their colocalization by immunofluorescence (Figure 3e).

IFITM3 protein, not IFITM1, was increased upon Cd151 ablation (Figure 3f), and their expressions were upregulated by pro-inflammatory stimulus (Figure S6c-d) in contrast to the downregulation of CD151 (Figure S1d-e). IFN-γ and LPS treatments increased CD151-IFITM3 colocalization (Figure S6e). Cd151 ablation or silencing enlarged IFITM3-containing vesicles and MVBs, enhanced IFITM3-LBPA co-localization, but unaltered MVB level (Figures 3g, S2e, and S6f). Super-resolution imaging analysis showed that IFITM3 and LBPA were distributed at both peripheral and intraluminal areas of MVBs in Mock group, but IFITM3 became accumulated in the peripheral area of MVBs in CD151-silenced group (Figures 3h and S6g), suggesting that CD151 supports sorting of IFITM3 from limiting membrane into intraluminal vesicle (ILV) or restricts reverse sorting of IFITM3 or ILV-limiting membrane fusion. CD151 removal also enhanced lysosomal presence of IFITM3 and enlarged lysosomes (Figure S6h). Hence, CD151 confines sizes of endolysosomes and prevents IFITM3 from accumulation in endolysosomes, especially in MVB limiting membrane.

Exosomal discharge of IFITM3 was increased in bEnd.3, MLEC, and HT1080 cells upon CD151 knockdown, so was cathepsin-D, a lysosomal protease (Figures 3i and S6i). In Cd151^iΔEC lung vasculature, especially vessels with intermediate size, cathepsin-D deposition was significantly increased in medium, but not within endothelium (Figure 3j), implying that endothelial ablation of Cd151 creates inflammation-prone microenvironment in vessel wall.

Despite lowering CD151 (Figure S7a), IFITM3 knockdown abolished or diminished CD151 removal-induced i) increase in exosomal release of ANGPT2 (Figure 3k), ii) upsurge of EV release (Figures 3l and S7b), and iii) decrease in EV size (Figure 3l), highlighting that CD151 restrains secretion of EVs and ANGPT2 by confining IFITM3 level. In vivo inhibition of IFITM3 with its down-regulator cyclosporin-H (CsH)²⁸ prolonged the survival of mice with LPS-caused sepsis and abrogated the difference in survival rate between Cd151^iΔEC and Ctrl. mice (Figures 3m and S7c-d), suggesting that IFITM3 potentiates systemic inflammation and counteracts CD151-mediated restraint of inflammation. Moreover, AAV2^QuadYF-mediated in vivo delivery of IFITM3 shRNA to endothelia markedly reduced LPS-caused inflammatory injury and macrophage infiltration of lung and abolished or diminished the differences between the two groups (Figures 3n-o and S7e-g). Therefore, CD151 interacts with IFITM3 physically and confines its level, to restrain MVB size and exosome release for less inflammation.

To affirm an essential role of IFITM3 in CD151 function in vivo, we crossed Ifitm3^−/− mice with Cd151^ΔEC mice but failed in generating Cd151^ΔEC;Ifitm3^−/− mice after examining 112 F1 pups. This failure is likely due to close adjacency of these two genes in the same chromosome (Cd151: Chr 7:141.05–141.05; Ifitm3: Chr 7:140.59–140.59). Such gene juxtaposition may reflect their functional linkage.

CD151 removal upregulates IFITM3 and its ubiquitination by dissociating VCP/p97 from endolysosomes.

Valosin-containing protein (VCP)/p97 is a member of AAA-ATPase family and directly regulates lysosomal trafficking and turnover of IFITM3²⁹. We determined VCP role in CD151-IFITM3 interactions. Notably, CD151 was colocalized with VCP, and this colocalization was largely disrupted by inflammatory stimulus LPS (Figures 4a and S8a). Neither CD151 nor IFITM3 knockdown changed VCP protein level; but CD151 formed complex with VCP, in an IFITM3-independent manner as IFITM3 knockdown didn't alter this association (Figure 4b). Atomistic molecular dynamics analysis on CD151 and VCP supports their direct association, which is stabilized, in part, by a positively charged patch formed by the cytoplasmic tails and loop of CD151 and a negatively charged patch adjacent to the N-terminal binding cleft of VCP (Figure 4c). CD151 palmitoylation, which occurs at multiple cysteine residues near the interface between inner leaflet and cytosol, is required for CD151-VCP association (Figures 3d and S6b). CD151 association with VCP was independent of p47 and Npl4, major cofactors of VCP³⁰, but appeared to accommodate Ufd1, another major cofactor of VCP (Figure 4d). CD151 was not required for VCP associations with p47, Npl4, and Ufd1 but confined Ufd1 level (Figure S8b). Moreover, reciprocal co-immunoprecipitation and PLA revealed the physical association between IFITM3 and VCP (Figure 4e-f). As expected, VCP and IFITM3 were colocalized, while CD151 silencing decreased this co-localization (Figures 4g and S8c) despite increased IFITM3 levels upon CD151 removal (Figure S8d-e). Also, IFITM3-VCP association was markedly reduced upon CD151 removal, as shown by both co-immunoprecipitation and PLA (Figures 4h-i and S8f), underlining CD151 intermediates the IFITM3-VCP association.

As VCP extracts and unfolds ubiquitinated IFITM3 for de-ubiquitination and subsequent degradation²⁹, we examined IFITM3 ubiquitination and found CD151 removal increased IFITM3 ubiquitination (Figure 4j), suggesting that CD151 anchors VCP onto MVB/lysosome and bridges VCP to IFITM3 for the IFITM3 de-ubiquitination (Figure 4k). Indeed, coalescence of VCP to MVBs, based on the VCP-LBPA co-localization, became diminished upon CD151 removal or VCP inhibitor ML240 treatment (Figure 4l); while CD151 localization to MVB was not altered by ML240 (Figure S8g), as anticipated. Similar to CD151 removal, ML240 treatment upregulated IFITM3 level (Figure 4m), supporting the notion that VCP promotes IFITM3 turnover. ML240 also enhanced IFITM3-CD151 colocalization (Figure S8h), underpinning the connection of CD151 to the IFITM3 buildup resulting from VCP inhibition. Total ubiquitination level of cellular proteins was also increased upon CD151 silencing (Figure S8i), suggesting general importance of CD151 in mediating protein deubiquitylation.

ML240 elevated EV and exosome productions in Mock and/or CD151 KD cells, and diminished or abrogated their difference in EV or exosome production, respectively (Figure 4n). In addition, ML240 enhanced IFITM3 expression and enlarged IFITM3-containing vesicles in Mock and CD151KD cells to equivalent levels (Figure 4o). In conclusion, CD151 bridges VCP-IFITM3 association by docking VCP to MVB/lysosome and prevents IFITM3 from accumulation in MVB/lysosome by maintaining VCP-mediated deubiquitination and then proteolytic degradation.

CD151 removal impairs endolysosome structure and function via VCP/p97 and IFITM3.

Both proteomics and genomics analyses underscore the relevance of lysosome and EV to CD151 functions (Figures 5a and S9). Indeed, CD151 removal increased i) number of ILVs (Figure 5b), i.e., exosome precursors, and ii) size of MVBs (Figures 5c-e, 3g, 4l, S6f, and S10a), despite no impact on MVB quantity (Figure 3g). IFITM3 silencing reversed the MVB enlargement induced by CD151 knockdown, and IFITM3 inhibition with CsH (2.5 μM²⁸) (Figure S7d) partially reversed the enlargement (Figure 5d-e), indicating IFITM3 importance in the MVB aberrancy resulting from CD151 silencing. In contrast, VCP inhibitor ML240 enlarged MVBs in both Mock and CD151 KD cells but diminished the difference in MVB size between the two groups (Figure 5d). Interestingly, LPS enhanced MVB localization of CD151(Figure S10b).

For lysosomes, CD151 silencing not only enlarged their size (Figures 5e-f and S10a), which was fully corrected by IFITM3 silencing (Figure 5e), but also decreased their numbers (Figure 5g). Besides cell-cell contact distribution (Figure S10d), CD151 is also present in lysosomes (Figure S10c), its effects on lysosomes are likely to be direct. For lysosomal function, CD151 knockdown elevated endolysosome pH (Figure 5h), implying compromised function of degradation and subsequent discharge of incompletely digested contents. Besides elevating lysosomal pH (Figures 5i and S10e), inhibitions of lysosomal proton pump/V-ATPase with bafilomycin A1 and/or VCP with ML240 also eliminated the differences in pH (Figure 5i) and size (Figure 5j) between two groups, suggesting that less VCP function contributes to higher endolysosomal pH and larger endolysosomal size in CD151-silenced cells.

Compromised lysosomal function due to CD151 removal was also reflected by cellular accumulation of the proteins that undergo lysosomal degradation such as insulin receptor (INSR)³¹, amyloid precursor protein (APP)³², and sequestosome-1 (SQSTM1)³³ (Figure 5k). Using purified lysosomes, we confirmed IFITM3 was also accumulated in the lysosomes upon CD151 removal (Figure 5l), further supporting dysfunctional lysosomes in CD151-removed ECs.

CD151 removal compromises the assembly of V-ATPase complex.

Because of the change in pH, we examined V-ATPase, the proton pump determining endolysosomal pH. Transmembrane ATPV0B and cytosolic ATPV1A subunits of V-ATPase were upregulated upon CD151 KO and/or KD (Figures 6a). In addition, lysosomal coalescence of ATP6V1A was markedly decreased (Figures 6b) while ubiquitination of ATP6V1A was increased, by either CD151 silencing or ML240 treatment (Figure 6c). Physical interactions between V0 and V1 subunits and between V0 subunits were diminished upon CD151 silencing (Figures 6d-e), indicating that CD151 is required for proper assembly of V-ATPase. IFITM3 was reported to regulate lysosomal pH^{34, 35} through physical interaction with V-ATPase³⁶. Indeed, IFITM3 associated with ATP6V1A, evidenced by co-IP and PLA analyses (Figures 6f and S10f). CD151 removal also markedly reduced IFITM3 associations with the V1 and V0 subunits (Figure 6f). Although IFITM3 silencing didn’t change protein levels of V-ATPase subunits (Figure S10g), the association between V1 and V0 subunits was diminished (Figure 6g), suggesting that IFITM3 is also needed for proper V-ATPase assembly.

As VCP inhibitor ML240 enlarges endolysosome, these observations collectively underscore the necessity of VCP-CD151-IFITM3 complex for assembling V-ATPase machinery to lower endolysosomal pH (Figure 6h). Taken together, VCP-CD151-IFITM3 complex sustains structural and functional integrity of MVBs/lysosomes.

Furthermore, inflammatory stimuli reduced CD151 level, enhanced IFITM3 and IFITM1 levels, but unaltered VCP level (Figures S1d-e and S6c-d), supporting a contrasting role of CD151 and IFITM3 in inflammatory response. In addition, IFITM3 knockdown did not change CD151 protein level (Figure S10h). At the transcriptional level (Figure S11), the majority of these genes are generally positively correlated (red color) with one another across multiple experiments, both in overall tissues (Upper panel) and even more so in endothelial tissues (Lower panel). While a smaller set are generally correlated with each other but anti-correlated (green) with the larger set.

Collectively, CD151 is an inhibitor of vascular inflammation and directly supports the formations of VCP-IFITM3 and IFITM3-V-ATPase complexes, to i) confine IFITM3 level for less pro-inflammatory exosomal release and ii) support V-ATPase assembly for proper endolysosomal functions (Figures 6i and S12). CD151 constitutes protein quality control machinery with VCP by docking VCP onto endolysosomes and supporting VCP activity, to maintain and balance the lysosomal degradation and exosomal discharge functions of cells.

Crosstalk of EC adhesion with endolysosome regulates inflammation.

In contrast to the earlier observation that EC interaction with matrix protein fibronectin provokes inflammation³⁸, we found that EC interaction with matrix protein laminin, which is mediated by CD151-associated laminin-binding integrins and strengthened by CD151^{22, 39, 40}, prevents inflammation. Hence, EC-laminin interaction has inflammation-preventive or -inhibitory function, departing from traditional view on EC adhesion onto matrix as simply an anchorage mechanism for endothelia.

Beyond the established understanding that EC-matrix adhesion supports EC-EC adhesion and then endothelial barrier function to avoid inflammation, our findings also suggest that CD151-reinforced EC-matrix adhesion coordinates with MVB/lysosome to limit endothelial production of pro-inflammatory exosomes and then avoid inflammation. More importantly, our study demonstrates that the presence of CD151 in MVB/lysosome directly sets a low inflammatory status (see below) for the cells and their microenvironment, in addition to its presence at the plasma membrane for enhancing EC-matrix adhesion. Hence, departing from the canonical function of CD151 as a cell adhesion enhancer, we reveal a non-canonical function of CD151, facilitating protein quality control or protein turnover in endolysosomes (Figure S12).

CD151 coupling with VCP-IFITM3/V-ATPase complexes maintain MVB/lysosome structure and function but confine exosome production.

MVB/lysosome aberrancies in CD151-removed cells, including less acidification, size enlargement, and number diminution, indicate CD151 necessity for proper structures and functions of MVB/lysosome. Although the in-depth mechanism(s) remains to be further delineated, CD151 supports MVB/lysosome by docking VCP to limit IFITM3 level and promote V-APTase assembly. VCP unfolds ubiquitinated proteins for proteolytic degradation. Interestingly, VCP anchors to ER via derlin, a tetraspanin-like 4-transmembrane protein, for retro-translocation, ubiquitination, and ER-associated degradation (ERAD) of proteins^{41, 42}. VCP function at endolysosomes is not well-understood⁴³, while our study shows that, with the support of CD151, VCP mediates endosome/lysosome-associated degradation (ELAD) of IFITM3, V1A subunit of V-ATPase, and likely other limiting membrane proteins, to maintain i) proper trafficking and function of endolysosomes and ii) proper formation of ILVs and then exosome. Hence, similar to derlin anchorage of VCP to ER for ERAD to alleviate ER stress, CD151 anchors VCP to endolysosome for ELAD to alleviate endolysosomal stress.

MVBs fuse with either plasma membrane for exosome release or lysosome for cargo degradation. Inhibition of lysosome increases the discharge of cargos via exosomes^44-46 and the release of exosomes by exocytosis^46-49. CD151 regulates exosome quantitatively and qualitatively. Increase in exosome production in primary Cd151-null ECs may not be a cell-autonomous effect, as the change was not observed in CD151-silenced cells; but decrease in exosome size and alteration in exosome component exist in all CD151-removed cells and likely result from accumulated IFITM3 in MVB limiting membrane. In CD151-removed cells, incomplete proteolysis due to i) higher pH caused by dysfunctional V-ATPases and/or ii) aberrant sorting into ILVs caused by IFITM3 buildup can lead to more release of exosome-carrying cytokines and proteases.

IFITM3 helps acidify endolysosomes^{34, 50} and form ILVs^{13, 50, 51}. CD151 limits ILV formation and exosome production likely by confining IFITM3 ubiquitination and then its level in MVB/lysosome. Indeed, IFITM3 silencing or inhibition ameliorates in vitro and in vivo phenotypes caused by CD151 removal. Notably, IFITM3 function in MVB requires its interaction with PIP3 through its intact, cytosol-facing lysine residues⁵², while excessive ubiquitination of IFITM3, as seen in CD151-removed cells, may impede and block this interaction and thereby compromise IFITM3 function. As pathophysiological functions of IFITM3 are expanded from earlier established preventive roles in the infection of enveloped viruses to newly discovered promotive roles in Alzheimer’s disease and B-cell leukemia^{12, 18, 53}, our finding of CD151 as a crucial partner for IFITM3 prompts future studies on the link of CD151 to these diseases.

Exosomal discharge of pro-inflammatory molecules upon endothelial CD151 removal

ANGPT2 released by ECs can cause endothelial barrier dysfunction and inflammation⁵⁴. Inflammatory phenotypes of endothelial Cd151 ablation are attributed largely to ANGPT2, although the contribution of cathepsin-D remains to be determined. The phenotypes are attenuated substantially by inhibition of exosomal exocytosis, underlining that CD151 prevents vascular inflammation by restraining exosome-mediated release of cytokines and proteases. Inhibition of exosomal release also reduces LPS-induced inflammation even when CD151 is present (Figure 2j-l), so does blockade of ANGPT2 (Figure 2m), suggesting their general importance in vascular inflammation.

Exosomal presence and release of ANGPT2, based on the lines of observations from this study, are not contradictory to the tradition view of ANGPT2 presence in and release from Weibel-Palade body in endothelial cells⁵⁵. First, other residents and/or markers of Weibel-Palade body such as P-selectin and CD63 are also present in MVB and traffic between Weibel-Palade body and late endosomes such as MVB^{56, 57}. Second, Weibel-Palade body can also form and discharge ILV⁵⁸. Hence, some exosomes from ECs may originate directly from Weibel-Palade body. Moreover, exosomal releases of ANGPT2, other angiopoietins, and angiopoietin-like proteins by various cells including ECs have been reported^59-61, although those studies didn’t reveal the subcellular compartment from which the ANGPT2-carrying exosomes are derived. In CD151-removed ECs, dysfunctional lysosomes may not fully degrade internalized ANGPT2, leading to more exosomal release of ANGPT2 from MVBs for recycled or extended ANGPT2 activity. This possibility could explain the potency or prominence of ANGPT2 in inflammation models in this study.

For TNF-α, although it is not elevated in the blood of Cd151^ΔEC mice and thereby not responsible for their phenotypes, increased production of TNF-α by ECs could also attribute to the more exosomes upon CD151 removal as exosomes mediate TNF-α release in ECs^{62, 63}.

CD151, IFITM3, and VCP/p97 as inflammation regulators for therapeutic benefits

CD151 physically associates with IFITM3, which is localized mainly in endolysosome. Upon Cd151 ablation, more IFITM3 are accumulated in MVBs/lysosomes, leading to aberrant MVB/lysosome functions and then increased exocytosis. IFITM3 affects cellular entry, endosomal fusion, and inter-cellular dissemination of various enveloped viruses⁶⁴, and our observation of IFITM3 as an inflammation amplifier reveals not only its unknown role in inflammation but also its potential as a therapeutic target against inflammation.

CD151 prevents but IFITM3 supports exosomal release, probably also sorting, of cytokines and proteases, which exacerbate inflammation. Given their accessibility at cell surface, CD151 and IFITM3 can be regulated for their levels for therapeutic purposes. Together with aforementioned anti-inflammation merits of endothelial CD151, these unique features place CD151 in a promising position for clinical prevention and treatment of vascular inflammation and inflammatory diseases in future.

In conclusion, by forming protein quality control machinery with VCP/p97, CD151 maintains protein quality, facilitates protein turnover, and determines the protein turnover routes (proteolytic degradation versus exosomal discharge) at the limiting membrane of endolysosomes, to support the structural and functional integrity of endolysosomes, restrain cellular inflammatory response, and subsequently inhibit vascular inflammation (Figure S12). Without exogenous challenge, endothelial Cd151 ablation appears to affect mouse lung selectively. Such selectivity results unlikely from differential expression of CD151 in vasculature of different organs, given CD151 abundance in almost all endothelia⁶⁵, but possibly from endothelium property and mechanical environment of different vascular beds. Also, how exosomal discharge of pro-inflammatory molecules from Cd151-null ECs leads to pulmonary infiltration of leukocytes is unclear. These unanswered questions warrant further mechanistic analysis in future studies.