Abstract
The mechanisms underlying clear cell renal cell carcinoma (ccRCC)
metastasis remain largely unexplored. We demonstrate that Deleted in
Split hand/Split foot protein 1 (DSS1), a critical cofactor of BRCA2 in
DNA repair, is upregulated in metastatic ccRCC and promotes both tumor
growth and distant metastasis. Mechanistically, DSS1 interacts with LC3
and promotes its degradation via TRIM25-mediated Lys63 (K63)-linked
polyubiquitination at LC3B-K51. This impairs (macro) autophagic flux
and leads to p62 accumulation, thereby stabilizing TWIST1 and
facilitating its nuclear translocation, ultimately activating
epithelial-mesenchymal transition (EMT). DSS1 highly expressed
(DSS1^hi) tumor cells are enriched in late-stage tumors and are
associated with microvascular invasion within a vascularized invasive
niche at the tumor-stromal interface, mediated by SPP1-ITGB1
interactions. Clinically, DSS1^hi tumor cells correlate with
therapeutic resistance and poorer patient outcomes. Collectively, these
findings provide new insights into the mechanisms of ccRCC metastasis
and suggest potential avenues for therapeutic intervention.
Subject terms: Renal cell carcinoma, Macroautophagy, Cancer
microenvironment
__________________________________________________________________
DSS1 is reported to be involved in the maintenance of DNA repair and
protein homeostasis. Here, authors show that DSS1 is upregulated in
metastatic ccRCC, promoting tumor growth and metastasis by impairing
autophagy and stabilizing TWIST1.
Introduction
Clear cell renal cell carcinoma (ccRCC) accounts for approximately 70%
of the kidney tumors^[34]1. Patients with distant metastasis have a
5-year survival rate of less than 10%^[35]2. Many patients experience
substantial drug toxicity or develop resistance following
cytokine-based or targeted therapies, including mechanistic target of
rapamycin (mTOR) inhibitors^[36]3,[37]4 and vascular endothelial growth
factor receptor (VEGFR) tyrosine kinase inhibitors (TKIs)^[38]2.
Although combination therapies involving VEGFR-TKIs and immune
checkpoint inhibitors show promise in improving survival, the benefits
remain modest^[39]5 and only a limited number of patients
respond^[40]6,[41]7, making the treatment of metastatic ccRCC
particularly challenging.
Although many tumor cells can be eliminated during treatment, surviving
cancer cell subclones may undergo reprogramming to acquire
epithelial-mesenchymal transition (EMT) characteristics, thereby
facilitating metastasis^[42]8,[43]9. This highlights EMT as a
fundamental process through which cancer cells gain metastatic
potential^[44]10,[45]11. Moreover, tumor metastasis relies on the
support of stromal cells (e.g., fibroblasts and endothelial cells) and
immune cells (e.g., T cells and macrophages), as reducing this support
significantly diminishes the invasive capacity of tumor
cells^[46]12,[47]13. Recent advances in the study of the tumor
microenvironment (TME) have revealed that EMT-enriched tumor niches are
spatially localized at the tumor margins^[48]8,[49]14. However, the
mechanisms by which TME components are organized to promote distant
metastasis remain poorly understood. Therefore, dissecting highly
invasive niches at the levels of molecular mechanisms and intercellular
communication is essential to uncovering the cellular and molecular
basis of metastatic ccRCC.
Deleted in Split hand/Split foot protein 1 (DSS1), initially identified
as an autosomal dominant candidate gene for split-hand/foot
malformation^[50]15, was subsequently found to bind directly to breast
cancer susceptibility gene 2 (BRCA2), where it plays a critical role in
regulating R-loop-associated DNA damage and transcription-related
genomic instability^[51]16. More recently, DSS1 has been implicated in
mRNA transport, RNA splicing, and protein degradation^[52]15. DSS1 has
also been found to be upregulated in glioblastoma and associated with
poor prognosis, although its functional role in this context remains
unclear^[53]17. Here, we identify DSS1 as a driver gene in ccRCC
metastasis. DSS1 interacts with LC3 to promote its ubiquitin-mediated
degradation, resulting in impaired (macro)autophagic flux, p62
accumulation, and TWIST1 stabilization, thereby triggering EMT.
Distinct from SPP1⁺ macrophages in hypoxic regions^[54]18,[55]19, DSS1
highly expressed (DSS1^hi) tumor cells engage microvascular cells via
SPP1-ITGB1 signaling within a vascularized invasive niche at the
tumor-stroma interface, thereby promoting metastasis. The
identification of DSS1^hi tumor cells may advance our understanding of
the mechanisms underlying metastasis in advanced ccRCC.
Results
DSS1 upregulation is associated with tumor metastasis in ccRCC
Integrative analysis identified DSS1 as a metastasis-associated gene in
ccRCC (Fig. [56]1a, b). Comparative analysis demonstrated aberrant
upregulation of both DSS1 mRNA (Fig. [57]1c, Supplementary Fig. [58]1a)
and DSS1 protein levels (Fig. [59]1e) in advanced ccRCC relative to
early-stage tumors and normal kidney tissues. This upregulation may not
be influenced by patient sex, as no significant sex-specific
differences were observed between tumor and normal tissues
(Supplementary Fig. [60]1b). Elevated expression of DSS1 was an
unfavorable factor for patient survival of ccRCC (Fig. [61]1d, f), with
no sex-specific differences observed (Supplementary Fig. [62]1c). DSS1
was broadly upregulated across multiple cancer types and associated
with poorer patient survival (Supplementary Fig. [63]1d, e,
Supplementary Fig. [64]2a, b).
Fig. 1. DSS1 is upregulated in metastatic ccRCC and promotes tumor
progression.
[65]Fig. 1
[66]Open in a new tab
a Schematic workflow for identifying DSS1 as a metastasis-associated
driver gene in ccRCC. We leveraged Catalogue of Somatic Mutations in
Cancer (COSMIC) cancer gene census due to its expert-curated cancer
hallmark annotations (e.g., invasion, metastasis) from experimental
evidence. b Scatter plot showing DSS1/SEC31B mRNA levels (log[2]
transformed count per million [log[2]CPM]) vs. wound healing scores in
TCGA-KIRC cohort (n = 578 distinct samples, two-sided Spearman’s rank
correlation test). Dashed lines: linear regression fit; Shaded area:
the 95% confidence interval. c DSS1 mRNA expression in ccRCC versus
adjacent normal tissues across datasets (n = 560 distinct samples,
two-tailed Welch’s t-test). Boxplot: Center line = median; box = 25th
to 75th percentiles; whiskers = minima to maxima. d Kaplan-Meier
survival curves for DSS1 expression (DESeq2 log[2]CPM) in TCGA-KIRC
cohort (n = 522 distinct patients, Log-rank test). e
Immunohistochemistry (IHC) of DSS1 protein in ccRCC tumor and paired
normal tissues (n = 74 distinct patients; boxes: zoom-in regions; scale
bar: 200×, 100 μm, 400×, 50 μm; error bar: mean (centre) ± Standard
Deviation [SD]; two-tailed Welch’s t-test). ISUP, the International
Society of Urologic Pathologists. f Survival analysis stratified by
DSS1 IHC levels in the in-house cohort (n = 74 distinct patients;
Log-rank test). g, i Representative images of metastatic lung lesions
from xenograft mouse models (BALB/c-nu, male, 7 weeks before harvest,
tail vein injection) using Caki-1 cells (g) or of 786-O cells (i).
Scale bar: lung, 5 mm; 5×, 2 mm; 200×, 50 μm. Boxes: zoom-in regions.
h, j Quantification of metastatic foci number and total lesion area for
Caki-1 based models (h; Lentivirus [Lv]-shNC as control) or 786-O based
models (j; Lv-Vector as control). k Subcutaneous tumors (axillary fossa
injection) from Caki-1 cells-based mouse models (BALB/c-nu, male,
Lv-shNC as control, 4 weeks before harvest). l, m Tumor growth curves
(l) and final weights (m) of subcutaneous tumors. Lv-shNC as control.
g–m n = 6 mice per group, Error bar: mean ± SD, two-tailed Welch’s
t-test. Statistics are provided in the source data. Source data are
provided as a Source Data file.
DSS1 expression was associated positively with wound healing scores
(Fig. [67]1b), based on analysis of previously reported gene
signatures^[68]20,[69]21. To assess the role of DSS1 in regulating the
in vivo migratory capacity of ccRCC cells, we injected Caki-1 cells,
derived from a cutaneous metastasis of ccRCC, stably expressing either
a DSS1 knockdown construct Lentivirus (Lv)-shDSS1 or a negative control
(Lv-shNC) into nude mice via the tail vein. Among the shRNAs tested,
DSS1 shRNA#3, which exhibited the highest knockdown efficiency, was
selected for use in this and subsequent experiments (Supplementary
Fig. [70]2c). Mice in the Lv-shDSS1 group exhibited a marked reduction
in both the number and volume of lung metastatic nodules compared to
the Lv-shNC group seven weeks post-injection (Fig. [71]1g, h). To
further evaluate whether DSS1 enhances the metastatic potential of
non-metastatic ccRCC cells, we selected 786-O cells, which were
originally derived from a primary ccRCC tumor. 786-O cells have
previously been shown to successfully establish lung metastasis models
in vivo^[72]4,[73]22. Importantly, DSS1 expression in 786-O cells is
lower than that in Caki-1 cells (Supplementary Fig. [74]2d). To
establish metastatic mouse models, we injected 786-O cells stably
expressing either the DSS1 overexpression construct (Lv-DSS1) or the
empty vector control (Lv-Vector) via the tail vein. The efficacy of
DSS1 overexpression is shown in Supplementary Fig. [75]2e. Mice
injected with Lv-DSS1 cells developed a greater number and larger
volume of lung metastatic nodules compared to those injected with
Lv-Vector cells, as assessed seven weeks after injection (Fig. [76]1i,
j).
The sustained proliferative capacity of cancer cells is essential for
tumor progression. DSS1 expression was positively associated with a
proliferation-related gene signature (Supplementary Fig. [77]3a). Cell
Counting Kit-8 (CCK-8) and EdU assays demonstrated that DSS1 knockdown
suppressed the proliferation of ccRCC cells, whereas DSS1
overexpression enhanced their proliferative capacity (Supplementary
Fig. [78]3b–d). To further evaluate the role of DSS1 in tumorigenicity
in vivo, subcutaneous xenograft models were established using Caki-1
cells in nude mice. Compared with the Lv-shNC control group, DSS1
knockdown resulted in the formation of smaller, slower-growing tumors,
supporting a pro-tumorigenic role for DSS1 (4 weeks post-injection,
Fig. [79]1k−m).
DSS1 promotes migration and invasion of ccRCC cells by blocking autophagic
flux
Pathway enrichment analysis suggested that autophagy may be negatively
regulated by DSS1 in ccRCC metastasis, as indicated by a normalized
enrichment score (NES) of −1.69 in the kidney renal clear cell
carcinoma cohort from The Cancer Genome Atlas (TCGA-KIRC, Fig. [80]2a).
This hypothesis was supported by Gene Set Enrichment Analysis (GSEA,
DSS1_hi vs. DSS1_lo, median as a cut-off) of three additional datasets,
including [81]GSE251905^[82]23 (NES = −1.81), [83]GSE3538^[84]24
(NES = −1.49), [85]GSE254566^[86]25 (NES = −1.20, Supplementary
Fig. [87]4a). In contrast, GSEA results for the mTOR pathway showed
inconsistent NES directions across datasets: TCGA-KIRC (NES = −1.30)
and [88]GSE251905 (NES = −1.64) showed negative enrichment, whereas
[89]GSE3538 (NES = 0.99) and [90]GSE254566 (NES = 1.00) showed positive
enrichment (Supplementary Fig. [91]4a). These discrepancies suggest
that DSS1 expression is unlikely to be consistently associated with the
mTOR pathway. Since GSEA was based on transcriptomic profiles, we
further investigated the correlation between DSS1 expression and the
RNA levels of autophagy-related genes. We performed RT-qPCR on 37 genes
commonly contributing to autophagy enrichment. Among these, 12 genes
were significantly upregulated in DSS1-knockdown cells and are known to
regulate key stages of autophagy, including initiation (TRAF6, ULK2),
elongation/maturation (TBK1, SMCR8), PI3K-III complex and nucleation
(ATG14, STX17, UVRAG), LC3 lipidation (ATG4C, ATG7), and
autophagosome-lysosome fusion (STX17, SNAP29, NBR1, CALCOCO2)
(Supplementary Fig. [92]4b). Together, these transcriptomic and
experimental findings support a potential inhibitory role of DSS1 in
the regulation of autophagy.
Fig. 2. DSS1 suppresses autophagic flux in ccRCC cells.
[93]Fig. 2
[94]Open in a new tab
a Schematic diagram illustrating KEGG autophagy pathway negatively
associated with DSS1 expression (DESeq2 log[2]CPM). b Transmission
electron microscopy (TEM, left) images and quantification (right) of
autophagic vacuoles (arrows) in ACHN and Caki-1 cells transfected with
siDSS1 or siNC (n = 3 independent experiments; scale bars: 8000×, 2 μm,
20000×, 0.5 μm; boxes: zoom-in regions; error bar: mean ± SD;
two-tailed Welch’s t-test, Benjamini-Hochberg [BH] adjustment). c
Confocal microscopy images of LC3 puncta (magenta) and LAMP1
(turquoise) co-localization in cells treated with siNC, siDSS1, with or
without chloroquine (CQ: 10 μM, control: 0.02% DMSO, 24 h before
harvest; n = 3 independent experiments; scale bar: 10 μm; error bar:
mean ± SD; two-tailed Welch’s t-test, BH adjustment). COSMIC, Catalogue
of Somatic Mutations in Cancer. Statistics are provided in the source
data. Source data are provided as a Source Data file.
Transmission electron microscopy revealed an increased accumulation of
autophagic vacuoles in DSS1-silenced ccRCC cells (Fig. [95]2b). DSS1
knockdown also led to a significant increase in LC3 puncta formation
and LC3-LAMP1 colocalization, both in the presence and absence of
chloroquine (CQ), a lysosomal acidification inhibitor that blocks
autophagosome degradation (Fig. [96]2c). These findings support the
role of DSS1 as a negative regulator of autophagic flux. Furthermore,
Transwell assays demonstrated that DSS1 silencing markedly suppressed
the invasive and migratory abilities of ccRCC cells, whereas DSS1
overexpression enhanced both processes (Fig. [97]3a, b). However, in
the presence of CQ, the differences in invasion and migration between
DSS1-silenced and control cells were no longer significant
(Fig. [98]3a, b), suggesting that DSS1 modulates invasion and migration
in an autophagy-dependent manner. Morphometric analysis further showed
that control Caki-1 cells exhibited a typical spindle-shaped
morphology, while DSS1-silenced cells displayed a rounded, pebble-like
appearance, indicative of reduced migratory and invasive potential
(Fig. [99]3c).
Fig. 3. DSS1 enhances ccRCC cell migration and invasion via autophagy
inhibition.
[100]Fig. 3
[101]Open in a new tab
a, b Transwell migration and Matrigel invasion assays in ACHN and
Caki-1 cells treated with shNC, shDSS1, pcDNA3.1, DSS1 plasmids, with
or without CQ (Chloroquine, representative images; scale bar: 100 μm).
Quantification of migrated/invaded cells (error bar: mean ± SD;
two-tailed Welch’s t-test; n = 3 independent experiments). c Morphology
images of ccRCC cells between DSS1 knockdown (pebble-shaped) and
negative control (spindle-shaped; scale bar: 100 μm; error bar:
mean ± SD; two-tailed Welch’s t-test). The average cell aspect ratio
was determined from 100 randomly selected cells per group per
experiment (n = 3 independent experiments). d, e Immunoblotting (IB) of
autophagy markers (LC3-II, p62) and EMT proteins (E-cadherin,
N-cadherin, Vimentin) in siNC- and siDSS1-treated (with or without CQ,
0.02% DMSO as control) cells. Endogenous control: β-actin. Densitometry
quantification (e, n = 3 independent experiments, error bar: mean ± SD,
two-tailed Welch’s t-test). The samples derived from the same
experiment were run on parallel gels, with each gel probed for a
different antibody. f RT-qPCR analysis of MAP1LC3B and SQSTM1 mRNA
levels in siDSS1 vs. siNC cells (normalized to GAPDH; n = 3 independent
experiments; error bar: mean ± SD; two-tailed Welch’s t-test).
Statistics are provided in the source data. Source data are provided as
a Source Data file.
We next investigated the mechanism by which DSS1 blocks autophagic flux
and promotes cell migration and invasion. DSS1 silencing increased
LC3-II protein levels regardless of CQ treatment, suggesting that DSS1
depletion enhances LC3 induction rather than impairs autophagosome
clearance (Fig. [102]3d, e). In addition, DSS1 knockdown led to
upregulation of E-cadherin and downregulation of p62, Vimentin, and
N-cadherin (Fig. [103]3d, e). However, in the presence of CQ, the
levels of these proteins did not differ significantly between
DSS1-silenced and control cells (Fig. [104]3d, e), indicating that the
observed changes of EMT markers are dependent on functional autophagic
flux. Moreover, a low correlation was observed between DSS1 mRNA levels
and those of MAP1LC3B or SQSTM1 (Supplementary Fig. [105]3e).
Consistently, no significant differences in MAP1LC3B or SQSTM1
transcript levels were detected between DSS1-silenced cells and
controls (Fig. [106]3f). These findings suggest that DSS1 inhibits
autophagic flux in ccRCC cells by suppressing LC3-II induction, thereby
promoting EMT activation.
DSS1 promotes EMT activation by regulating TWIST1 in ccRCC
Next, we investigated how DSS1-mediated autophagy suppression leads to
EMT activation. Transforming growth factor-β (TGF-β), a canonical
upstream inducer of EMT, has also been reported to inhibit
autophagy^[107]26. We therefore examined whether DSS1 modulates TGF-β
expression. However, TGF-β levels were not altered in DSS1-silenced
ccRCC cells compared to negative controls (Supplementary Fig. [108]5a,
b). Given that the protein levels of E-cadherin, N-cadherin, and
Vimentin were affected by autophagy inhibition (Fig. [109]3d, e), we
hypothesized that DSS1 may promote EMT by regulating EMT-inducing
transcription factors, such as ZEB1/2, SNAI1/2, and TWIST1. To identify
transcription factors potentially involved in the DSS1-autophagy-EMT
regulatory axis, we developed a Consensus Scoring of Genes in Cascade
(CSGC) algorithm (Fig. [110]4a). Candidate transcription factors were
expected to meet three criteria: (1) upregulated in ccRCC or metastatic
ccRCC; (2) positively correlated with both EMT activity and DSS1
expression; and (3) negatively correlated with autophagy activity. The
activities of EMT and autophagy in ccRCC samples were quantified using
single-sample gene set enrichment analysis (ssGSEA) based on
established gene signatures (Fig. [111]4a). Among the screened factors,
TWIST1 was identified as the top candidate.
Fig. 4. DSS1 drives epithelial-mesenchymal transition (EMT) in ccRCC through
autophagy inhibition and TWIST1 stabilization.
[112]Fig. 4
[113]Open in a new tab
a Identification of TWIST1 as a potential downstream effector of
DSS1-autophagy axis. Gene signatures (Molecular Signature DataBase
v2022.1) were analyzed by single-sample gene set enrichment analysis
(ssGSEA) in TCGA-KIRC (n = 522 distinct tumors). b Fold changes
(DESeq2) of transcription factor (TF) mRNA expression in metastatic
(Met.) vs. primary ccRCC and primary ccRCC vs. normal kidney
(TCGA-KIRC). c Immunoblotting of EMT-TFs (ZEB1, ZEB2, SNAI1,
SLUG/SNAI2, FOXC2, TCF3) in shDSS1 vs. shNC cells (two-tailed Welch’s
t-test, Benjamini-Hochberg [BH] adjustment). β-actin: endogenous
control. d Immunoblotting showing TWIST1 protein levels in cells
treated with siNC/siDSS1, autophagy inhibitor (CQ, 10 μM; Control:
0.02% DMSO; 24 h before harvest), or both (two-tailed Welch’s t-test).
e Immunoblotting showing TWIST1 levels in cells treated with siNC,
TWIST1 knockdown, pcDNA3.1 or TWIST1 plasmids (two-tailed Welch’s
t-test). f Rescue experiments showing protein levels of EMT markers in
ccRCC cells transfected with different combinations of control, DSS1,
and TWIST1 siRNAs, and DSS1 and TWIST1 plasmids (two-tailed Welch’s
t-test). g Transwell assays following the same treatments as in panel
(f) (scale bar: 100 μm, two-tailed Welch’s t-test). h
Immunohistochemistry using lung metastatic lesions of xenograft mouse
models generated by tail vein injection of Caki-1 cells treated with
Lv-shDSS1 or control Lv-shNC (n = 5 independent samples per group;
scale bar: 100 μm; two-tailed Welch’s t-test). IOD, Integrated Optical
Density. i Caki-1 cells were analyzed by immunoprecipitation with
antibody to the LC3 or TWIST1 epitope, followed by SDS-PAGE and
immunoblotting (IgG: negative control, input: 5% lysate). j
Immunofluorescence microscopy showing the co-localization of Bcl-2
(magenta), p62 (orange), and TWIST1 (turquoise) in ACHN cells treated
with DSS1 knockdown or control (scale bar: 5 μm). DAPI (blue): nucleus.
c–f The samples derived from the same experiment were run on parallel
gels, with each gel probed for a different antibody. c–g, i, j n = 3
independent experiments. c–h Error bars: mean ± SD. Statistics are
provided in the source data. Source data are provided as a Source Data
file.
Previous studies have reported that both TWIST1 mRNA and TWIST1 protein
levels are upregulated in ccRCC^[114]27. Immunoblotting confirmed that
DSS1 silencing reduced TWIST1 protein levels (Fig. [115]4d), while the
expression of other EMT-related transcription factors that are commonly
upregulated in ccRCC or metastatic ccRCC remained unaffected
(Fig. [116]4b, c). Notably, in the presence of CQ, TWIST1 levels were
comparable between DSS1-silenced and control cells (Fig. [117]4d),
indicating that DSS1 modulates TWIST1 protein levels through autophagy
regulation. To further assess the role of TWIST1 in DSS1-mediated EMT,
we performed rescue experiments. TWIST1 silencing in
DSS1-overexpressing cells led to increased E-cadherin levels and
decreased N-cadherin and Vimentin levels (Fig. [118]4e, f). Conversely,
TWIST1 overexpression in DSS1-silenced cells reversed these effects,
resulting in decreased E-cadherin and increased N-cadherin and Vimentin
expression (Supplementary Fig. [119]5c, d). In contrast, TWIST1
overexpression abolished the effects of DSS1 knockdown on these EMT
markers, with no significant changes observed in E-cadherin,
N-cadherin, or Vimentin levels between DSS1-silenced and control groups
(Fig. [120]4f). Functionally, TWIST1 silencing suppressed the invasive
and migratory capabilities of DSS1-overexpressing ccRCC cells, whereas
DSS1 knockdown had no significant impact on these behaviors in
TWIST1-overexpressing cells (Fig. [121]4g). These findings indicate
that DSS1 promotes EMT, migration, and invasion through upregulation of
TWIST1. Consistently, analysis of lung metastatic lesions from
xenograft mouse models (tail vein injection of Caki-1 cells) revealed
that DSS1 knockdown resulted in lower p62 and TWIST1 protein levels,
alongside increased LC3 and E-cadherin levels, compared to lesions from
control mice (Fig. [122]4h). Together, these in vitro and in vivo
findings demonstrate that DSS1 suppresses autophagy to promote EMT via
TWIST1 regulation.
We next investigated how DSS1-mediated autophagy suppression regulates
TWIST1 protein levels. Co-immunoprecipitation (Co-IP) assays did not
detect an interaction between LC3 and TWIST1 (Fig. [123]4i), suggesting
that TWIST1 regulation is unlikely to involve selective autophagy.
Instead, it may depend on p62-mediated stabilization, as previous
studies have shown that p62 accumulation inhibits TWIST1
degradation^[124]28. In addition, Bcl-2, a known upstream regulator of
both autophagy and apoptosis, has been reported to interact with
TWIST1, facilitating its nuclear translocation and promoting
angiogenesis and metastasis^[125]29. Interestingly, Bcl-2 protein
levels were reduced in DSS1-knockdown ccRCC cells, whereas Beclin-1
levels remained unchanged (Supplementary Fig. [126]5a, b). To explore
whether Bcl-2 and p62 are involved in DSS1-mediated EMT activation, we
performed immunofluorescence analyses. Compared to control cells, DSS1
knockdown reduced both the nuclear co-localization of Bcl-2 and TWIST1,
and the cytoplasmic co-localization of p62 and TWIST1 (Fig. [127]4j,
Supplementary Fig. [128]5e). These findings suggest that DSS1
suppresses autophagy to promote p62 accumulation, thereby stabilizing
TWIST1 and enhancing its association with Bcl-2. This facilitates
TWIST1 nuclear translocation and contributes to EMT activation.
DSS1 interacts with LC3 and promotes TRIM25-mediated polyubiquitination
degradation of LC3
We next investigated the mechanism by which DSS1 inhibits LC3-II
induction. As mTOR is a central regulator of autophagy initiation and
the autophagic lysosome reformation cycle^[129]30, we examined whether
DSS1 modulates autophagy by altering mTOR activity. However, the
protein levels of total mTOR and phosphorylated mTOR (p-mTOR) remained
unchanged between DSS1-knockdown and control ccRCC cells (Supplementary
Fig. [130]5a, b), indicating that DSS1 suppresses autophagy via an
mTOR-independent pathway. LC3 is a key protein involved in the
elongation of phagophores and the maturation of autophagosomes^[131]31.
Given that DSS1 knockdown increased LC3-II levels (Fig. [132]3d, e),
and that DSS1 is a known component of the 19S proteasome regulatory
particle complex, we hypothesized that DSS1 facilitates the degradation
of ubiquitinated LC3. Co-IP assays demonstrated a physical interaction
between DSS1 and LC3 (Fig. [133]5a, b). This interaction was further
validated by glutathione-S-transferase (GST) pulldown assays using
recombinantly expressed human proteins (Fig. [134]5c). GST-tagged DSS1
migrated at ~38 kDa on SDS-PAGE, consistent with previous
reports^[135]32. Recombinant full-length human LC3B (pro-LC3B) migrated
at ~14 kDa, as pro-LC3B runs similarly to LC3B-II in SDS-PAGE^[136]33.
Fig. 5. DSS1 interacts with LC3 and promotes LC3 polyubiquitination and
degradation.
[137]Fig. 5
[138]Open in a new tab
a Co-Immunoprecipitation (Co-IP) with antibody to the FLAG epitope,
followed by SDS-PAGE and immunoblotting with antibody to LC3
(endogenous, IgG: isotype control; Input: 5% lysate). Short/long:
exposure time of the same membrane. b Co-IP with antibody to the LC3
epitope, followed by SDS-PAGE and immunoblotting showing endogenous
DSS1. a, b IgG light chain-specific secondary antibody was used. c
Glutathione-S-Transferase (GST)-pulldown analysis for DSS1-LC3B
interaction using 50 ng recombinantly expressed human full-length LC3B
and 1 mg GST/GST-DSS1 (GST: negative control; Purified LC3B: positive
control; Glutathione agarose beads: 50 µL). d Cells were lysed with
EDTA (10 mM) or not, followed by metal bath at 95 °C for 5 min and
immunoblotting with antibody to DSS1. e GST-pulldown analysis showing
the interaction of LC3 and DSS1 mutant using 50 ng recombinantly
expressed human LC3 and 1 mg GST-DSS1^W27GW39GW43GF52A (Glutathione
agarose beads: 50 µL). f Caki-1 cells were transfected with plasmids
encoding HA-DSS1^W27GW39GW43GF52A. Cell lysates and cell
immunoprecipitants were immunoprecipitated with antibody to the LC3
epitope and HA epitope, respectively. g Knockdown of DSS1 or treatment
with MG-132 (20 μM, 6 h before harvest) elevates LC3 protein levels in
HEK293T cells (control: siNC and/or 0.02% DMSO, error bar: mean ± SD,
two-tailed Welch’s t-test). h Immunoblotting showing the indicated
protein levels in HEK293T cells transfected with control or DSS1 siRNA,
and/or treated with 0.02% DMSO or MG-132 (20 μM, 6 h before harvest). i
Immunoblotting of cycloheximide (CHX, 50 μg/mL) chase assay in shDSS1
cells compared to shNC cells (error bar: mean ± SD, two-tailed Welch’s
t-test). c–e, g–i The samples derived from the same experiment were run
on parallel gels, with each gel probed for a different antibody. a–i
n = 3 independent experiments. Statistics are provided in the source
data. Source data are provided as a Source Data file.
Interestingly, human DSS1 (70 amino acids; predicted molecular weight
~8 kDa) exhibited aberrant migration on SDS-PAGE, appearing at ~18 kDa,
~37 kDa, and higher molecular weights^[139]34,[140]35. This aberrant
migration is partially attributed to its highly acidic nature, which
impairs SDS binding^[141]36 (theoretical isoelectric point [pI] ~3.81,
calculated using the ExPASy Compute pI/Mw tool:
[142]https://web.expasy.org/compute_pi/). The additional bands may
represent oligomeric forms or adducts of DSS1. A sharp ~37 kDa DSS1
band was consistently detected in anti-LC3 immunoprecipitates
(Fig. [143]5b). DSS1 has previously been shown to form SDS-resistant
oligomers and adducts with cellular proteins, mediated by four critical
residues (W27, W39, W43, and F52), which are also implicated in
ubiquitination-related degradation^[144]34. The 37 kDa oligomeric form
has been observed in both endogenous and purified preparations of DSS1,
independent of redox status or metal ion availability^[145]34,[146]35.
Reactive oxygen species (ROS), induced by UV or H₂O₂ treatment, promote
DSS1 adduct formation, whereas scavengers such as EDTA or vitamin C
suppress adduct formation. Notably, these treatments do not affect the
levels or migration pattern of the DSS1 oligomer^[147]34. To assess
whether the 37 kDa band observed in anti-LC3 IP products represents a
DSS1 adduct, we treated cells with EDTA. However, EDTA treatment did
not eliminate the 37 kDa band (Fig. [148]5d), suggesting it is not a
ROS-mediated adduct. Furthermore, GST pulldown assays using purified
LC3B and mutant GST-DSS1^W27GW39GW43GF52A confirmed that LC3B can still
bind to DSS1 in the absence of the adduct-forming residues
(Fig. [149]5e), consistent with co-immunoprecipitation results from
ccRCC cells (Fig. [150]5f). Taken together, the ~37 kDa DSS1 band
likely represents a DSS1 oligomer, as it lacks features of
glycosylation or ubiquitination (e.g., smearing), shows no evidence of
methylation, acetylation, or lipidation (e.g., subtle mass shifts), and
remains resistant to reductants (e.g., DTT)^[151]34 and ROS scavengers.
DSS1 knockdown or treatment with MG-132, a proteasome inhibitor, in
HEK293T cells led to increased levels of both total LC3 and
ubiquitinated LC3B (Fig. [152]5g, h), indicating that LC3 degradation
was inhibited. In contrast, MG-132 treatment equalized the levels of
ubiquitinated LC3 and restored LC3 protein abundance, as proteasomal
degradation was blocked regardless of DSS1 expression (Fig. [153]5g,
h). These findings suggest that DSS1 knockdown impairs LC3 degradation,
resulting in the accumulation of ubiquitinated LC3. To further
strengthen the mechanistic link between DSS1-mediated autophagy
suppression and LC3 degradation, we assessed LC3 protein stability
following treatment with cycloheximide (CHX), a protein synthesis
inhibitor. Time-course analyses revealed that LC3 degradation was
significantly slower in DSS1-knockdown cells compared to controls
(Fig. [154]5i). We next examined the subcellular localization of
DSS1-LC3 interactions using immunofluorescence. PSMD3 (also known as
Rpn3), a subunit of the 19S regulatory particle of the proteasome and a
known DSS1 interactor^[155]37, was used as a proteasomal marker.
Co-localization of DSS1 and LC3 with PSMD3 was observed predominantly
in the cytoplasm (Fig. [156]6a, Supplementary Fig. [157]6a).
Collectively, these results demonstrate that DSS1 binds to LC3 and
promotes its degradation via the ubiquitin-proteasome system.
Fig. 6. TRIM25 mediates LC3 polyubiquitination and degradation.
[158]Fig. 6
[159]Open in a new tab
a Representative images of immunofluorescence showing the subcellular
localization of DSS1 (magenta), LC3 (yellow), and PSMD3 (azure) in ACHN
cells (n = 3 independent experiments). Nuclei: DAPI (blue). Scale bar:
5 μm. b LC-MS/MS analysis of LC3 interactomes identifies E3 ligases
potentially associated with LC3 (n = 3 independent experiments;
Proteome Discoverer 2.4, Thermo Scientific; False discovery rate
<0.01). c The mRNA expression of TRIM25 in ccRCC patients from the
TCGA-KIRC dataset (n = 595 distinct samples, tumor vs. normal,
two-tailed Welch’s test). Boxplot: Center line = median; box = 25th to
75th percentiles; whiskers = minima to maxima. d Representative images
of immunohistochemistry analysis showing TRIM25 protein levels
upregulated in ccRCC tissues vs. normal (n = 66 distinct pairs; scale
bar: 50 μm; error bar: mean (centre) ± SD; two-tailed paired t-test).
e, f Immunoblotting showing TRIM25 knockdown increases LC3 levels,
while TRIM25 overexpression reduces LC3 levels (n = 3 independent
experiments, error bar: mean ± SD, two-tailed Welch’s t-test). g Co-IP
analysis in Caki-1 cells showing the interaction between endogenous
TRIM25 and LC3 (n = 3 independent experiments). h GST-pulldown analysis
for TRIM25-LC3B interaction using 50 ng recombinantly expressed human
full-length LC3B and 1 mg GST-TRIM25 (n = 3 independent experiments;
Glutathione agarose beads: 50 µL). Asterisk: GST-TRIM25. i
Immunoblotting of HEK293T cells transfected with HA-Ub, pcDNA3.1, and
Myc-LC3B (wildtype) or Myc-LC3B mutant plasmids (n = 3 independent
experiments). j Immunoblotting of HEK293T cells transfected with
Myc-LC3B and HA-Ub-wildtype or HA-Ub-K48R/K63R mutant plasmids (n = 3
independent experiments). k Recombinantly expressed human GST-TRIM25
was incubated with recombinant Ub, LC3B, UbcH5a/UbcH5c, and His-UBE1 in
a reaction system containing MgATP at 37 °C for 4 h. Samples were
analyzed by SDS-PAGE and immunoblotting with antibody to Ub epitope
(n = 3 independent experiments). As shown in the result of GST-TRIM25,
the polyubiquitination of LC3B (LC3B-Ub) is shown only when all the
components are present in the mix. e, f, h–j The samples derived from
the same experiment were run on parallel gels, with each gel probed for
a different antibody. Statistics are provided in the source data.
Source data are provided as a Source Data file.
We next investigated the mechanism by which LC3 is ubiquitinated.
LC-MS/MS analysis of anti-LC3 immunoprecipitates identified two
candidate E3 ubiquitin ligases: RNF213 and TRIM25 (Fig. [160]6b).
TRIM25, but not RNF213, exhibited a negative correlation with LC3
protein levels based on immunohistochemistry (IHC) data from renal
cancer tissues in the Human Protein Atlas (HPA) database^[161]38
(Supplementary Fig. [162]6b). TRIM25 was significantly upregulated in
ccRCC, as shown by both TCGA-KIRC transcriptomic data and our IHC
analysis (Fig. [163]6c, d), with no observed sex-specific differences
(Supplementary Fig. [164]6c). Functionally, TRIM25 knockdown increased
LC3 protein levels (Fig. [165]6e), while TRIM25 overexpression
decreased LC3 levels (Fig. [166]6f), indicating that TRIM25 promotes
LC3 degradation. Co-IP confirmed that TRIM25 interacts with LC3
(Fig. [167]6g), and this interaction was further validated by GST
pulldown assays using recombinantly expressed human TRIM25 and LC3B
(Fig. [168]6h).
To identify the ubiquitination site on LC3B, we individually mutated
each of its ten lysine (K) residues to arginine (R), which cannot be
ubiquitinated (Supplementary Fig. [169]6d). Among these, only the K51R
mutation abolished ubiquitination, indicating that lysine 51 is the
primary site of polyubiquitination on LC3B (Fig. [170]6i). Furthermore,
LC3B ubiquitination was reduced by K63R, but not K48R mutant ubiquitin,
suggesting that TRIM25 mediates K63-linked polyubiquitination of LC3B
(Fig. [171]6j). To directly test whether TRIM25 catalyzes LC3B
polyubiquitination, we performed an in vitro ubiquitination assay using
recombinant LC3B, ubiquitin, E1 (His-UBE1), E2
(UbcH5a/UbcH5c)^[172]39,[173]40, a source of ATP (MgATP), and
GST-tagged human TRIM25 purified from E. coli. Immunoblot analysis of
the complete reaction mixture with recombinant GST-TRIM25 revealed
high-molecular-weight bands consistent with polyubiquitinated LC3B, as
detected by anti-ubiquitin antibody (Fig. [174]6k). Together, these
findings demonstrate that TRIM25 directly mediates K63-linked
polyubiquitination of LC3B at lysine 51.
Functional states and clinical value of DSS1-driven cells in ccRCC
To evaluate the clinical significance of DSS1-driven cells in ccRCC, we
analyzed single-cell RNA sequencing (scRNA-seq) data from 49 ccRCC
samples across three independent datasets. After quality control, a
total of 201,810 cells were retained and integrated (Fig. [175]7a).
Following batch correction and cell-type annotation, we identified
30,662 epithelial cells (EPCAM and KRT8), 22,180 fibroblasts (ACTA2 and
MYL9), 18,821 endothelial cells (ENG and CLEC14A), 2768 B cells (BANK1
and MS4A1), 933 plasma cells (MZB1 and TNFRSF17), 83,987 T/NK cells
(CD2, CD3, NKG7, and GZMK), 39,758 myeloid cells (CD68 and FCGR3A), and
2701 mast cells (KIT and VWA5A, Fig. [176]7b, c). To identify malignant
epithelial cells, we isolated the epithelial population for
re-clustering and copy number variation (CNV) analysis using the
inferCNV algorithm. Prior to CNV estimation, four clusters (5893 cells)
expressing marker genes from multiple lineages (e.g., immune cells)
were removed as likely doublets. The analysis identified 11,501
malignant epithelial cells (Fig. [177]7d, e). At single-cell
resolution, DSS1 expression was significantly higher in malignant
epithelial cells compared to normal kidney epithelial cells
(Fig. [178]7f).
Fig. 7. Single-cell transcriptomic profiling reveals clinical relevance of
DSS1-driven malignant cells in ccRCC.
[179]Fig. 7
[180]Open in a new tab
a Quality control for single-cell RNA-seq data from ccRCC tissues
(n = 49 distinct samples). b, c Uniform Manifold Approximation and
Projection (UMAP) of cells of ccRCC tissues before (b) and after batch
effect correction (c). Right panel: annotation of major cell types. d,
e Identification of malignant cells using inference of copy number
variation (inferCNV, CNV score threshold: 0.04) of epithelial cells.
Del, deletion; Amp amplification. f Comparison of DSS1 pseudo-bulk
expression between malignant cells (tumor samples) and normal
epithelial cells (normal samples, error bar: mean ± SD, n = 47 distinct
samples with detected DSS1 RNA; two-tailed Welch’s t-test). g DSS1
amplification correlates with higher DSS1 expression (DESeq2 log[2]CPM)
in tumors (n = 513). h Functional module analysis for malignant cells.
Upper graph: hierarchical clustering dendrogram of genes; lower graph:
gene module (x-axis, different genes). i UMAP showing DSS1-driven ccRCC
cells and other malignant cells. j Comparison of the
epithelial-mesenchymal transition (EMT)/autophagy module scores between
DSS1-driven cells and other malignant cells. KEGG v113.0, Kyoto
Encyclopedia of Genes and Genomes. k Proportion of DSS1-driven cells
across different stages in treatment-naive patients (I-IV: n = 1, 6, 6,
3 patients, respectively). l DSS1-driven cell signature scores in
patients with different grades and pathologic stages (n = 521). m
Comparison of DSS1-driven cell signature scores across molecular
subtypes defined by Clinical Proteomic Tumor Analysis Consortium
(CPTAC) in TCGA-KIRC dataset (n = 495). n High DSS1-driven cell
signature predicts poorer survival (Log-rank test, n = 506 and 509,
respectively). HR, Hazard Ratio. o–q Differences in signature scores of
DSS1-driven cells between treatment-naive/response and
treatment/resistance groups (VEGFR inhibitors) in ccRCC patients or
patient-derived xenograft (PDX) models (left to right, TCGA-KIRC,
n = 430; [181]GSE65615, n = 122; [182]GSE64052, n = 28; [183]GSE76068,
n = 16). MUT, mutant; WT, wildtype. g, j, l, m, o–q Two-tailed
Mann-Whitney U test. Box plots (Turkey style): Center line = median;
box = 25th to 75th percentiles; whiskers = ±1.5×interquartile range;
Violin plots: data distribution (minima to maxima) via kernel density
estimation. Statistics are provided in the source data. Source data are
provided as a Source Data file.
DSS1 is located on chromosome 7q21.3, and consistent with prior genomic
studies^[184]41 gains of chromosome 7 were observed. To examine whether
DSS1 expression is affected by CNV status, we analyzed matched
transcriptomic and genomic profiles. Samples harboring DSS1
amplification showed significantly elevated DSS1 expression compared to
those without amplification (Fig. [185]7g), with no significant
sex-specific differences (Supplementary Fig. [186]6e). Notably, DSS1
amplification and its associated transcriptional upregulation were
recurrent across 32 human cancer types (Supplementary Fig. [187]6f),
underscoring its potential as a broadly relevant oncogenic driver.
Given the high degree of heterogeneity among malignant cells, we
applied high-dimensional weighted gene coexpression network analysis
(hdWGCNA) to characterize their functional states (Fig. [188]7h). This
analysis revealed 14 coexpression modules associated with diverse
biological processes, including cell metabolism (e.g., ALDHs, FABPs),
EMT (e.g., LAMC2, VIM, FN1, COL1A1), and programmed cell death (e.g.,
LC3B) (Fig. [189]7h, Supplementary Fig. [190]7a). We identified 1,403
DSS1-driven malignant cells (referred to as DSS1^hi) with elevated DSS1
expression—defined as at least 50% higher than the average expression
in normal epithelial cells (Fig. [191]7i, Supplementary Fig. [192]7c).
Compared to other malignant cells, DSS1^hi cells exhibited
significantly higher DSS1 levels (log₂ fold change = 2.08, P = 0,
two-tailed Mann-Whitney U test). Functionally, DSS1^hi cells showed
higher EMT module scores and lower autophagy scores than other
malignant cells (Fig. [193]7j). In patients with metastatic ccRCC
(stage IV), DSS1^hi cells had significantly elevated EMT and wound
healing scores compared to those from non-metastatic cases
(Fig. [194]7j, Supplementary Fig. [195]7b). Additionally, with disease
progression, DSS1^hi cells displayed increased cell cycle activity,
modestly enhanced angiogenic signatures, and reduced programmed cell
death compared to other malignant cells (Supplementary Fig. [196]7b).
In treatment-naive patients, the frequency of DSS1^hi cells was
significantly higher in stage IV tumors than in earlier-stage tumors
(Fig. [197]7k), further underscoring their potential role in tumor
progression and metastasis.
Given the accessibility and clinical relevance of bulk RNA sequencing
data, we investigated the significance of DSS1-driven cells across bulk
ccRCC cohorts. Marker genes of DSS1^hi cells were used to calculate
signature scores for each bulk sample. Consistent with single-cell
findings, higher signature scores were associated with poorer tumor
differentiation (Grade), larger tumor size (T), and the presence of
distant metastasis (M) (Fig. [198]7l). When stratified by immune
subtype^[199]42, the signature score was highest in the CD8⁻ inflamed
subtype, which was characterized by elevated endothelial, fibroblast,
and macrophage infiltration, high EMT, oxidative phosphorylation, and
angiogenesis, as well as frequent chromosome 7 gains. Scores were also
elevated in the CD8⁺ inflamed subtype, which exhibits high immune
checkpoint expression and interferon response. In contrast, scores were
lowest in the metabolic/VEGF immune-desert subtype, marked by high
MTORC1 signaling, HIF-1/glycolysis activity, and angiogenesis
(Fig. [200]7m).
To assess the prognostic relevance, patients were classified into high-
and low-score groups based on the median signature score. High-score
patients had significantly worse survival outcomes (Fig. [201]7n). It
is expected that signature scores were associated with chemotherapy, as
DSS1 was demonstrated to be a critical protein of DNA repair, such as
double-strand breaks^[202]43 (Fig. [203]7o). Signature scores were
significantly higher in patients resistant to anti-VEGFR therapy than
in those with progression-free disease, independent of Von
Hippel-Lindau (VHL) mutation status (Fig. [204]7o). These results were
validated in an independent cohort^[205]44, where patients treated with
sunitinib exhibited higher DSS1 signature scores than treatment-naive
patients, again independent of VHL mutation status (Fig. [206]7p).
Consistently, RCC cell lines (with VHL mutations)^[207]45 treated with
anti-VEGFR agents showed enrichment of DSS1-driven transcriptional
signatures compared to untreated cells (Fig. [208]7q). Moreover, in
patient-derived xenograft (PDX) models^[209]46 treated with sunitinib
(40 mg/kg/day orally for 4 weeks), tumors from non-responders had
significantly higher levels of DSS1-driven cells than those from
responders (Fig. [210]7q). These findings collectively suggest that
DSS1-driven cells are associated with aggressive clinical features,
resistance to anti-VEGFR therapies, and adverse patient outcomes.
To identify potential therapeutic strategies targeting DSS1-driven
cells, we applied two predictive frameworks: beyondcell^[211]47 and
scTherapy^[212]48. Drug sensitivity analysis using beyondcell revealed
candidate compounds targeting key downstream effectors of DSS1, such as
SQSTM1 (e.g., glucocorticoid receptor agonist prednisolone), Bcl-2
(e.g., CCR8 agonist ZK-756326), and TWIST1 (e.g., GABA receptor agonist
NCS-382). However, no single compound was predicted to simultaneously
target all three effectors (Supplementary Fig. [213]7d). In contrast,
scTherapy analysis based on DSS1-driven cell transcriptomes identified
several compound classes with potential efficacy, including proteasome
inhibitors, histone deacetylase (HDAC) inhibitors, and autophagy
activators (Supplementary Fig. [214]7d). These findings suggest that a
therapeutic combination involving proteasome inhibition and autophagy
activation may represent a rational strategy for selectively targeting
DSS1-driven tumor cells.
DSS1-driven cells communicate with microvascular endothelial cells through
invasion-related ligand-receptor interactions
To elucidate the spatial distribution and microenvironmental context of
DSS1-driven cells, we analyzed spatial transcriptomics data from 38
ccRCC specimens. Cell type deconvolution revealed the spatial landscape
of DSS1-driven and tumor microenvironment (TME) cell populations across
different tumor stages (Fig. [215]8a, Supplementary Fig. [216]8). Among
the endothelial subsets, microvascular endothelial cells were most
frequently co-localized with DSS1-driven cells within the same spatial
spots. In contrast, hypoxic, lymphatic, and inflammatory endothelial
cells showed less co-occurrence (Fig. [217]8b, c).
Fig. 8. Spatial characteristics and cell-cell communications in a DSS1-driven
invasive niche of advanced ccRCC.
[218]Fig. 8
[219]Open in a new tab
a Cell type deconvolution of DSS1-driven cells using spatial
transcriptome data from ccRCC tissues. b Spearman correlations of
proportions between DSS1-driven cells and tumor microenvironment (TME)
cells in spatial spots. c Correlations between the proportion of
DSS1-driven cells and signature scores (AddModuleScore) of endothelial
subsets in spots. Scores were rescaled (to a 0–1 range), then adjusted
by dividing by the summed weights of TME cell types within a spot. d
Re-clustering (single-cell RNA-seq) and annotation of endothelial cell
subsets. e Cell-cell communication analysis between DSS1-driven cells
(n = 1403 cells) and microvascular cells (n = 7916 cells, single-cell
RNA-seq). f Correlations between DSS1-driven cell weights and
coexpression strength of predicted ligand-receptor pairs in spots
(pT1/pT2: n = 8 distinct tumors, pT3: n = 27 [21 distinct tumors with 1
slide, and 3 distinct tumors with 2 slides each], pT4/IV: n = 2
distinct tumors available, error bar: mean ± standard error of mean).
Ligand-receptor pair coexpression (log-normalized counts) strength:
rescaled (to 0-1) product of ligand L and receptor R expression
(R1 + R2, if two receptors). g SPP1-integrin colocalizations in
late-stage sample slides. h Multiplex immunohistochemistry showing
expression and colocalization of ligand SPP1 from DSS1-driven cells
(DSS1^high and LC3^low, and pan-CK^+) and receptor ITGB1 of
microvascular cells (CD105^+) (n = 4 distinct patients; boxes: zoom-in
regions; scale bar: 20 μm). i Quantification of DSS1-driven cells and
colocalized puncta in ccRCC patient samples (n = 4 distinct patients).
j Proportions of SPP1-ITGB1 interaction within 25, 50, or 75 μm radius
of CD105 puncta. b, c Boxplot: Center line: median; box: 25th to 75th
percentiles; whiskers: ±1.5×interquartile range. a–c, f, g. n = 38 (32
distinct tumors with 1 section each, 3 distinct tumors with 2 sections
each). Source data are provided as a Source Data file.
The microvascular cells were identified in single-cell RNA-Seq data by
markers such as FLT1 (VEGFR1), NOTCH4, and ESM1, and were distinguished
from the broader endothelial compartment (Fig. [220]8d). To investigate
potential intercellular communication between DSS1-driven and
microvascular cells, we performed ligand-receptor interaction analysis
using single-cell RNA-seq data. Among the identified ligand-receptor
pairs, VEGF-VEGFR1/R2, SPP1-ITGA5/ITGB1, ANGPTL4-ITGA5/ITGB1, and
FN1/COL4A1/COL4A2-SDC4 were recurrently identified (Fig. [221]8e).
Coexpression of these ligand-receptor pairs was enriched in late-stage
lesions, whereas early-stage lesions exhibited weaker interactions in
spots (Fig. [222]8f). Notably, SPP1-ITGA5/ITGB1 interactions showed a
consistent positive correlation with DSS1-driven cell abundance in
advanced ccRCC (Fig. [223]8f). The SPP1-ITGB1 interaction, recurrently
co-occurring with VEGF-VEGFR1/R2 interactions, was enriched at the
tumor-normal interface (Fig. [224]8g, Supplementary Fig. [225]9a, b).
ITGB1, a central member of the integrin family, mediates
cell-extracellular matrix adhesion and intracellular signaling.
Compared to SPP1-ITGA5, SPP1-ITGB1 interaction demonstrated stronger
co-localization with DSS1-driven cells and higher coexpression
intensity at the tumor boundary (Fig. [226]8g). Multiplex
immunohistochemistry of ccRCC specimens confirmed that DSS1-driven
tumor cells (DSS1^hi, pan-CK^+) exhibited reduced autophagy (LC3^lo)
and were enriched within a vascularized invasive niche at the
tumor-stromal interface. These cells were in spatial proximity to
CD105^+ microvascular endothelial cells, with evidence of SPP1-ITGB1
interactions (Fig. [227]8h, boxes 1 and 2). With disease progression,
both the proportion of DSS1-driven cells and the frequency of
SPP1-ITGB1 interactions increased (n = 4 patients, Fig. [228]8i).
Spatial transcriptomic data revealed that spots enriched with
DSS1-driven cells exhibited higher EMT scores and lower autophagy
scores (Supplementary Fig. [229]9c). Consistently, multiplex
immunohistochemistry demonstrated that DSS1-driven cells with reduced
E-cadherin and elevated Vimentin levels were frequently located near
CD31-positive vascular endothelial cells at the tumor-stromal boundary.
This spatial arrangement suggests that the interaction between
DSS1-driven cells and vascular endothelial cells involves the induction
of EMT via DSS1 (n = 4 patients, Fig. [230]9a).
Fig. 9. EMT and SPP1 in DSS1-driven cells-related niche.
[231]Fig. 9
[232]Open in a new tab
a Multiplex immunohistochemistry showing spatial distribution of
DSS1-driven cells with a high epithelial-mesenchymal transition (EMT)
feature (E-cad^low, Vimentin^high) around vascular cells (CD31^+,
arrows, n = 4 distinct patients). Scale bar: 40 μm. b SPP1 RNA and SPP1
protein levels in tissues of diverse cancer types (kidney cancer: RNA,
n = 35 distinct cell lines, protein, n = 110 distinct samples) from the
Human Protein Atlas (HPA, [233]www.proteinatlas.org). c SPP1 Protein
levels in normal tissues from the HPA database (kidney, high expression
in proximal tubules, n = 3 slides). d Pseudo-bulk RNA levels of SPP1 in
epithelial and immune cell types. AvgExp, Average expression. e Protein
abundance of SPP1 between tumor and normal tissues in Clinical
Proteomic Tumor Analysis Consortium (CPTAC)-ccRCC dataset (n = 194
distinct samples, two-tailed Mann-Whitney U test). Boxplot: Center line
= median; box = 25th to 75th percentiles; whiskers = minima to maxima.
f Spearman correlation between SPP1 and DSS1 mRNA expression (log[2]
Count Per Million, two-tailed spearman’s rank correlation test). g
Immunoblotting showing SPP1 protein levels in ccRCC cells (shDSS1 vs.
shNC, n = 3 independent experiments, error bar: mean ± standard
deviation [SD], two-tailed Welch’s t-test). The samples derived from
the same experiment were run on parallel gels, with each gel probed for
a different antibody. h Spearman correlation between CD68 expression
and coexpression of SPP1-ITGB1 in spots of sample sections from either
tumor core or tumor-stromal interface (Li2022 spatial dataset, n = 13
sections from 10 distinct patients; error bar: mean ± SD). i Schematic
diagram (by Figdraw.com) illustrates the proposed mechanistic model of
DSS1-driven cells in ccRCC metastasis: DSS1 (oligomer) interacts with
pro-LC3B or LC3B-I, promoting LC3B degradation via E3 ubiquitin ligase
TRIM25-mediated Lys-63 (K63)-linked polyubiquitination at LC3B-K51,
leading to impaired macroautophagic flux and p62 accumulation, TWIST1
stabilization and increased TWIST1 nuclear transport, promoting EMT
activation. DSS1 highly expressed (DSS1^hi) tumor cells were increased
in late-stage tumors and linked to microvascular invasion within a
vascularized invasive niche at the tumor-stromal interface, mediated by
SPP1-ITGB1 interactions. Statistics are provided in the source data.
Source data are provided as a Source Data file.
SPP1 is a secreted phosphoprotein that modulates diverse forms of
cell-cell signaling. Although both normal and cancerous kidney tissues
exhibit high levels of SPP1 mRNA and SPP1 protein (Fig. [234]9b, c),
the protein level of SPP1 is slightly reduced in kidney cancer tissues
compared to normal tissues (Fig. [235]9e), with no sex-specific
differences observed in expression patterns (Supplementary
Fig. [236]9d). Correlation analysis revealed that DSS1 expression was
not significantly associated with SPP1 expression (Fig. [237]9f), a
finding consistent with immunoblot results showing that silencing DSS1
did not alter SPP1 protein levels in ccRCC cells compared to the
negative control (Fig. [238]9g). Notably, more than 50% of the positive
SPP1-ITGB1 interactions were located within a 50 µm radius of
CD105-positive microvascular puncta, and over 70% were within a 75 µm
radius (Fig. [239]8h, box 3; Fig. [240]8j), indicating that SPP1-ITGB1
interactions are spatially enriched around microvascular cells, despite
no overall increase in SPP1 expression at the bulk-tissue level
(Fig. [241]9e).
SPP1 expression was higher in DSS1-driven cells than in other cell
types, including SPP1-positive tumor-associated macrophages
(Fig. [242]9d). This raises the question of whether SPP1-secreting
macrophages contribute to DSS1-driven niches. Previous studies have
shown that SPP1-positive macrophages are enriched in hypoxic and
necrotic tumor regions^[243]18,[244]19, which are spatially distinct
from the areas enriched in DSS1-driven cells (Fig. [245]8c, h;
Fig. [246]9a) and from regions where SPP1-ITGB1 colocalization occurs
(Fig. [247]9h), both of which are associated with vascularized niches.
Furthermore, SPP1 puncta were detected in close proximity to
DSS1-driven cells, suggesting that in these vascularized niches, SPP1
is likely secreted primarily by DSS1-driven cells rather than
macrophages (Fig. [248]8h). Taken together, these observations suggest
that DSS1 enhances the invasive and migratory capabilities of ccRCC
cells by promoting local SPP1 expression within vascularized niches and
facilitating spatial interactions with microvascular endothelial cells,
thereby contributing to microvascular metastasis. This supports a
mechanistic model in which DSS1-driven invasive niches play a critical
role in ccRCC metastasis (Fig. [249]9i).
Discussion
Through integrative analyses of transcriptomic datasets,
patient-derived tissues, and animal models, we demonstrate that DSS1 is
upregulated in ccRCC and promotes metastatic progression.
Mechanistically, DSS1 binds to LC3 and facilitates its proteasomal
degradation via TRIM25-mediated Lys63 (K63)-linked polyubiquitination
at lysine 51 of LC3B. This degradation impairs macroautophagic flux,
resulting in the accumulation of p62 and stabilization of TWIST1,
thereby enhancing its nuclear translocation and activating EMT. We
further identified a distinct subset of ccRCC cells characterized by
high DSS1 expression. These DSS1-driven cells were enriched in
advanced-stage and metastatic tumors, as shown by single-cell
transcriptomic, bulk RNA-Seq, spatial transcriptomics, and multiplex
immunohistochemistry analyses. Notably, the frequency of DSS1-driven
cells was inversely correlated with patient survival, suggesting their
potential as a prognostic biomarker. Spatial profiling revealed that
DSS1-driven cells preferentially reside at the tumor-stroma interface,
where they engage in cell-cell communication with microvascular
endothelial cells. These interactions are mediated by
invasion-associated ligand-receptor pairs, including VEGF-VEGFRs and
SPP1-ITGB1/ITGA5. Such spatially restricted signaling likely
facilitates microvascular invasion and dissemination of tumor cells.
Together, our findings delineate a mechanistic model in which DSS1
suppresses autophagy and activates EMT, thereby promoting a
pro-metastatic cellular state in ccRCC. This model not only deepens our
understanding of ccRCC progression but also identifies DSS1-driven
cells as potential therapeutic targets for mitigating metastasis in
advanced disease.
Autophagy plays a critical role in maintaining cellular homeostasis in
renal tubular epithelial cells by eliminating damaged organelles and
senescent proteins^[250]49. In tumor cells, autophagy can theoretically
support survival by degrading intracellular components to generate
metabolic substrates. However, autophagy appears to be highly
niche-dependent in ccRCC. The tumor suppressor VHL contributes to
autophagic homeostasis by promoting the ubiquitin-mediated degradation
of LC3 through direct binding^[251]50. Loss of VHL function occurs in
approximately 60% to 80% of ccRCC cases^[252]51. In hypoxic niches, VHL
deficiency enables transient receptor potential melastatin 3 (TRPM3) to
activate autophagy by stimulating AMPK and ULK1 signaling, thereby
facilitating phagophore formation and supporting tumor cell
survival^[253]52. In contrast, within vascularized niches, ccRCC cells
secrete succinate, which induces autophagy in PDGFR-β⁺ GPR91⁺
pericytes. These pericytes, in turn, release metabolites that nourish
adjacent tumor cells^[254]53. Our findings suggest that DSS1 may
function as an alternative effector in the context of VHL loss. By
promoting the polyubiquitination and proteasomal degradation of LC3,
DSS1 suppresses autophagy while concurrently activating EMT, thereby
facilitating microvascular metastasis. The highly vascularized nature
of ccRCC may also help explain the limited clinical efficacy of
autophagy inhibitors. For instance, hydroxychloroquine demonstrated
minimal therapeutic benefit in a phase II trial, with only 2 of 33
patients achieving partial responses^[255]54. These observations are
consistent with our data and prior studies showing reduced autophagy
activity in ccRCC tissues compared with matched adjacent non-tumorous
tissues^[256]55. Furthermore, autophagy levels were found to be lower
in advanced or metastatic ccRCC relative to localized disease^[257]55.
This pattern correlates with the progressive upregulation of DSS1 and
the relatively low-hypoxia phenotype of DSS1-driven cells. Thus,
elucidating the molecular mechanisms underlying autophagy suppression
in ccRCC is crucial for understanding disease progression and
metastasis, and may inform the development of more effective
therapeutic strategies.
Autophagy comprises a series of tightly regulated steps, including
initiation, nucleation, elongation, maturation, and the autophagic
lysosome reformation (ALR) cycle^[258]30. mTOR serves as a critical
negative regulator by inhibiting autophagy initiation and promoting the
ALR cycle. In our study, DSS1 knockdown did not alter total or
phosphorylated mTOR (p-mTOR) protein levels. Instead, it attenuated the
ubiquitin-mediated degradation of LC3, suggesting that DSS1 inhibits
autophagic flux independently of mTOR signaling. Beclin 1, together
with VPS15 and VPS34, mediates phagophore nucleation. However, its
activity can be suppressed by Bcl-2 through direct interaction^[259]30.
Our findings suggest that DSS1 may impair autophagy by promoting LC3
ubiquitination and degradation as well as upregulating Bcl-2
expression, although the mechanism underlying Bcl-2 regulation remains
unclear. Therefore, DSS1-mediated autophagy suppression may represent a
distinct regulatory pathway, operating independently of both mTOR and
VHL signaling.
The loss of VHL function leads to the stabilization of
hypoxia-inducible factors (HIFs), which activate downstream angiogenic
programs^[260]51. This mechanism underpins the benefits of VEGFR
inhibitors in treatment. However, resistance to anti-VEGFR therapy
remains a major clinical challenge, and the molecular mechanisms
underlying this resistance remain unclear^[261]56. Recent studies have
identified ITGB1 as a key mediator of tumor angiogenesis and metastatic
niche formation^[262]57,[263]58. In our analysis, VEGF-VEGFR and
SPP1-ITGB1 ligand-receptor interactions frequently co-occurred within
tumor regions, particularly at invasive fronts. This spatial
co-localization suggests that SPP1-ITGB1 signaling may compensate for
suppressed VEGF signaling during anti-VEGFR treatment, thereby
maintaining the invasive phenotype of tumor cells. Furthermore, the
increased abundance of DSS1-driven cells in patients who had received
anti-VEGFR therapy supports the hypothesis that these cells contribute
to therapy resistance. Taken together, our data imply that DSS1-driven
cells, through their engagement in alternative angiogenic signaling
pathways such as SPP1-ITGB1, may play a critical role in the failure of
anti-angiogenic therapies in advanced ccRCC.
DSS1 is an intrinsically disordered protein that lacks a stable
tertiary structure under physiological conditions, posing substantial
challenges for the development of small-molecule inhibitors. Upstream
of DSS1, genomic copy number amplification is associated with its
overexpression across various cancers, including ccRCC. Previous
studies have also suggested that DSS1 expression may be linked to
elevated levels of ROS^[264]34. In addition, DSS1 levels are reduced
upon treatment with 12-O-tetradecanoylphorbol-13-acetate (TPA, also
known as PMA), a protein kinase C activator^[265]59, although the
underlying regulatory mechanism remains unclear. A comprehensive
investigation of the regulatory mechanisms driving DSS1 overexpression
in ccRCC may facilitate the identification of upstream therapeutic
targets. Downstream of DSS1, TWIST1, lacks enzymatic activity and
possesses a flexible structure, rendering it undruggable using
conventional approaches. Currently, no approved therapies directly
target TWIST1. Interestingly, DSS1 interacts with structured proteins
such as components of the proteasome, although proteasome complexes
lacking DSS1 remain functional in yeast^[266]60,[267]61. Given these
limitations, targeting DSS1 or TWIST1 through proteolysis-targeting
chimera (PROTAC) technology may offer a promising therapeutic strategy
for ccRCC. PROTACs have recently demonstrated substantial advantages in
degrading previously undruggable proteins^[268]62. Alternatively,
combination strategies that exploit vulnerabilities downstream of DSS1
may also be beneficial. For example, proteasome inhibitors such as
carfilzomib, when combined with autophagy agonists such as MLN0128, may
synergistically suppress DSS1-driven tumor progression by counteracting
the autophagy blockade induced by DSS1. However, the therapeutic
efficacy and safety of such combinations require rigorous validation
through preclinical models and clinical trials.
Methods
Ethical compliance
This study complied with all relevant ethical regulations. The human
research protocol was approved by the Ethics Committee of the First
Affiliated Hospital of Chongqing Medical University (Approval no.
2021-465), and the animal research protocol was approved by the Ethics
Committee of Chongqing Medical University.
Integrative analysis of DSS1 as a driver gene in ccRCC metastasis
* i)
Genes with positive expression and differentially expressed in
ccRCC
Differentially expressed genes were identified using data from
TCGA. Specifically, the TCGA-KIRC dataset, comprising 523 tumor
samples and 72 normal kidney tissue samples, was analyzed. The
HTSeq-count expression matrix was downloaded from the Genomic Data
Commons (GDC) data portal ([269]https://portal.gdc.cancer.gov/).
Differential expression analysis was performed using the
DESeq2^[270]63 package (version 1.46.0) in R. Genes with a
Benjamini-Hochberg adjusted P-value < 0.05 were considered
statistically significant. Basal gene expression levels in ccRCC
and normal kidney tissues were evaluated using both the TCGA-KIRC
dataset and the HPA database (version 19.3;
[271]http://www.proteinatlas.org)^[272]38. For mRNA-level analysis,
an average Fragments Per Kilobase of transcript per Million mapped
reads (FPKM) value > 1 in either tumor or normal tissues was used
to define gene expression abundance. Genes with low expression
abundance (FPKM ≤ 1 in both tissue types) were excluded from
further analysis. For protein-level evaluation,
immunohistochemistry data from the HPA database were used. Only
genes with protein expression detected in at least three
independent samples were considered for downstream analyses.
* ii)
Genes associated with patient survival and metastasis
First, univariate and multivariate Cox proportional hazards
regression analyses were performed to identify genes associated
with overall patient survival. Kaplan-Meier survival curves were
generated, and genes with log-rank P values < 0.05 in univariate
analysis were further evaluated by multivariate Cox regression.
Clinical covariates, including age, sex, tumor stage, grade, and
TNM classification, were included in the multivariate model to
assess whether gene expression was an independent prognostic
factor. Genes with P values < 0.05 after adjustment for these
clinical variables were considered statistically significant.
Second, a 154-gene signature related to cancer invasion and
metastasis was curated as a subset of Catalogue of Somatic
Mutations in Cancer (COSMIC) Cancer Gene Census (version 92,
[273]https://cancer.sanger.ac.uk/cosmic)^[274]64. Pearson
correlation analysis was conducted between the expression of these
154 genes and candidate prognostic genes. Gene pairs with an
absolute Pearson correlation coefficient (|r | ) > 0.4 and a false
discovery rate (FDR) < 0.001 were considered significantly
correlated. The COSMIC Cancer Gene Census was selected for this
analysis because it not only catalogs genetic mutations but also
provides expert-curated annotations linking genes to cancer
hallmarks such as invasion and metastasis. These annotations are
supported by strong experimental evidence, including knockdown,
knockout, and overexpression studies, which elucidate the
functional roles of these genes in cancer progression mechanisms.
Metastasis-related hub genes with high connectivity in the gene
regulatory network were further selected based on their degree of
interaction. Predicted regulatory interactions were considered
valid if supported by at least one of nine computational tools:
RNA22 v2.0^[275]65, miRcode v11^[276]66, TargetScan v7.2^[277]67,
PITA v6^[278]68, PicTar 2007^[279]69, miRDB v6.0^[280]70,
mirTarBase v2019^[281]71, miRanda v3.3a^[282]72, RNAhybrid
v2.1.2^[283]73. For putative interactions, only those with
P-values < 0.05 (when applicable) were retained. Additionally,
interactions directly supported by low-throughput experimental
evidence were included. Genes with a degree of connectivity ≥ 5
were defined as high-degree hub genes.
* iii)
Identification of candidate genes
To evaluate the concordance of gene differential expression in
ccRCC at the protein level, data from the HPA database were
analyzed. Only genes showing high concordance between mRNA and
protein expression levels were retained. For example, if
differential expression analysis identified a gene as upregulated
at the mRNA level, the corresponding immunohistochemistry data from
HPA tumor samples should show high or medium expression in at least
two-thirds of cases, while normal kidney samples should show no or
low expression in at least two-thirds of cases.
Subsequently, a comprehensive literature review was performed to
identify genes previously reported in ccRCC or cancer metastasis.
An independent gene signature analysis using Spearman correlation
was conducted to confirm associations between candidate genes and
metastasis-related signatures
(CHANG_CORE_SERUM_RESPONSE_UP^[284]20,
CSR_Activated_15701700^[285]21). A correlation coefficient |ρ | >
0.4 with P-value < 0.01 was considered statistically significant.
Next, independent dataset analyses were performed to validate gene
dysregulation in cancer metastasis. Specifically, GEO datasets
[286]GSE113204^[287]74 (n = 12) and [288]GSE72304^[289]75 (n = 14)
were analyzed using two-tailed Welch’s t-tests to assess
differential expression of candidate genes.
Functional and pathway analyses
First, pathway enrichment analysis based on hypergeometric tests was
performed using a DSS1-related gene set to identify candidate pathways.
Pathways with a Benjamini-Hochberg adjusted P-value < 0.05 were
considered statistically significant. Second, Gene Set Enrichment
Analysis (GSEA, version 4.3.3)^[290]76 was conducted with default
parameters to explore potential pathways associated with DSS1
expression. Samples were divided into high- and low-DSS1 expression
groups based on the median expression value of DSS1, and
DESeq2-normalized log[2] count per million (CPM) values were used.
Pathways were ranked by NES and nominal P values.
Identification of downstream transcription factors regulated by DSS1 through
autophagy
The Consensus Scoring of Genes in Cascade (CSGC) algorithm was
developed to identify candidate genes that potentially play a role in
the DSS1-mediated cascade associated with autophagy and metastasis.
EMT-inducing transcription factors (TF, i) ZEB1/2, SNAI1/2, TWIST1,
CTNNB1, FOXC1/2, TCF3, and KLF8 were considered as candidates. First,
these TFs should be upregulated in ccRCC or metastatic ccRCC because
they are considered EMT inducers. Fold change (FC) was calculated based
on DESeq2. Second, TFs should be correlated (Coef, Spearman) positively
with DSS1 and EMT, while correlated negatively with autophagy to be
logically consistent with the hypothesis that the TFs are regulated by
DSS1-suppressed autophagy. Log[2]CPM expression (DESeq2, TCGA-KIRC) was
used for calculation of pathway activities (ssGSEA^[291]77, MSigDB gene
sets v2022.1 [e.g., autophagy, EMT]) and Spearman correlation. This was
performed by introducing a logical variable (Sign, 1 or −1). Variable
Sign was assigned to 1 if the expected correlation was shown, otherwise
−1. A CSGC score of a candidate TF[i] was calculated as follows:
[MATH: CSGCi=(FCi1+FCi2)*
∑nj1Coefi,j<
/mrow>*Signi,j<
/mrow> :MATH]
1
Where j represents the activities of elements (n; j₁, j₂, j[n]) of the
DSS1-related cascade (DSS1, autophagy pathway, EMT process).
Proteomics-based analysis can, in theory, offer more accurate insights
than transcriptome-based analysis for identifying functional regulatory
cascades. The Clinical Proteomic Tumor Analysis Consortium (CPTAC)
project^[292]42 has quantified the protein levels of approximately
10,000 proteins in ccRCC tissues using trypsin-based mass spectrometry.
However, several proteins interested in this analysis were not detected
in the CPTAC-ccRCC cohort (trypsin-based mass spectrometric analysis),
such as SNAI1/2, TWIST1, FOXC2, KLF8, and DSS1. While transcriptomic
data do not fully account for post-transcriptional and
post-translational regulation, the RNA expression levels of key
EMT-related genes (e.g., TWIST1^[293]78) have been shown to correlate
with EMT activation^[294]42. Furthermore, it has been demonstrated
that, in general, mRNA levels positively correlate with protein
abundance across a broad range of genes in cells^[295]79. Therefore,
despite inherent limitations, transcriptome data offer a
high-throughput and broad-coverage platform that enables the
quantitative estimation of pathway activities, such as
autophagy^[296]80,[297]81, and have been widely used for hypothesis
generation and pathway modeling.
Analysis of mTOR and autophagy pathways
It is well established that the mTOR pathway negatively regulates
autophagy. GSEA using the TCGA-KIRC dataset (comparing DSS1-high vs.
DSS1-low, defined by the median expression level) showed negative NES
for both mTOR signaling and autophagy-related pathways. However, the
mTOR pathway exhibited a modestly negative NES and ranked relatively
low among all tested pathways. To evaluate whether this inverse
association between DSS1 expression and both mTOR/autophagy pathways is
consistently observed, we performed additional GSEA (v4.3.3) analyses
under the same settings (default parameters, DSS1-high vs. DSS1-low,
median as cut-off) across three independent ccRCC datasets. The
[298]GSE3538 dataset^[299]24 includes 177 human ccRCC tumor samples
across stage I–IV (n = 49, 29, 40, and 59, respectively).
Log[2]-transformed expression data from different microarray platforms
were merged and quantile-normalized. The [300]GSE251905 dataset^[301]82
contains RNA-seq data from 54 samples (32 primary and 22 metastatic
ccRCC tumors), and the [302]GSE254566 dataset^[303]83 includes RNA-seq
data from 91 ccRCC tumors. For RNA-seq datasets, expression profiles
(log[2]CPM, DESeq2) were downloaded.
Single-cell transcriptome data
Single-cell RNA sequencing (10x Genomics platform) data from ccRCC
tumor tissues and adjacent normal tissues were obtained from the Gene
Expression Omnibus (GEO) database, including [304]GSE159115
(n = 14)^[305]84, [306]GSE178481 (n = 26)^[307]85, and [308]GSE210038
(n = 9)^[309]86, comprising a total of 251,681 cells. These datasets
were integrated using the merge function in the Seurat package
(v4.3)^[310]87. High-quality cells were retained based on the following
filtering criteria: nFeature_RNA > 500, nCount_RNA > 1000, and
mitochondrial gene percentage (percent.mt) <20%. Cells with
nFeature_RNA > 7000 were identified as potential doublets and excluded.
Additionally, clusters containing cells expressing markers from
multiple lineages were flagged as doublets and removed^[311]14. The raw
count data were normalized, and 2000 highly variable genes were
selected. Batch effects among datasets were corrected using the FastMNN
algorithm from the SeuratWrappers package (v0.3.1,
[312]https://github.com/satijalab/seurat-wrappers). Dimensionality
reduction and clustering were conducted using the standard Seurat
pipeline, including RunUMAP and FindNeighbors, based on MNN-corrected
principal components (top 30 PCs). Cell clusters were identified using
the Louvain algorithm via the FindClusters function.
Cell type annotation and inferCNV analysis
Major cell types were annotated based on canonical marker genes.
Specifically, B cells were identified using BANK1 and MS4A1, plasma
cells by MZB1 and TNFRSF17, mast cells by KIT and VWA5A, myeloid cells
by CD68 and FCGR3A, fibroblasts by ACTA2 and MYL9, endothelial cells by
ENG and CLEC14A, epithelial cells by EPCAM and KRT8, and T/NK cells by
CD2, CD3, NKG7, and GZMK. Endothelial cell subtypes were further
delineated using distinct markers: inflammatory endothelial cells by
HLA-DRA and ICAM1, microvascular endothelial cells by FLT1, NOTCH4, and
ESM1, hypoxic endothelial cells by HIF1A and FGF2, and metabolically
altered endothelial cells by LDHA and PKM. Pericytes were identified
using RGS5 and ACTA2. To distinguish malignant epithelial cells, we
applied inferCNV (v1.10.0)^[313]35 using normal epithelial cells and
immune cells as reference controls. Copy number variation (CNV) scores
were estimated, scaled, and subsequently rescaled to a range between −1
and 1. The mean squared value of rescaled CNV estimates across the
genome was calculated as the CNV score. Cells with a CNV score > 0.04
were classified as malignant^[314]88.
High-dimensional weighted gene coexpression network analysis (hdWGCNA)
To investigate the functional programs within malignant cells,
high-dimensional weighted gene coexpression network analysis (hdWGCNA,
R package v0.3.00)^[315]89 was performed following the developer’s
guidelines. Briefly, metacells were constructed by grouping every 10
cells, with the maximum number of shared cells between any two
metacells set to 10. A soft-thresholding power was selected to
construct a signed coexpression network. Functional annotations of each
module were determined using Gene Ontology (GO, July 2024 release)
enrichment analysis. Module signature scores for individual cells were
calculated using the AddModuleScore function in Seurat, based on the
expression of all genes within each module.
Identification of DSS1-driven cells and analysis of the clinical value
DSS1-highly expressed ccRCC cells (referred to as DSS1-driven cells)
were defined as those exhibiting expression levels at least 50% higher
than the mean expression of normal epithelial cells, comprising the top
12% of malignant cells. This threshold yielded an average log₂ fold
change (log₂FC) of 2.08 between DSS1-driven cells and all other
malignant cells, exceeding the empirical four-fold cutoff used by the
HPA (v24.0)^[316]38 to classify elevated expression. To assess the
clinical relevance of DSS1-driven cells using bulk RNA-Seq data,
signature scores were computed as the mean scaled expression of marker
genes identified via the FindMarkers function in Seurat. Comparisons of
signature scores across tumor stages and molecular subtypes were
performed using the two-tailed Mann-Whitney U test. The prognostic
significance of the signature score in ccRCC patients was evaluated via
survival analysis (log-rank test) based on the TCGA-KIRC cohort.
Associations between DSS1-driven cell signature scores and anti-VEGFR
treatment efficacy were analyzed using the Mann-Whitney U test across
multiple datasets, including [317]GSE65615^[318]44,
[319]GSE64052^[320]46, and [321]GSE76068^[322]45.
Analysis of potential drugs or compounds targeting DSS1-driven cells
To identify potential drugs that may effectively inhibit ccRCC
metastasis, we analyzed the single-cell RNA sequencing (scRNA-seq) data
of DSS1-driven cells using two computational tools: scTherapy^[323]48
and beyondcell^[324]47. Both tools were employed to predict responses
to FDA-approved drugs, preclinical candidates, and laboratory
compounds. scTherapy^[325]48 (v1.0.0) is a methodological framework
that integrates scRNA-seq data with therapeutic response prediction by
identifying cell-type- or cell-population-specific drug targets and
biomarkers. Differentially expressed genes between DSS1-driven cells
and other malignant cells were identified using the FindMarkers
function in Seurat. Genes with |avg_log₂FC | > 0.5 and adjusted
P-values < 0.05 (Bonferroni correction) were used as input for drug
response prediction. All other parameters were set to default.
Beyondcell^[326]47 is a computational tool designed to predict drug
sensitivity at the single-cell level by integrating scRNA-seq data with
pharmacogenomic resources such as the NIH LINCS project. We applied
beyondcell to DSS1-driven cells to compute a switch point score
(ranging from 0 to 1) for each cell in response to each drug. A lower
switch point indicates higher predicted drug sensitivity. Following the
developer’s guidelines, drugs with switch point values < 0.1 were
considered potentially effective against DSS1-driven cells.
Spatial transcriptome analysis
Spatial transcriptome data of ccRCC tumor tissues were obtained from
three datasets: [327]GSE175540 (n = 2)^[328]90, [329]GSE210041
(original n = 24)^[330]86, and Li2022 dataset (original
n = 16)^[331]91. Li2022 dataset is available on the CELLxGENE
platform^[332]92. Samples lacking clinical staging information (n = 1
from [333]GSE210041) or determined to be of low quality according to
the original publication (n = 3 from Li2022) were excluded, resulting
in a final dataset of 38 spatial transcriptomics samples. Cell type
deconvolution was performed using the RCTD algorithm, implemented in
the spacexr (v2.2.1) package^[334]93, with our annotated single-cell
RNA-seq dataset as the reference. To analyze spatial relationships
between DSS1-driven cells and other tumor microenvironment (TME) cell
types, the raw weights assigned by RCTD were normalized: for each spot,
the weight of a given TME cell type was divided by the sum of all TME
cell weights within that spot. Similarly, to assess the correlation
between DSS1-driven cell abundance and functional states of endothelial
subsets, signature scores for endothelial subtypes were computed using
AddModuleScore, rescaled to [0–1], and divided by the summed TME cell
weights in each spot. Endothelial subtypes were defined by curated
marker genes. Microvascular endothelial cells were defined by PECAM1,
KDR, VCAM1, NOS3, PROCK, ANGPT2, TEK, ITGAL, ESM1, FLT1, NOTCH4, and
TNFRSF1A. Lymphatic endothelial cells were defined by PROX1, LYNE1,
FLT4, and PDPN. Hypoxia endothelial cells were defined by HIF1A.
Inflammatory endothelial cells were defined by SELE, ICAM1, VCAM1, and
IL6. To define tumor-normal boundaries, spots were first categorized as
tumor-cell dominant if more than 50% of their total RCTD-derived cell
weights were from malignant cells. All other spots were classified as
non-tumor-cell dominant. To further identify boundary spots, Euclidean
distances between spatial coordinates of each non-tumor-cell dominant
spot and its nearest tumor-cell dominant spot were calculated.
Non-tumor-cell dominant spots located within a distance of less than 2
units from a tumor-cell dominant spot were defined as tumor-normal
boundary spots.
Cell-cell communication analysis
Cell-cell communication between DSS1-driven cells and microvascular
endothelial cells was analyzed using CellChat (v2.1.2)^[335]94 based on
ligand-receptor pairs curated in CellChatDB v2^[336]94. The TriMean
method was applied to compute communication probabilities, and
interactions involving fewer than 10 cells were excluded to ensure
robustness. To validate these ligand-receptor interactions in spatial
transcriptomics data, a coexpression strength metric was calculated for
each spot by computing the product of the ligand expression and the sum
of receptor expressions, followed by rescaling to a 0–1 range. To
assess the relationship between the abundance of DSS1-driven cells and
the strength of ligand-receptor coexpression, Spearman correlation
coefficients were calculated across spatial spots for each individual
sample (slide).
Human tissues and ethics approval
Formalin-fixed and paraffin-embedded (FFPE) tissue samples from
patients with ccRCC were randomly selected from the pathology archives
(November 2019 to June 2022) of the First Affiliated Hospital of
Chongqing Medical University using a random number generator.
Eligibility criteria included: (a) histologically confirmed diagnosis
of ccRCC; (b) complete clinical information, including sex (biological
attribute), age, and pathological grading or staging; and (c)
sufficient residual tissue from both tumor and adjacent normal regions.
No subjective criteria were applied for sample exclusion. In total,
FFPE sections from 74 patients (47 males and 27 females; median age: 63
years) were collected, including both tumor and paired adjacent normal
kidney tissues (left/right kidney). Among these, 46 patients were
diagnosed with World Health Organization/International Society of
Urological Pathology (WHO/ISUP) grade 1 or 2, and 28 patients with
grade 3 or 4. All tissue sections were stored at 4 °C prior to
analysis. Written informed consent was obtained from all participants,
and the study protocol was approved by the Ethics Committee of the
First Affiliated Hospital of Chongqing Medical University (Approval No.
2021-465). Patient identities were anonymized prior to any downstream
analysis.
Immunohistochemistry (IHC) and H&E staining
FFPE human and murine tissue sections were processed following standard
protocols. Briefly, sections were deparaffinized in xylene, rehydrated
through graded ethanol series, and subjected to heat-induced epitope
retrieval in 10 mM sodium citrate buffer (pH 6.0) using a microwave for
15 min. Endogenous peroxidase activity was quenched with 3% hydrogen
peroxide for 10 min at room temperature (RT), followed by blocking in
10% goat serum for 1 hour at RT. Sections were then incubated overnight
at 4 °C with primary antibodies diluted in antibody diluent. The
following antibodies were used: DSS1 (Proteintech, 13639-1-AP, 1:100),
LC3 (Proteintech, 14600-1-AP, 1:100), SQSTM1/p62 (Proteintech,
18420-1-AP, 1:100), TWIST1 (Proteintech, 25465-1-AP, 1:100), TRIM25
(Proteintech, 12573-1-AP, 1:100), and CDH1/E-cadherin (Cell Signaling
Technology, 3195, 1:100). After washing, sections were incubated with a
biotinylated secondary antibody at 37 °C for 30 min and developed with
DAB (Absin, abs9210) according to the manufacturer’s instructions.
Nuclear counterstaining was performed using hematoxylin. The IHC assay
was performed using the rabbit-enhanced polymer detection kit
(ZSGB-BIO, PV-9001). IHC staining was visualized and imaged using a
bright-field microscope (Nikon or equivalent). Tumor cell content and
staining intensity were independently evaluated by two pathologists
blinded to clinical information. DSS1 IHC was performed in 74 matched
tumor and adjacent normal tissue pairs, and TRIM25 IHC was performed in
66 pairs with sufficient tissue availability. H&E staining was
conducted using a standard hematoxylin and eosin kit (ServiceBio,
G1076-500ML) for morphological assessment.
Multiplex immunohistochemistry (mIHC)
FFPE tissue sections from four patients with ccRCC were used for
multiplex immunohistochemistry (mIHC) analysis. Slides were incubated
overnight at 4 °C with primary antibodies against DSS1 (Proteintech,
13639-1-AP; 1:200), LC3 (Proteintech, 14600-1-AP; 1:200), SPP1 (Abcam,
ab214050; 1:200), ITGB1 (Abcam, ab30394; 1:200), pan-Cytokeratin
(pan-CK, Abcam, ab7753; 1:200), and CD105 (Abcam, ab231774; 1:200). To
analyze the autophagic status of DSS1-driven cells in tissue, FFPE
sections from an additional four ccRCC patients were processed using
the same protocol, with additional primary antibodies against CD31
(Proteintech, 66065-2-Ig; 1:200), E-cadherin (Cell Signaling
Technology, 3195; 1:200), and Vimentin (Cell Signaling Technology,
5741; 1:200). Signal amplification was performed using a tyramide
signal amplification kit (RecordBio Co. Ltd, RC0086Plus-67RM) according
to the manufacturer’s instructions.
Cell lines and cell culture
The 786-O (CRL-1932), ACHN (CRL-1611), Caki-1 (HTB-46), and HEK293T
(CRL-11268) cells were obtained from the American Type Culture
Collection (ATCC). Cell line authentication was performed by Short
tandem repeat (STR) profiling at the time of purchase. Cells were
passaged at ~80% confluency using 0.25% trypsin-EDTA (Gibco) and
monitored daily for morphological consistency. Mycoplasma contamination
was not detected in cells. The 786-O cells were cultured in Roswell
Park Memorial Institute (RPMI)−1640 medium, ACHN cells in Minimum
Essential Medium (MEM) medium, Caki-1 in McCoy’s 5 A medium, and
HEK293T cells in Dulbecco’s Modified Eagle Medium (DMEM). All media
were supplemented with 10% fetal bovine serum (FBS; ExCell, South
America origin) and 1% penicillin-streptomycin (Gibco). Cells were
maintained at 37 °C in a humidified incubator with 5% CO₂.
Cell morphology
Cell morphology was quantified by calculating the aspect ratio, defined
as the ratio of the major axis length (longest diameter) to the minor
axis length (shortest perpendicular diameter). Measurements were
performed using the Fit Ellipse function in ImageJ software (v1.53t,
NIH). Independent experiments were performed in triplicate.
Inhibitors
Chloroquine (CQ, HY-17589A, 10 μM), MG-132 (HY-13259, 20 μM), and
Cycloheximide (CHX, 50 μg/mL) were purchased from MedChemExpress® and
administered to cultured cells at the indicated concentrations
following the manufacturer’s instructions.
RNA extraction and reverse transcription-quantitative (RT-q) PCR
Total RNA was extracted from cells using TRIzol^® reagent (Invitrogen,
Thermo Fisher Scientific) and 500 ng RNA was reverse transcribed into a
20 µl final volume of cDNA using Reverse Transcription kit (Takara,
RR092A), according to the manufacturer’s instructions. RNA purity was
assessed by A260/A280 ratio (Nanodrop 2000, Thermo Fisher), and
integrity was confirmed via agarose gel electrophoresis (28S/18S ratio
>1.8). MAP1LC3B and SQSTM1 RNA expression were measured using qPCR on
the QuantStudio Real-Time PCR system (Thermo Fisher Scientific). qPCR
was conducted using TB Green^® Premix Ex Taq™ II (Takara, RR820A) and
analyzed using the 2^-ΔΔCt method. qPCR reaction conditions were as
follows: Initial denaturation at 95 °C for 2 min, followed by 39 cycles
at 95 °C for 15 s and 60 °C for 30 seconds. GAPDH was used as
endogenous controls.
Additionally, GSEA analyses showed a consistent negative correlation
between DSS1 and KEGG (v113.0) autophagy, with a big number of
autophagy genes (Core Enrichment gene set) negatively correlated with
DSS1 expression. To validate the mRNA-level expression correlation
between DSS1 and autophagy-related genes, we conducted RT-qPCR to
evaluate the mRNA expression of autophagy-related genes (Core
Enrichment gene set) between DSS1-knockdown cells and control cells.
Specifically, the genes identified as Core Enrichment genes by at least
3 of 4 datasets (TCGA-KIRC, [337]GSE3538^[338]24,
[339]GSE251905^[340]23, and [341]GSE254566^[342]25) were obtained (37
genes) and were analyzed using RT-qPCR. Primer specificity was
confirmed by a single peak in melt curve. This experiment was
biologically replicated 3 times, with 3 technical replicates for each
sample. Primer sequences were provided in Supplementary Table [343]1.
Western blot
Total proteins were extracted using RIPA lysis buffer supplemented with
PMSF, phosphatase, and protein inhibitors, following the Whole Cell
Lysis Assay protocol (KeyGEN Biotech, KGP2100). The protein
concentration was determined using a BCA reagent kit (Beyotime,
P0010S). The proteins were separated by SDS-PAGE and then transferred
to PVDF membranes. After blocking in 5% bovine serum albumin or 5%
nonfat milk at RT for 2 h, the membranes were incubated with primary
antibodies against DSS1 (Proteintech, 13639-1-AP, 1:1000),
CDH1/E-cadherin (Proteintech, 20874-1-AP, 1:1000), CDH2/N-cadherin
(Proteintech, 22018-1-AP, 1:1000), VIM/Vimentin (Proteintech,
10366-1-AP, 1:1000), LC3 (Proteintech, 14600-1-AP, 1:1000), SQSTM1/p62
(Proteintech, 18420-1-AP, 1:1000), TWIST1 (Proteintech, 25465-1-AP,
1:1000), TRIM25 (Proteintech, 12573-1-AP, 1:1000), ubiquitin (Abcam,
ab134953, 1:1000), Myc-tag (Proteintech, 16286-1-AP, 1:1000), HA-tag
(Proteintech, 51064-2-AP, 1:1000), mTOR (Abways, CY5306, 1:1000),
p-mTOR (Abways, CY5996, 1:1000), SPP1 (Abcam, ab214050, 1:1000), Bcl-2
(Abways, CY5032, 1:1000), Beclin 1 (HUABIO, HA721216, 1:1000), TGF-β
(Immunoway, YM8257, 1:1000), ZEB1 (Cell Signaling Technology, 70512,
1:1000), ZEB2 (Cell Signaling Technology, 97885, 1:1000), SNAI1
(Proteintech, 13099-1-AP, 1:1000), SNAI2/SLUG (Cell Signaling
Technology, 9585, 1:1000), TCF3 (Proteintech, 21242-1-AP, 1:1000),
FOXC2 (Proteintech, 23066-1-AP, 1:1000), GST-tag (Proteintech,
66001-2-Ig, 1:1000) and ACTB/β-actin (Proteintech, 20536-1-AP, 1:1000)
at 4 °C overnight. Subsequently, the membranes were incubated with an
HRP-conjugated secondary antibody (Proteintech, SA00001; or Abbkine,
IPKine™ A25022 light chain specific; 1:5000 dilution) for 1 h at RT and
detected using the ECL system. Independent experiments were performed
in triplicate.
Plasmids, small interfering RNA (siRNA), and cell transfection
Small interfering RNAs (siRNAs) targeting DSS1, TRIM25, and TWIST1 were
purchased from GenePharma Co., Ltd. Plasmids encoding FLAG-DSS1
(N-terminal), HA-DSS1^W27GW39GW43GF52A (N-terminal), DSS1, TRIM25,
TWIST1, LC3B, Myc-LC3B (N-terminal), Myc-LC3B mutants (K5R-K8R,
K30R-K39R, K42R-K48R, K51R, K65R, K103R, LC3B-K122R, all N-terminal),
HA-Ub (wildtype), HA-Ub-K48R, HA-Ub-K63R (N-terminal) were constructed
using the pcDNA3.1 vector. Empty vector (pcDNA3.1) was obtained from
Tsingke Biotech Co., Ltd. For the Glutathione-S-Transferase (GST)
pull-down assay, the full-length sequence of DSS1,
DSS1^W27GW39GW43GF52A, and TRIM25 were cloned into a pGEX-4T-1 vector
(Tsingke) containing an N-terminal GST tag. The empty pGEX-4T-1 vector
(Tsingke) was used as a control. Plasmid transfection was performed
using Lipofectamine™ 2000 (Invitrogen) according to the manufacturer’s
protocol. siRNAs were transfected into cells using GP-Transfect-Mate
(GenePharma, [344]G04009). The sequences of the human siRNAs are as
follows:
DSS1 siRNA: 5’- GACAAUGUAGAGGAUGACUUCUCUA-3’.
TWIST1 siRNA: 5’- CCUGAGCAACAGCGAGGAATT-3’.
TRIM25 siRNA-1: 5’- AUGGAUUUUCUCUAAGAGGAA-3’.
TRIM25 siRNA-2: 5’- UAUGGAUUUUCUCUAAGAGGA-3’.
Control siRNA (siNC): 5’- GCAGGCGAUUCAGAUCUGGUGCUUA-3’.
Lentivirus and short hairpin RNA (shRNA)
The lentiviral overexpression construct for DSS1 (vector GV492:
Ubi-MCS-3FLAG-CBh-gcGFP-IRES-puromycin) and the lentiviral shRNA
knockdown construct for DSS1 (vector GV493:
hU6-MCS-CBh-gcGFP-IRES-puromycin) were generated by Genechem Co., Ltd.
Lentivirus infection was conducted according to the manufacturer’s
instructions. Seventy-two hours post-infection, 786-O, ACHN, and Caki-1
cells expressing green fluorescent protein (GFP) were selected. Stable
DSS1-overexpressing or DSS1-knockdown cell lines were established by
culturing in medium containing 2 μg/mL puromycin. The efficiency of
DSS1 overexpression or knockdown was confirmed by western blot
analysis. Target sequences for ShDSS1: ShDSS1 #1:
5’-GGTAGACTTAGGTCTGTTAGA-3’; ShDSS1 #2: 5’-GAGTTTGAAGAGTTCCCTGCC-3’;
ShDSS1 #3: 5’-GACGACGAGTTTGAAGAGTTC-3’. Target sequence for shNC:
5’-TTCTCCGAACGTGTCACGT-3’.
Cell proliferation assay
Cell proliferation was assessed using the Cell Counting Kit-8 (CCK-8;
Yeasen Biotechnology, 40203ES60). Briefly, 2000 cells were seeded per
well in 96-well plates. At indicated time points, 10 μL of CCK-8
reagent was added to each well and incubated at 37 °C for 1 hour in the
dark. Absorbance was measured at 450 nm using a microplate reader every
24 h. Additionally, proliferation was evaluated by
5-ethynyl-2′-deoxyuridine (EdU) incorporation using the BeyoClick™ EdU
Cell Proliferation Kit (Beyotime, C0071S) according to the
manufacturer’s protocol. Cells were stained with Alexa Fluor
488-conjugated dye at RT for 30 min in the dark and imaged by
fluorescence microscopy (Nikon Eclipse Ts2R). All experiments were
performed in biological quintuplicates (CCK-8) or triplicates (EdU
assay). CCK-8 experiment was biologically replicated 5 times, with 3
technical replicates for each sample. EdU experiment was biologically
repeated 3 times.
Transwell assay
8 μm-pore transwell chambers (BIOFIL, TCS020024) coated with Matrigel
(BD Bioscience) were used for cell invasion assay, while those without
any pre-treatment were for cell migration assay. 2 × 10^4 (migration)
and 5 × 10^4 (invasion) suspended transfected cells in 200 μL FBS-free
medium were loaded into the upper chamber of each 24-well transwell
chamber, while 600 μL 10%-FBS medium was added into the lower chamber.
After culture for 48 h at 37 °C with 5% CO[2], the cells on the lower
surface were fixed with methanol for 30 min and stained with a 0.5%
crystal violet solution for 15 min after the non-adhering cells in the
upper chamber were scraped. Images were captured using an inverted
microscope and analyzed using ImageJ software. Independent experiments
were performed in triplicate.
Co-immunoprecipitation (Co-IP)
The total proteins (5% of proteins for input) were incubated with
primary antibodies against Myc-tag (Proteintech, 16286-1-AP, 1:50),
FLAG-tag (Proteintech, 20543-1-AP, 1:50), LC3 (Proteintech, 14600-1-AP,
1:50), TWIST1 (Proteintech, 25465-1-AP, 1:50), TRIM25 (Proteintech,
12573-1-AP, 1:50), and IgG (Abcam, ab172730, 1:50), followed by
precipitation with Protein A-Agarose beads (Santa Cruz, sc-2001). After
washing the beads with wash buffer (Beyotime Biotechnology, P0013) and
retrieving the proteins, they were subjected to western blot analysis.
Independent experiments of co-IP were performed in triplicate.
GST pull-down
The recombinant plasmids of GST-DSS1, GST-DSS1^W27GW39GW43GF52A,
GST-TRIM25, and pGEX-4T-1 vector were transfected into E. coli BL21
(DE3), followed by the selection of monoclonal colonies for extended
culture. Expression of recombinant plasmids were induced by addition of
0.5 mM IPTG and incubation at 25 °C for 6 h. 1 mg of fusion protein was
immobilized in 50 µL of glutathione agarose and incubated at 4 °C for
4 h with gentle rocking. Purified human LC3B protein was purchased from
MedChemExpress^® (HY-[345]P70909). 50 ng of LC3B fusion proteins was
added to the immobilized GST-DSS1, GST-DSS1^W27GW39GW43GF52A,
GST-TRIM25, and GST-tag and then incubated overnight at 4 °C with
gentle rotation. Following this, the complex of agarose and bound
proteins was washed, and the bound proteins were eluted for
immunoblotting. This experiment was conducted using a GST pull-down kit
(FITGENE, FI88808). Independent experiments were performed in
triplicate.
DSS1 adducts analysis
Cells were lysed using Cell Lysis Solution (Beyotime Biotechnology,
P0013J) supplemented with Protease Inhibitor Cocktail (Beyotime
Biotechnology, P1005), with or without 10 mM EDTA. Lysates were then
heated at 95 °C for 5 min in a dry bath incubator prior to immunoblot
analysis. The recombinant DSS1^W27GW39GW43GF52A plasmid was also
utilized to assess the formation of DSS1 adducts.
Immunofluorescence
Cells were transfected with siDSS1 or siNC and cultured for 48 h at
37 °C, then seeded onto glass slides in 24-well plates. After an
additional 36 h of culture at 37 °C, cells were fixed with 4%
paraformaldehyde for 30 min, permeabilized with 0.5% Triton X-100 for
20 min, and blocked with 10% goat serum for 1 hour at room temperature.
Slides were incubated overnight at 4 °C with primary antibodies against
LAMP1 (Abcam, ab25630, 1:100) and LC3 (Proteintech, 14600-1-AP, 1:100).
Following washes, slides were incubated for 1 h at RT with DyLight
488-conjugated goat anti-mouse secondary antibody (for LAMP1) and
DyLight 549-conjugated goat anti-rabbit secondary antibody (for LC3).
Nuclei were counterstained with DAPI (BOSTER Biological Technology,
AR1176) for 10 min in the dark at RT. For multiplex immunofluorescence,
primary antibody against DSS1 (Proteintech, 13639-1-AP, 1:200), LC3
(ServiceBio, GB11124, 1:500), PSMD3 (Proteintech, 12054-1-AP, 1:200),
Bcl-2 (ServiceBio, [346]GB154830, 1:500), TWIST1 (Proteintech,
25465-1-AP, 1:200), and p62 (Proteintech, 18420-1-AP, 1:200) were
applied. Signal amplification was performed using the Tyramide Signal
Amplification Kit (RecordBio Co. Ltd, RC0086-34RM) according to the
manufacturer’s instructions. Imaging was conducted on a Leica SP8
confocal microscope with LAS X software (v3.5.7) or scanned using an
automated slide scanner (3DHISTECH, Pannoramic MIDI). Independent
experiments of immunofluorescence were performed in triplicate.
Liquid Chromatography-Tandem Mass Spectrometry (LC-MS/MS)
Protein samples (Caki-1) separated by SDS-PAGE were excised as gel
bands and subjected to in-gel digestion. Gel pieces were destained with
alternating washes of 50% acetonitrile (ACN) in 25 mM ammonium
bicarbonate (NH[4]HCO[3]) until complete decolorization, followed by
dehydration in 100% ACN. Proteins were reduced with 10 mM
dithiothreitol (DTT) at 57 °C for 60 min and alkylated with 55 mM
iodoacetamide (IAA) in darkness for 45 min. After sequential washes
with NH[4]HCO[3] and ACN, tryptic digestion was performed using
sequencing-grade trypsin (Promega) at 37 °C overnight
(enzyme-to-substrate ratio 1:50). Peptides were extracted using 50%
ACN/0.1% formic acid (FA) with ultrasonic assistance, followed by
vacuum concentration. Desalting was performed using Pierce C18 Spin
Tips (Thermo Scientific) according to manufacturer’s protocol. Purified
peptides were reconstituted in 0.1% FA and analyzed on a Q-Exactive HF
mass spectrometer (Thermo Scientific) coupled with an EASY-nLC 1200
nanoflow HPLC system. Chromatographic separation was achieved using a
C18 analytical column (75 μm × 15 cm, 1.9 μm particles) with a
300 nL/min gradient from 2% to 35% ACN in 0.1% FA over 60 min. MS
acquisition employed data-dependent acquisition (DDA) mode. Raw data
were processed using Proteome Discoverer 2.4 (Thermo Scientific)
against the SwissProt Homo sapiens database (release 2023_03) with the
following search parameters: tryptic digestion (max 2 missed
cleavages), carbamidomethylation (C) as fixed modification, oxidation
(M) and N-terminal acetylation as variable modifications, 10 ppm
precursor mass tolerance, and 0.02 Da fragment mass tolerance. FDR was
controlled at ≤ 0.01. Independent experiments were performed in
triplicate.
Transmission electron microscope
The sample was pre-fixed with 3% glutaraldehyde at 4 °C overnight,
avoiding light, and subsequently fixed with 1% osmium tetroxide.
Following fixation, the samples underwent dehydration in a graded
series of acetone solutions. Next, the samples were treated with a
mixture of embedding agent (SPI-Pon™ 812) and acetone (at ratios of
1:3, 1:1, and 3:1 by volume, for 1 h each), followed by embedding with
the embedding agent (SPI-Pon™ 812). Sectioning of the samples was
conducted using a Leica UC7rt (at thicknesses of 60–90 nm). After
staining with uranyl acetate for 15 min and lead citrate for 2 min, the
samples on copper grids were imaged using a transmission electron
microscope JEM-1400FLASH (JEOL, Japan). Two pathologists engaged in
identifying and labeling cellular structures, such as autophagosomes.
Independent experiments were performed in triplicate.
Ubiquitination assay
HEK293T cells were selected for ubiquitination assays due to their high
transfection efficiency, low background ubiquitin-proteasome activity,
and broad compatibility with established protocols. Cells were
transfected with Myc-LC3B and HA-Ub, along with control siRNA (siNC) or
DSS1-targeting siRNA (siDSS1), and cultured for 48 h at 37 °C. Prior to
harvest, cells were treated with either 0.02% DMSO (vehicle control) or
20 μM MG-132 (MedChemExpress) for 6 h at 37 °C to inhibit proteasomal
degradation. Cells expressing Myc-LC3B were lysed on ice for 20 min in
Cell Complete Lysis Buffer (Beyotime, P0037) supplemented with a
protease inhibitor cocktail (Beyotime, P1005). Lysates were clarified
by centrifugation at 14,000 g for 15 min at 4 °C. The supernatant was
incubated with anti-Myc magnetic beads at 4 °C for 12 h to
immunoprecipitate Myc-LC3B. Beads were then washed, and bound proteins
were eluted for immunoblot analysis. In vitro ubiquitination analysis
was performed based on E2 Select Ubiquitin Conjugation Kit (Yeasen
Biotechnology, 20440ES10) according to the manufacturer’s instructions.
20 µL reaction system (2 µL 10× MgATP, 4 µL 5× Ubiquitin, 2 µL 10×
His-tagged E1 enzyme UBE1, 1 µL 20× E2 enzyme UbcH5a/UbcH5c, 0.5 µg E3
enzyme GST-TRIM25, 0.5 µg substrate LC3B [MedChemExpress,
HY-[347]P70909], 2 µL 10× Reaction Buffer, and ddH[2]O) was prepared
and incubated for 4 h at 37 °C. Samples were analyzed by
immunoblotting. Independent experiments were performed in triplicate.
Xenograft model
Male BALB/c-nu nude mice (4 weeks old) were obtained from HuaFuKang
Bioscience (Beijing, China) and housed under pathogen-free conditions
in individually ventilated cages (IVCs) with autoclaved corncob
bedding. Mice were maintained on a 12 hour light/dark cycle with free
access to food and water. Environmental conditions were controlled at
22 ± 1 °C temperature and 55 ± 5% relative humidity, with HEPA-filtered
air ventilation providing ≥15 air changes per hour. Given the higher
incidence of ccRCC in males and the absence of observed sex-related
differences in DSS1 effects in patients, male mice were used
exclusively. Inclusion criteria comprised all healthy animals, while
exclusion criteria included procedural errors or unexpected mortality;
no animals were excluded, and all completed the study. Body weights
were recorded weekly.
For the experimental metastasis model, 1 × 10^6 Caki-1 or 786-O cells
stably expressing lentiviral constructs (Lv-shDSS1, Lv-shNC, Lv-DSS1,
or Lv-Vector) were injected into the lateral tail vein of mice (n = 6
per group; total n = 24), randomized prior to injection. Mice were
monitored for up to seven weeks and euthanized by cervical dislocation
unless humane endpoints were met earlier (> 20% body weight loss from
baseline or signs of distress). Lung tissues were harvested for
evaluation of metastatic foci and histopathological analyses.
For the subcutaneous tumor model, 1 × 10^6 Caki-1 cells stably
transduced with Lv-shDSS1 or Lv-shNC (n = 6 per group; total n = 12)
were injected into the axillary fossa. Tumor dimensions (length, width,
height) were measured weekly using a vernier caliper, and tumor volume
was calculated as π/6 × (L × W × H). Animals were euthanized four weeks
post-inoculation or earlier if humane endpoints were reached (tumor
volume ≥ 1500 mm³, ulceration, or > 20% body weight loss). No animals
required early euthanasia.
All experimental data were analyzed without subjective exclusions.
Investigators performing experiments and outcome assessments were
blinded to group allocation, and data analysts remained blinded until
completion of statistical analyses. One sample per group was excluded
from immunohistochemical analysis due to tissue processing damage. All
animal procedures were approved by the Ethics Committee of Chongqing
Medical University. Throughout the study, tumor sizes remained within
institutional ethical limits (maximum diameter 1.8 cm; volume
1800 mm³), ensuring compliance with all relevant ethical guidelines.
Statistical analysis
Omics data of ccRCC patients were analyzed using R 3.6.1 software
([348]https://www.r-project.org/). Images were evaluated using ImageJ
software (v1.53t, [349]https://imagej.nih.gov/ij/). Experimental data
were analyzed using Prism 10.4 software (GraphPad Prism Software Inc.).
A two-tailed Welch’s t-test (effect size: Welch-corrected Cohen’s d, or
Hedges’ g if n < 20) or Mann-Whitney U test P < 0.05 (effect size: Δ
with 95% confidence interval) was considered statistically significant.
Descriptive mean difference (MD) with 95% confidence interval was also
calculated. Benjamini-Hochberg adjustment for P values was performed
for multiple tests. Error bars (mean ± standard deviation) of data were
shown in experiments. Detailed statistics for each experiment or
analysis were shown in Source Data.
Reporting summary
Further information on research design is available in the [350]Nature
Portfolio Reporting Summary linked to this article.
Supplementary information
[351]Supplementary Information^ (10.5MB, pdf)
[352]Reporting Summary^ (120.5KB, pdf)
[353]Transparent Peer Review file^ (651.8KB, pdf)
Source data
[354]Source Data^ (36.3MB, xlsx)
Acknowledgements