Autophagy-targeted NBR1-p62/SQSTM1 Complexes Promote Breast Cancer Metastasis by Sequestering ITCH

Abstract

Autophagy deficiency in breast cancer promotes metastasis through the accumulation of the autophagy cargo receptor NBR1. Here, we show that autophagy normally suppresses breast cancer metastasis by enabling the clearance of NBR1-p62/SQSTM1 complexes that instruct p63-mediated pro-metastatic basal differentiation programs. When autophagy is inhibited, the autophagy cargo receptors NBR1 and p62/SQSTM1 accumulate within biomolecular condensates in cells, which drives basal differentiation in both mouse and human breast cancer models. Mechanistically, these NBR1-p62/SQSTM1 complexes sequester ITCH, an ubiquitin ligase that degrades and negatively regulates p63 in breast cancer cells, thereby stabilizing and activating p63. Accordingly, mutant forms of NBR1 unable to sequester ITCH into NBR1-p62/SQSTM1 complexes do not promote basal differentiation and metastasis in vivo. Overall, our findings illuminate how proteostatic defects arising in the setting of therapeutic autophagy inhibition modulate epithelial lineage fidelity and metastatic progression.

Metastasis, the spread of cancer cells from the primary tumor to distant locations, is the leading cause of cancer-related deaths. To successfully metastasize, a tumor cell must overcome a range of challenges, including hypoxia, nutrient scarcity, reduced cell-matrix adhesion, and evasion of the adaptive and innate immune system¹. Hence, there is immense interest in targeting pathways that shield metastatic cells from these diverse stressors. One stress response pathway under intense investigation is macroautophagy (hereafter referred to as autophagy), a highly conserved lysosomal degradation process that promotes the metabolic fitness and survival of normal and cancerous cells^{2, 3}. Indeed, anti-malarials such as hydroxychloroquine have been repurposed as autophagy inhibitors in multiple clinical oncology trials, and multiple next-generation autophagy inhibitors are being developed as potential cancer therapeutics^{4, 5}.

Despite the promises of targeting autophagy in treating certain advanced cancers, its role in preventing or treating metastasis remains controversial because evidence indicates that autophagy suppresses metastasis⁶. For example, our previous study employing polyoma middle T (PyMT) mouse mammary cancer models uncovered that genetic autophagy inhibition strongly attenuates primary tumor growth, yet paradoxically promotes spontaneous metastasis to the lung and enables the outgrowth of disseminated tumor cells into overt macro-metastases⁷. Additional work similarly demonstrates autophagy inhibition in dormant disseminated tumor cells gives rise to aggressive metastatic subpopulations, which exhibit cancer stem-like properties and enhanced rates of proliferation^8–10. Taken together, these results demonstrate that autophagy restricts key rate-limiting steps in the metastatic cascade, namely colonization and outgrowth at distant tissue sites^{6, 11}.

Basal differentiation in breast cancer is associated with poor patient prognosis and the presence of aggressive and pro-metastatic traits^{12, 13}. These cancer cells show elevated expression of cytokeratins 5 and 14 (CK5 and CK14) and the activation of the transcription factor p63, a master regulator of basal epithelial differentiation^{14, 15}. In both PyMT primary tumors and metastases, the genetic loss of autophagy causes a significant increase in the population of tumor cells exhibiting basal differentiation markers, including CK14 and p63. Similar phenotypes are observed when autophagy in inhibited in 4T1 mammary cancer models in vivo or in human triple negative breast cancer cell lines in vitro⁷. Furthermore, these effects on pro-metastatic differentiation and metastasis arise due to impaired autophagy-dependent turnover of Neighbor of BRCA1 (NBR1), an autophagy cargo receptor that mediates selective autophagy and functions as a multidomain cellular signaling scaffold^{7, 16, 17}. Accordingly, we hypothesized that increased NBR1 in autophagy-deficient cells promotes lineage infidelity resulting in aggressive basal subpopulations with enhanced metastatic potential. In support, the enforced ectopic expression of NBR1 is sufficient to promote metastasis and pro-metastatic basal differentiation programs⁷.

Until now, the precise mechanisms through which the autophagic turnover of NBR1 suppresses metastatic colonization and outgrowth or impairs pro-metastatic basal differentiation have remained obscure. In addition to NBR1, the archetypal autophagy cargo receptor p62/SQSTM1 is the other major ACR that robustly accumulates in diverse autophagy-deficient breast cancer cells^{7, 18}. NBR1 forms oligomeric complexes with p62/SQSTM1 and these two ACRs are proposed to cooperate in the selective autophagy of various targets^{17, 19}. Interestingly, multiple recent studies demonstrate that NBR1 and p62/SQSTM1 form phase-separated biomolecular condensates, also termed p62 bodies, which profoundly accumulate in the autophagy-deficient cells^20–23. Importantly, the ectopic overexpression of NBR1 is sufficient to catalyze the formation of phase separated biomolecular condensates containing p62/SQSTM1, which triggers the activation of the KEAP1-NRF2 anti-oxidant pathway in hepatocytes, thereby broaching an important role for ACR condensates in fine-tuning signaling pathways that control cell fate and survival²³. Motivated by these results, we sought to further dissect the epistatic relationships between NBR1 and p62/SQSTM1 in the control of breast cancer metastasis. In addition, we scrutinized whether and how NBR1-p62/SQSTM1 governs pro-metastatic basal differentiation in breast cancer cells.

Results

p62/SQSTM1 Promotes Metastasis Of Autophagy-Deficient Tumors

Genetic deletion of autophagy in polyoma middle T (PyMT) mammary tumors results in the significant increase of tumor cells expressing the basal transcription factor p63⁷. To determine whether increased p63 observed in these tumor cells correlates with p62/SQSTM1 accumulation, we performed immunofluorescence staining for p62/SQSTM1 and p63 in autophagy-deficient tumors due to the genetic deletion of either ATG5 or ATG12. Autophagy inhibition resulted in p62/SQSTM1 protein accumulation in tumor cells from both primary tumors and lung metastases; remarkably, increased nuclear p63 expression was more commonly observed in tumor cells exhibiting high levels of p62/SQSTM1 (Extended Data Fig. 1a–1c). To determine whether p62/SQSTM1 functionally contributed to metastasis, we performed experimental metastasis assays using PyMT R221A mammary cancer cells expressing shRNA against the key autophagy component ATG7 in combination with two independent shRNAs targeting p62/SQSTM1 (Fig. 1a). Whereas stable ATG7 depletion in R221A cells increased both the size and number of metastases compared to control shRNA, co-depletion of p62/SQSTM1 reversed these metastatic phenotypes (Fig. 1b, 1c and Extended Data Fig. 1d). Furthermore, autophagy-deficient metastases displayed a significant increase in tumor cells positive for p63 and CK14; co-depleting either p62/SQSTM1 or NBR1 in these autophagy deficient metastases profoundly reduced both p63 and CK14 expression (Fig. 1d, 1e and Extended Data Fig. 1e–1g). Hence, both p62/SQSTM1 and NBR1 are functionally required for pro-metastatic basal differentiation and the increased metastatic potential of autophagy-deficient mammary tumor cells.

Fig. 1. p62/SQSTM1 Required for Pro-Metastatic Basal Differentiation Upon Autophagy Inhibition.

a, R221A cells immunoblotted for indicated proteins. Samples run on separate gels; loading control (actin) corresponds to p62 blot. b, Quantification of lung metastasis size in mice inoculated with R221A cells expressing shCTRL (n = 10), shATG7 (n = 22), shATG7+shp62, #1 (n = 13), shATG7+shp62, #2 (n = 13). c, H&E staining of lung metastases from indicated groups. d, Immunofluorescent staining of p63 (green) and nuclei (DAPI, blue) in lung metastasis. Images also shown in Extended Data Fig. 1f. e, Quantification of p63 and CK14 positive cells. Each point represents individual mice (N=5 per cohort) with 300 tumor cells counted from 4–8 metastasis. f, R221A, 4T1 and MDA-MB-231 cells immunoblotted for indicated proteins. Samples run on separate gels with equal loading of total protein. g-h, MDA-MB-231 cells stained for p63 (red, g), CK5 (green, h), CK14 (red, h) and DAPI (blue). i-j, Percent of MDA-MB-231 (i) or R221A (j) cells positive for p62 (left) and double positive for p62 and p63 (right) quantified by flow cytometry. k, Primary human breast organoid cells from high-risk prophylactic mastectomy patient (Org 130) treated with solvent, Ulk-i or HCQ for 18h. Top row: Staining for p63 (green) and nuclei (DAPI, blue). Bottom rows: Staining for p62 (green), CK5 (red) and nuclei (DAPI, blue). l, Quantification of cells from human breast prophylactic mastectomy (PM) organoids positive for p63 and CK5. 250 or more cells counted from each of five (N=5) individual patient organoid preparations. m, Quantification of CK14 positive cells from prophylactic mastectomy (PM, circles) and breast cancer (BC, triangles) patient organoids shown in Extended Data Fig. 2f–g. 250 or more cells counted from each individual PM (n=3) and BC (n=3) organoid. Statistics: Statistical significance determined by one-way ANOVA followed by Dunnett’s multiple comparisons test (b, e, l) or 2-tailed t test (i, j, m). Data represented by mean ± S.E.M. Each data point represents one mouse (b, e), one representative experiment (I, j) or one patient organoid preparation (l, m). Statistical data and unprocessed blots available as source data.

To extend these results, we evaluated the role of p62/SQSTM1 in mouse triple negative mammary cancer cells (R221A, 4T1), the human triple negative breast cancer line MDA-MB-231 and a non-cancerous basal breast epithelial cell line MCF10A. Autophagy inhibition, achieved via shRNA-mediated depletion of ATG7 or ATG12, resulted in increased expression of p63 and basal keratins (CK5, CK14), which was reversed upon co-depletion of p62/SQSTM1 (Fig. 1f and Extended Data Fig. 1h). Immunofluorescence staining for p63, CK5 and CK14 further corroborated that p62/SQSTM1 is required for the development of basal differentiation in autophagy deficient breast cancer cells (Fig. 1g, 1h, and Extended Data Fig. 1i, 2a).

We next asked whether pharmacological inhibitors targeting early and late stages of the autophagy trafficking pathway impacted basal differentiation. To more precisely quantify the association between p62/SQSTM1 accumulation and p63 activation, we performed flow cytometric analysis for p62/SQSTM1 and p63 expression in MDA-MB-231 and R221A cells. Cells were treated with MRT68921 (Ulk-i), a pharmacological inhibitor targeting Ulk1/2 kinases required for early autophagosome initiation, or alternatively, with the lysosome inhibitor hydroxychloroquine (HCQ), which blocks the late steps of autophagic degradation^{24, 25}. Similar to ATG7 knockdown, both autophagy inhibitors resulted in significant accumulation of p62/SQSTM1 and the concurrent induction of p63 in approximately 40–50% of the cells (Fig. 1i, 1j and Extended Data Fig. 2b). In addition, we generated R221A cell lines expressing a recombinant dual reporter autophagic flux probe²⁶. Upon ATG7 depletion or treatment with Ulk-i, nuclear p63 was significantly increased in individual cells with low autophagic flux (green) compared to those with increased autophagy (Extended Data Fig. 2c–2e). Lastly, we determined whether pharmacological autophagy inhibition in organoids from primary human breast tissue (Extended Data Fig. 2f) resulted in increased basal differentiation. We first evaluated organoids prepared from high-risk patients who had undergone prophylactic mastectomy (PM organoid)^{27, 28}. Autophagy inhibition using Ulk-i or HCQ significantly increased the number of cells expressing nuclear p63 and the basal keratins CK5 and CK14 (Fig. 1k–1m, Extended Data Fig. 2g). Similarly, autophagy inhibitor treatment of triple negative breast cancer organoids (BC organoids) caused a significant increase in p63, CK14 and CK5 positive cells (Fig. 1m, Extended Data Fig. 2h–2i). Overall, these results from diverse breast epithelial and cancer cells, including cells from primary patient organoids, corroborate that inhibiting distinct steps in the autophagy pathway promotes nuclear p63 expression and the expression of basal cytokeratins.

p62 Supports NBR1-Driven Basal Differentiation and Metastasis

p62/SQSTM1 forms complexes with multiple ACRs, including NBR1 and TAX1BP1^{29, 30}. Recently, these ACRs have been shown to co-localize in biomolecular condensates, both in vitro and in vivo, which play important roles in the selective autophagic degradation of proteins^{20, 23, 29}. Accordingly, we asked whether these individual ACRs functionally influenced basal differentiation in breast cancer cells when autophagy was inhibited. Upon treating cells with Ulk-i, MDA-MB-231 cells depleted of either NBR1 or p62/SQSTM1 did not exhibit increased nuclear p63 or basal keratin expression (CK5, CK14). In contrast, Tax1BP1 knockdown did not reverse basal differentiation in response to Ulk-i treatment (Fig. 2a and Extended Data Fig. 3a).

Fig. 2. p62/SQSTM1 Supports NBR1-Driven Pro-Metastatic Basal Differentiation and Metastasis.

a, Immunofluorescent staining for p63 (green) and CK14 (red) in MDA-MB-231 cells expressing indicated shRNAs, treated with Ulk1/2 kinase inhibitor (Ulk-i), or both. b, Immunoblot of R221A cell cohorts used for in vivo lung metastasis assay in Fig. 2c–d. Samples run on separate gels with equal protein loading; loading control (actin) corresponds to p62 blot. c, Quantification of lung metastatic size in mice inoculated with R221A cells expressing shCTRL (n = 7), WT NBR1 (n = 6), WT-NBR1+shp62, #1 (n = 8), WT-NBR1+shp62, #2 (n = 9). d, Representative H&E staining of lung metastases for indicated groups. e-f, Representative images of MDA-MB-231 cells expressing indicated constructs and stained with indicated markers. g, Immunoblot of MDA-MB-231 cells expressing the indicated constructs. For total cell lysate (right), separate gels run with equal total protein loading; loading control (actin) corresponds to p62 blot. h, MDA-MB-231 cells overexpressing WT NBR1 or D50R stained for FLAG (NBR1, red), p62 (green) and counterstained with DAPI (blue). Graph: Enumeration of p62, FLAG (NBR1) or double positive (p62⁺FLAG⁺) puncta per cell. Puncta from 10+ cells counted per condition; each dot indicates an individual cell. i-j, Immunofluorescent staining of MDA-MB-231 cells expressing indicated constructs. Cells immunostained for p63 (green, i) or CK5 (green, j) and counterstained with DAPI (blue). k, Quantification of percentage of p63 (green, nuclei) and CK5 positive MDA-MB-231 cells from Fig. 2i and 2j. 400 or more cells from each of three individual experiments counted for each group. l, Quantification of lung metastatic size in mice inoculated with R221A cells expressing shCTRL (n = 12), WT NBR1 (n = 11), D50R-NBR1 (n = 10). Mice in CTRL and WT NBR1 cohorts also included as controls in Fig. 2c. m, Representative H&E staining of lungs metastases for indicated groups. Statistics: Statistical significance determined by one-way ANOVA followed by Dunnett’s multiple comparisons test (c, h, k, l). Data represented by mean ± S.E.M. Each dot represents one animal (c and l) or percent cells from one individual experiment (k). Statistical data and unprocessed blots available as source data.

Based on these results, we hypothesized that p63 activation and basal keratin expression in autophagy-deficient cells resulted from the accumulation of complexes containing both p62/SQSTM1 and NBR1. To test this prediction, we further scrutinized whether the interaction between these two ACRs influenced pro-metastatic phenotypes. We previously demonstrated that ectopic overexpression of NBR1 (WT NBR1), but not p62/SQSTM1, in PyMT-R221A cells was sufficient to promote pro-metastatic basal differentiation and tumor metastasis⁷. Wild-type NBR1 (WT NBR1) overexpression at levels comparable to that observed upon autophagy inhibition resulted in increased p62/SQSTM1 puncta in breast cancer cells, consistent with previous findings that NBR1 promotes biomolecular condensate formation^{20, 23} (Extended Data Fig. 3b, c). We next tested whether p62/SQTM1 was functionally required for NBR1-driven metastasis by overexpressing WT NBR1 in combination with p62/SQSTM1 shRNAs in PyMT R221A cells (Fig. 2b). WT NBR1 overexpression resulted in significant increases in both size and number of lung metastases in vivo, which was reversed upon concurrent shRNA depletion of p62/SQSTM1 (Fig. 2c, 2d, and Extended Data Fig. 3d, 3e). Furthermore, p62/SQSTM1 depletion reversed the ability of WT NBR1 overexpression to promote p63 and CK14 expression (Extended Data Fig. 3f, 3g). To further verify these results in human cell lines, we overexpressed NBR1 in MDA-MB-231 cells, either alone or in combination with p62 shRNAs. Similar to R221A cells, NBR1 overexpression resulted in increased expression of p63 and basal keratins (CK5), which was reversed upon concomitantly depleting p62/SQSTM1 (Fig. 2e, 2f).

To complement these studies, we evaluated whether disrupting the ability of NBR1 to interact with p62/SQSTM1 impacted metastasis. We generated cells expressing a FLAG-tagged NBR1 carrying a D50R mutation in the PB1 domain, which disrupts binding to p62/SQSTM1^{16, 31}. Immunoprecipitation studies confirmed that the D50R-NBR1 mutant exhibited reduced interaction with p62/SQSTM1 compared to WT NBR1 (Fig. 2g). Furthermore, compared to WT NBR1, D50R-NBR1 exhibited reduced co-location with p62/SQSTM1, and diminished NBR1-p62/SQSTM1 double positive puncta formation (Fig. 2h). In contrast to WT NBR1, D50R-NBR1 did not promote increased nuclear p63 or basal cytokeratin expression in MDA-MB-231 cells and R221A cells (Fig. 2i–2k and Extended Data Fig. 3h, 3i). Also, unlike WT NBR1, ectopic overexpression of D50R-NBR1 in R221A cells was not sufficient to enhance lung metastasis (Fig. 2l, 2m and Extended Data Fig. 3j) or to promote the increased expression of p63 or CK14 within these metastatic lesions in vivo (Extended Data Fig. 3k, 3l). Overall, these results reveal that the ability of NBR1 to promote metastasis critically requires its interaction with p62/SQSTM1. They further substantiate that NBR1 and p62/SQSTM1 functionally cooperate to promote pro-metastatic differentiation, both in vitro and in vivo.

Enhanced p63 Transcription In Autophagy-Deficient Cells

We next tested whether autophagy inhibition and p62/SQSTM1 regulated the transcription of p63 or its targets. shRNAs targeting ATG7 or ATG12 were used to genetically inhibit autophagy, either alone or in combination with p62/SQSTM1 knockdown, in both human (MDA-MB-231) and mouse (R221A, 4T1) triple negative breast cancer cells. qPCR analysis revealed that multiple p63 targets, including CK5, CK14 and ITGA7, were upregulated in ATG7 and ATG12 deficient cells; the concomitant shRNA-mediated depletion of p62/SQSTM1 reversed the upregulation of these targets (Fig. 3a, 3b and Extended Data Fig. 4a). In addition, we analyzed individual tumor cells from primary and metastatic tumors from breast cancer patients using published single cell RNA sequencing datasets from triple negative and HER2 positive human breast tumors as well as from breast cancers that had metastasized to brain^{32, 33}. These results confirmed that in breast cancer patients, tumor cells with a high autophagy biogenesis gene signature score display a striking inverse correlation with multiple p63 target genes (Fig. 3c)³⁴. Although ATG7/12 knockdown or p62/SQSTM1 depletion did not significantly increase p63 mRNA expression (Extended Data Fig. 4a–4c), cycloheximide chase assays demonstrated that ATG7 knockdown and HCQ treatment increased p63 protein stability compared to controls (Extended Data Fig 4d–e). Altogether, these findings suggested that decreased autophagy in human breast cancers is associated with increased p63 target expression³⁵. In further support, enforced overexpression of p63 (full length p63α) was sufficient to promote basal cytokeratin expression in MDA-MB-231 cells (Fig. 3d). In R221A cells, WT p63 overexpression increased lung metastases at levels comparable to WT NBR1 overexpression; increased CK5 and CK14 was observed within these metastatic lesions in vivo (Fig. 3e–h and Extended Data 4f). p63 overexpression. Notably, because p63 knockdown in MDA-MB-231 and R221A cells caused extensive cell death, we were unable to assess if p63 was functionally required for basal differentiation or increased metastasis upon autophagy inhibition (Extended Data Fig. 4g). Based on these results, we propose that autophagy inhibition stabilizes p63 protein levels, resulting in increased expression of p63-driven pro-metastatic targets (Extended Data Fig. 4h).

Fig. 3. Autophagy Inhibition and p62/SQSTM1 Promote Transcription of p63 Targets.

a-b, qRT-PCR analysis of p63-target mRNAs (CK5, CK14 and ITGA7) for indicated groups in MDA-MB-231 (a) and R221A cells (b). c, Heatmap of autophagy gene signature from primary and metastatic tumors from breast cancer patients using publicly available single cell RNA sequencing datasets, including triple negative (top), HER2 positive (middle) human breast tumors and from breast cancers that had metastasized to brain (bottom). d, Left panel: Immunoblot analysis of MDA-MB-231 cells expressing CTRL or myc-tagged WT-p63 for the indicated proteins. Samples run on separate gels with equal loading of total proteins; loading control (actin) corresponds to the p62 blot. Right: Immunofluorescence staining for myc tag (WT-p63, green) and CK5 (red). e, Top-left: Immunoblot of R221A cells expressing either WT-NBR1 or WT-p63 used for in vivo lung metastasis assay in Fig. 3f–h. Samples run on same gel, including loading control (actin). Top-right and bottom: Representative H&E staining of lung metastases for indicated groups. f-g, Quantification of metastatic size (f) and numbers of metastases (g) in mice inoculated with R221A cells expressing CTRL (n = 8), WT NBR1 (n = 9), WT-p63 (n = 12). h, Representative immunofluorescence staining for p63 (green), CK5 (red) and nuclei (DAPI, blue) in lung metastasis from indicated groups. Statistics: Data (a, b, f and g) presented by mean ± S.E.M. Each dot in (a) and (b) represents a technical replicate within one representative biological replicate. In (f) and (g), statistical significance determined by ordinary one-way ANOVA followed by Dunnett’s multiple comparisons test; each dot represents one animal. Statistical data and unprocessed blots available as source data.

Accumulated NBR1-p62/SQSTM1 Complexes Sequester ITCH

Growing evidence suggests that p62/SQSTM1 accumulation in autophagy-deficient cells can sequester proteins that negatively regulate pro-tumorigenic transcriptional regulators and enzymes^{10, 11, 36, 37}. This is best exemplified by the sequestration of KEAP1, an adaptor subunit of the Cullin 3-based E3 ubiquitin ligase, into p62/SQSTM1 condensates, which impedes the ability of Cullin 3 to degrade NRF2, a transcription factor that promotes antioxidant gene expression and cancer cell survival^{36, 38}. Given our findings that both NBR1 and p62/SQSTM1 functionally impacted metastasis downstream of autophagy inhibition, we postulated that complexes containing both ACRs recruited and sequestered a negative regulator of p63 activation, resulting in increased basal differentiation (Extended Data Fig. 4h). To further understand how accumulation of NBR1-p62/SQSTM1 leads to the stabilization of p63, we overexpressed epitope-tagged NBR1, both WT and the D50R mutant, in HEK293T cells. Similar to our results in breast cancer cells (Fig. 2h and Extended Data Fig. 3c), WT NBR1 overexpression was sufficient to robustly promote the formation of puncta containing both NBR1 and p62/SQSTM1 in these cells (Fig. 4a). Tandem mass tag mass-spectrometric analysis of immunoprecipitated NBR1 complexes (Fig. 4b, 4c, and Extended Data Fig. 5a, 5b, Supplementary Table 1) identified several known NBR1-interacting proteins, including its binding partners p62/SQSTM1 and TAX1BP1, as well as KEAP1. Intriguingly, this NBR1 interactome contained ITCH, a NEDD4-family E3 ligase previously implicated in promoting p63 turnover via the ubiquitin-proteasome system (UPS)³⁹ (Fig. 4c and Extended Data Fig. 5b).

Fig. 4. NBR1-p62/SQSTM1 Complexes Sequester ITCH in Autophagy Deficient Cells.

a, Indicated HEK293T cohorts immunostained for FLAG (green) and p62 (red). b, Coomassie staining of α-FLAG immunoprecipitates utilized for TMT-MS. c, Volcano plot of proteins identified by TMT-MS from α-FLAG immunoprecipitates from HEK293T cells overexpressing empty vector (CTRL) versus WT NBR1 (both 3XFLAG-tagged). Proteins plotted according to −log₁₀ P values as determined by two-tailed t-test and log₂-fold enrichment (n=3 biological replicates). Red: proteins significantly enriched in WT NBR1, identified with P>0.05 and log₂-fold change>|0.5|. Blue: significantly enriched in CTRL. Grey: not enriched in either group. d, Immunoblotting of α-p62 or α-FLAG immunoprecipitates from indicated MDA-MB-231 cohorts. Samples run on separate gels with equal total protein loading. e, α-ITCH or α-TAX1BP1 immunoprecipitation from MDA-MB-231 cells expressing indicated constructs or treated with Ulk-i. Samples run on separate gels. f, Indicated MDA-MB-231 cohorts cultured in full media (control), starved in 0.1% serum, HCQ treated, or both for 18h; cells lysed and immunoblotted for indicated proteins. Samples run on separate gels with equal total protein loading; loading control (actin) corresponds to ITCH blot. g, Double positive (p62⁺ITCH+) puncta enumerated from indicated cells stained for p62 and ITCH. N=10 cells per condition from three individual experiments. h, Indicated MDA-MB-231 cells stained with p62 (green), ITCH (red) and DAPI (blue). i, Cells from high-risk prophylactic mastectomy organoids (Org 130) treated with solvent, Ulk-i, or HCQ and stained for p62 (green), ITCH (red) and DAPI (blue). j, MDA-MB-231 cells expressing indicated constructs stained for ITCH (red), p62 (green) and nuclei (DAPI, blue). k, Quantification of percent colocation (relative to total number of puncta) of p62 and ITCH puncta in MDA-MB-231 and R221A cells. Minimum of 300 puncta from three individual fields (N=3) representing at least 25 individual cells were counted. Statistics: Each dot indicates a cell (g) and each data point (k) represents percent of colocalized puncta enumerated from two (R221A) or three (MDA-MB-231) independent fields. Data represented by mean ± S.E.M. Statistical significance (g and k) determined by two-way ANOVA followed by Dunnett’s multiple comparisons test. Statistical data and unprocessed blots available as source data.

We next validated that NBR1 biochemically interacted with these identified components (Extended Data Fig. 5c) and that p62/SQSTM1, TAX1BP1 and ITCH all co-localized with FLAG-NBR1 in puncta (Fig. 4a and Extended Data Fig. 5d–5e). Interestingly, the PB1 mutation (D50R) in NBR1 reduced its interaction not only with p62/SQSTM1, but also with ITCH in both MDA-MB-231 and HEK293T cells (Fig. 4d and Extended Data Fig. 5c, 5f). Reverse immunoprecipitation with an anti-ITCH antibody (Fig. 4e, lanes 1–4) corroborated that ITCH is not able to interact with NBR1 or p62/SQSTM1 in D50R overexpressing cells (Fig. 4e and Extended Data Fig. 5g). In contrast, immunoprecipitation with anti-TAX1BP1 antibody (lanes 5–8) demonstrated that TAX1BP1 interacted with NBR1, but unlike p62/SQSTM1 and ITCH, this interaction was not influenced by mutation of the NBR1 PB1 domain (lanes 6, 7). Remarkably, ITCH co-immunoprecipitated with NBR1 and p62/SQSTM1 in cells treated with Ulk-i (lanes 4, 8).

Because ITCH interacted with these ACRs, we tested whether ITCH was degraded via autophagy. However, in contrast to p62/SQSTM1, ITCH protein levels were not significantly altered by serum starvation, lysosomal inhibition using HCQ or upon WT NBR1 or D50R overexpression (Fig. 4f). In parallel, we determined whether autophagy inhibition impacted ITCH sequestration within p62/SQSTM1 bodies. Indeed, ATG7 knockdown caused a significant increase in the punctate co-location of ITCH and p62/SQSTM1 in both MDA-MB-231 and R221A cells (Fig 4g, 4h). Similarly, we observed increased ITCH colocalization within p62/SQSTM1 puncta in human breast organoids treated with pharmacological autophagy inhibitors (Fig. 4i). Moreover, the overexpression of WT NBR1, but not D50R-NBR1, resulted in the punctate colocation of ITCH with p62/SQSTM1 (Fig. 4j–k). Finally, in WT NBR expressing cells, p62/SQSTM1 knockdown resulted in the significant reduction in NBR1-ITCH colocalization, whereas in TAX1BP1 depleted cells, this colocalization was maintained (Extended Data Fig. 5h and 5i). Based on these results, we propose that autophagy inhibition promotes the sequestration of ITCH into NBR1-p62/SQSTM1 puncta.

ITCH Degrades p63 In Autophagy Competent Breast Cancer Cells

Because ITCH promotes p63 ubiquitination³⁹, we next investigated whether NBR1 and p62/SQSTM1 regulated the ability of ITCH to degrade p63 via the UPS (Fig. 5a). p63 was stabilized in MDA-MB-231 cells overexpressing WT NBR1 (lane 2) or treated with Ulk-i (lane 5); shRNA depletion of either p62/SQSTM1 (lane 6) or NBR1 (lane 7) reversed the effects of Ulk-i treatment on p63. Nonetheless, treatment with the proteasome inhibitor MG132 for 4hr or overnight (O/N, 18h) led to p63 stabilization in all of these conditions. Furthermore, shRNA depletion of ITCH also stabilized p63 protein, corroborating that the UPS-mediated degradation of p63 required ITCH-mediated ubiquitination (lane 8). To further verify the role of ITCH in p63 ubiquitination, we overexpressed HA-tagged ubiquitin in MDA-MB-231 cells to assess p63 ubiquitination status in response to various perturbations (Fig. 5b). Increased p63 ubiquitination was observed upon treatment with proteasome inhibitor (MG132, lane 2) but not with treatment with autophagy inhibitor (Ulk-i) or upon shRNA-mediated ITCH depletion, despite the stabilization of p63 (lanes 3–5). Furthermore, WT-ITCH augmented p63 ubiquitination, resulting in p63 destabilization (lane 6) unless proteasome inhibitor was present (lane 7). Finally, Ulk-i mediated autophagy inhibition counteracted the effects of ITCH overexpression on the stability of p63 (lane 8).

Fig. 5. ITCH Promotes UPS-Mediated Degradation of p63 in Breast Cancer Cells and Forms Trimeric Complex With NBR1 and p62/SQSTM1 In Silico.

a, Immunoblotting of MDA-MB-231 cells expressing the indicated constructs. As indicated, cells were treated with the Ulk kinase inhibitor (Ulk-i) and the proteasome inhibitor MG132 for 4 or 18 hours (overnight, O/N). Samples run on separate gels with equal loading of total proteins; loading control (actin) corresponds to the p63 (-MG132) blot. b, Immunoblot of HA-(ubiquitin) immunoprecipitated complexes (top) or total cell lysate (bottom) from MDA-MB-231 cells expressing HA-ubiquitin along with indicated constructs. As indicated, cells were treated with Ulk-i or MG132 for 18h. For total cell lysates, separate gels were run with equal loading; loading control (actin) corresponds to the ITCH blot. c-h, Protein structure prediction of the trimeric complex consisting of the N-terminal regions of ITCH, NBR1 and p62, modelled using ColabFold and visualized in ChimeraX. c, Illustration of the predicted relevant binding between ITCH, NBR1, and p62. d, Overview of the high-level protein structure. e, rotated view (180°) highlighting key domains, including the ITCH WW domains (WW1-WW4) and the PB1 and Zinc Finger (ZnF) domains of both NBR1 and p62. f-h, Enlarged views of the predicted contact sites between (f) NBR1 and p62 PB1 domains; (g) NBR1 PB1 domain and ITCH WW3 domain; and (h) NBR1 ZnF domain and ITCH WW4 domain. Contact lines predicted by ColabFold visualized using default ChimeraX settings (3 Å) and color-coded by PAE (Predicted Aligned Error) values (in angstroms), with a linear color gradient ranging from 5 Å (blue) to 10 Å (yellow) to 20+ Å (red). Unprocessed blots are available as source data.

Interactions Between NBR1-p62/SQSTM1 and ITCH

Using the AlphaFold2 algorithm via ColabFold to predict the structure of heteromeric complexes^{40, 41}, we computationally modelled the trimeric interactions between the N-termini of NBR1, p62/SQSTM1 and ITCH (Fig. 5c–5e). We corroborated previous experimentally described interactions between the PB1 domains of NBR1 and p62/SQSTM1 in this multimeric structure, supporting the validity of this in silico model (Fig. 5f). In addition, this model uncovered multiple points of direct interaction between the PB1 and ZnF domains of NBR1 with the WW3 and WW4 domains of ITCH, respectively (Fig 5g, 5h).

Based on this model, we first made multiple deletion and point mutants in the PB1 domain of NBR1. These PB1 domain mutants were unable to bind both p62/SQSTM1 and ITCH, similar to the phenotype observed with D50R-NBR1 (Fig. 6a). We also uncovered multiple interactions between the NBR1 ZnF and the ITCH WW3 domains. Accordingly, the deletion of the ZnF domain in NBR1 (amino acids 211–256) attenuated its interaction with ITCH but not p62/SQSTM1 (Fig. 6a). Because several interactions were located in the vicinity of the three di-Cysteine motifs of the NBR1 ZnF domain, we mutated these motifs, both individually and in combination (Extended Data Fig. 6a). The individual mutation of two di-Cys motifs, CC1 (residues 217/220) and CC3 (residues 240/243) reduced binding to ITCH, whereas the mutation in CC2 (residues 231/234) had modest effects. Combined mutations of any two di-Cys motifs further reduced binding to ITCH as did the mutation of all three di-Cys motifs; these effects were all comparable to the effects of ZnF deletion. Remarkably, none of these mutations impacted binding to p62/SQSTM1, supporting that they did not completely disrupt the structure of NBR1. Furthermore, the deletion of either the WW3 and WW4 domains in ITCH reduced its interactions with both NBR1 and p62/SQSTM1 (Fig. 6b) and attenuated ITCH sequestration into p62 bodies in the presence of HCQ (Extended Data 6b). Overall, these results support that the PB1 and ZnF domains of NBR1 interact with the WW3 and WW4 domains of ITCH.

Fig. 6. Validation of Protein-Protein Interactions Between NBR1, p62/SQSTM1 and ITCH.

a, α-FLAG immune complexes (upper) and total cell lysates (lower) from indicated MDA-MB-231 were immunoblotted for the indicated proteins. Samples run on separate gels with equal loading of immune complex (upper) or total cell lysate (lower). For immune complexes, loading control (NBR1) corresponds to p62 blot and for total cell lysates, loading control (actin) corresponds to the p63 gel. Also see Extended Data Fig. 6a. b, α-GFP immune complexes (upper) and total cell lysates (lower) from MDA-MB-231 cells expressing indicated ITCH constructs immunoblotted for the indicated proteins. Samples run on separate gels with equal loading of immune complex (upper) or total cell lysate (lower); for immune complexes, loading control (GFP-ITCH) corresponds to p62 gel and for total cell lysates, loading control (actin) corresponds to GFP-ITCH. c, MDA-MB-231 cells expressing either PB1 or zinc-finger mutant of NBR1 (FLAG-tagged) stained for the indicated markers. d-f, In MDA-MB-231 cells expressing the indicated constructs from Fig. 6c, quantification of the percentage of double positive puncta with colocalization of: (d) FLAG (NBR1) and p62, (e) FLAG (NBR1) and ITCH or (f) p62 and ITCH. A minimum of ten (N=10) cells from four independent experiments counted for each group; each data point represents the percent of double positive puncta with the indicated colocalization within an individual cell. (g) Representative p63 (green) immunofluorescent staining of MDA-MB-231 cells expressing indicated NBR1 constructs. Nuclei stained with DAPI (blue). Statistics: Data (d-f) are represented by mean ± S.E.M. Statistical significance determined by one-way ANOVA followed by Dunnett’s multiple comparisons test; (NS)= non-significant (d-f). Statistical data and unprocessed blots are available as source data.

We also evaluated how this spectrum of NBR1 mutations impacted the colocation of ITCH with NBR1-p62/SQSTM1 puncta (Fig. 6c). Compared to WT NBR1, mutations in the PB1 domain, including D50R, all reduced NBR1 colocalization with both p62/SQSTM1 and ITCH (Fig. 6c–e). In contrast, mutations in the NBR1 ZnF domain, reduced its colocalization with ITCH but not with p62/SQSTM1 (Fig. 6c–e). Furthermore, both the D50R mutation and ZnF domain deletion in NBR1 reduced ITCH colocalization with p62/SQSTM1, further confirming that NBR1 is functionally required to incorporate ITCH into NBR1-p62/SQSTM1 complexes (Fig. 6f). Finally, compared to WT NBR1, the overexpression of NBR1 mutants resulted in reduced protein stability and nuclear localization of p63 (Fig. 6a, 6g and Extended Data Fig. 6a)

Dynamic Properties of p62/SQSTM1 and ITCH Condensates

Previous studies have shown p62/SQSTM1 in association with ubiquitin undergoes phase separation, which is proposed to mediate the formation of biomolecular condensates, termed p62 bodies, that sequester important molecules, such as KEAP1, during autophagy inhibition^{21, 22, 29}. Hence, we asked whether the ITCH and p62/SQSTM1 observed in puncta share similar chemical and dynamic properties. First, MDA-MB-231 cells treated with 5% 1, 6-hexanediol, an alkenediol that disrupts phase-separated condensates, but not 5% 2, 5-hexanediol, which is unable to dissolve condensates, exhibited a marked reduction in both p62/SQSTM1 and ITCH puncta. (Fig. 7a and Extended Data Fig. 6c). To verify the dynamic nature of these puncta, we examined the pattern of fluorescence recovery after photobleaching (FRAP) of mCherry-tagged p62 and GFP-tagged ITCH puncta. These proteins demonstrated comparable fluorescence intensity recovery kinetics at laser-bleached sites with almost full recovery within three to four minutes. The mobile fractions of mCherry-p62 and GFP-ITCH were calculated as 91.5% and 97.1%, respectively, and the time for half recovery of fluorescence intensity was 62.5s and 50.0s, respectively (Fig. 7b–7d and Supplementary Videos 1, 2). Finally, because p62/SQSTM1 condensates exhibit fission and fusion^{30, 42}, we examined whether similar exchanges occurred between p62/SQSTM1 and ITCH puncta. In MDA-MB-231 cells co-expressing mCherry-tagged p62 and GFP-tagged ITCH, we observed fission (Fig. 7e and Extended Data Fig. 6d, Supplementary Video 3) and fusion (Fig. 7f and Extended Data Fig. 6e, Supplementary Video 4) among these puncta. Overall, these data support that ITCH and p62/SQSTM1 co-localize into puncta within cells and exhibit chemical and dynamic properties consistent with their incorporation into biomolecular condensates.

Fig. 7. Dynamic Properties of p62/SQSTM1 and ITCH in Biomolecular Condensates.

a, MDA-MB-231 cells treated with HCQ overnight and then subject to 5% 2,5-hexanediol or 1,6-hexanediol treatment for indicated times. Puncta from ten individual cells (N=10) enumerated for each treatment group. Each data point represents mean number of puncta ± S.E.M. Also see Extended Data Fig. 6c. b, Recovery of fluorescence (FRAP) of GFP-ITCH (green) and mCherry-p62 (red) puncta in MDA-MB-231 cells scored at 10-seconds intervals following photobleaching. (N=26 for GFP-ITCH and N=22 for mCherry-p62). Arrows indicate time points for half-recovery. Data represented by mean ± S.E.M. c, Representative frames of FRAP of mCherry-p62 puncta in MDA-MB-231 cells for indicated time points. Bleached puncta indicated by white circles whereas unbleached (control) puncta encircled in yellow. Also see Supplementary Video 1. d, Representative frames of FRAP of GFP-ITCH puncta in MDA-MB-231 cells for the indicated time points. Bleached puncta indicated by red circles whereas unbleached (control) puncta encircled in yellow. Also see Supplementary Video 2. e, Representative frames of GFP-ITCH (green) and mCherry-p62 (red) puncta in MDA-MB-231 cells over indicated time points. Frames demonstrate course of fission event of multiple (#2, #3) green (ITCH) and p62 (red) puncta; point of fission indicated by arrow. Also see Extended Data Fig. 6d and Supplementary Video 3. Images in Fig. 6e also shown as part of Extended Data Fig. 6d. f, Representative frames of GFP-ITCH (green) and mCherry-p62 (red) puncta in MDA-MB-231 cells for indicated time points. Frames demonstrate course of fusion event of ITCH (green, #1) and p62 (red, #2) puncta; point of fusion indicated by arrow. Also see Extended Data Fig. 6e and Supplementary Video 4. Images in Fig. 6f are also shown as part of Extended Data Fig. 6e. Statistical data and unprocessed blots are available as source data.

ITCH Restricts Mammary Cancer Metastasis

Because the NBR1 ZnF domain promotes ITCH sequestration into NBR1-p62/SQSTM1 condensates and the activation of p63 in breast cancer cells, we next tested whether it influences metastasis in vivo. Experimental lung metastasis was reduced in PyMT-R221A cells expressing del-ZnF NBR1 compared to WT NBR1 (Fig. 8a–8c and Extended Data Fig. 7a). In addition, del-ZnF NBR1 metastases exhibited reduced p63 and CK14 expression compared to WT NBR1 (Fig. 8d–8f). Furthermore, ITCH loss-of-function in PyMT-R221A cells resulted in a significant increase in the number and size of metastases compared to non-targeting controls (Fig. 8g–8i and Extended Data Fig. 7b).

Fig. 8. ITCH Restricts Basal Differentiation and Metastatic Outgrowth of Mammary Tumors.

a, Average lung metastatic size in mice inoculated with R221A cells expressing shCTRL (n = 12), WT-NBR1 (n = 11) and del-ZnF-NBR1 (n=10). CTRL and WT NBR1 cohorts included as controls in Fig. 2c and 2l. b, H&E staining of lungs bearing metastases from indicated R221A cohorts. c, Immunoblot of R221A cell cohorts used for lung metastasis assays in Fig. 8a–b. Loading control (actin) corresponds to FLAG (NBR1) blot. d-f, Quantification (d) and representative images (e-f) of p63 and CK14 positive cells (green) in lung metastases from indicated cohorts. Nuclei counterstained with DAPI (blue). Each data point represents individual mice per condition with at least 250 tumor cells from multiple (4–8) metastases counted. N=5 individual mice from three independent experiments evaluated. e-f, g, Quantification of average metastatic size in in mice inoculated with R221A cells expressing shCTRL (n = 13), shATG7 (n=6), shITCH, #1 (n = 25), shITCH, #2 (n=24), shKEAP1, #1 (n=12) and shKEAP1, #2 (n=12). Related to Extended Data Fig. 7b. Mice in shCTRL and shATG7 cohorts also used in Fig. 1b and 3f. h, Representative H&E staining of lungs bearing metastases for indicated cohorts. i, Immunoblot of R221A cell cohorts expressing indicated shRNAs used for in vivo lung metastasis assays in Fig. 8g and 8h. Samples run on separate gels run with equal loading of total proteins; loading control (actin) corresponds to NRF2 blot. j-k, Representative immunofluorescent staining for CK14 (green) and NRF2 (red) in (j) or, p63 (green) in (k) of lung metastases from indicated groups. Nuclei counterstained with DAPI (blue). l, Quantification of p63 (nuclei) and CK14 positive cells in Fig. 8j and 8k. Minimum of 250 cells from multiple (4–8) metastases are counted per condition. N=5 individual mice from three independent experiments evaluated. Statistics: Each dot (a, d, g and l) represents one animal. Data are represented by mean ± S.E.M. Statistical significance determined by one-way ANOVA followed by Dunnett’s multiple comparisons test (a, d, g, and l). NS=non-significant (8g). Statistical data and unprocessed blots are available as source data.

Notably, our proteomics studies (Fig. 4c, Supplementary Table 1) also identified KEAP1, a negative regulator of NRF2 antioxidant pathway activation, as an NBR1-interacting protein²³. Because recent work demonstrates that NRF2 activation facilitates metastatic progression,⁴³ we scrutinized the effects of KEAP1 loss-of-function on metastasis. However, in contrast to ITCH, shRNA depletion of KEAP1 did not enhance breast cancer metastasis compared to controls (Fig. 8g–8i and Extended Data Fig. 7b). Remarkably, ITCH-depleted, but not KEAP1-depleted, metastases exhibited increased numbers of tumor cells with nuclear p63 and high CK14 expression in comparison to controls (Fig. 8j–8l). On the other hand, increased nuclear NRF2 was observed in KEAP1-depleted, but not ITCH-depleted metastases (Fig. 8j). ITCH knockdown in human MDA-MB-231 cells also resulted in the increased expression of nuclear p63 and basal cytokeratins (Extended Data Fig. 7c, 7d). Finally, no compensatory upregulation of NEDD4, a E3 ligase related to ITCH, was observed in ITCH depleted cells and NEDD4 knockdown did not affect either p63 or CK5 protein levels (Extended Data Fig. 7e). Altogether, these results support that the accumulation of the autophagy-targeted complex containing the ACRs NBR1 and p62/SQTM1 sequesters the ubiquitin ligase ITCH; as a result, these complexes facilitate the stability and activation of the transcription factor p63, thereby promoting basal differentiation and breast cancer metastasis (Extended Data Fig. 7f).

Discussion

Here, we uncover that autophagic clearance of NBR1-p62/SQSTM1 protein complexes governs the activation of pro-metastatic differentiation programs mediated by p63, a p53 family transcription factor that promotes basal and stem-like fates in both normal and cancerous epithelial tissues, including the breast^{35, 44}. Autophagy inhibition promotes the activation of p63 in multiple human and mouse breast cancer models, including primary breast organoids generated from patients with breast cancer or at high risk to develop breast cancer^{27, 28}. Moreover, interrogation of multiple single cell RNA sequencing datasets from breast cancer patients, including patients with brain metastasis, corroborates the inverse correlation between autophagy biogenesis gene signatures and the expression of multiple p63 target genes^{7, 32, 33}. Overall, these results highlight a crucial proteostatic role for the autophagy pathway that suppresses p63-mediated differentiation programs and impedes metastatic phenotypes both in vitro and in vivo.

Breast cancer is a heterogenous disease comprised of distinct subtypes. Among these subtypes, basal-like breast cancers are notable for their aggressive nature, with a higher risk of recurrence and metastasis¹³. Despite the expression of basal cytokeratins, such as CK14, basal-like breast cancers consist of mixed cell lineages proposed to originate from luminal progenitor cells in the normal breast^{12, 45}. Furthermore, evidence suggests that basal-like breast cancer cells can dynamically transition between CK14+ and CK14- states, and the mechanisms governing these transitions remain a subject of active investigation^{14, 45}. Our results reveal an important mechanism through which autophagy suppresses lineage transitions toward basal states by suppressing p63 transcriptional programs. Importantly, these beneficial effects of autophagy in preventing aggressive basal differentiation may not be limited to metastatic settings, as we have observed increased basal differentiation when autophagy is acutely inhibited in non-cancerous breast organoids from high-risk patients. Notably, recent high-resolution mapping of histologically normal human breast tissue has identified a sub-population of cells, termed basal-luminal (BL) cells. These cells express transcriptional signatures associated with basal-like breast cancer, show impaired lineage fidelity, and their numbers tend to increase with age²⁸. Hence, an important matter for future investigation is determining whether and how reduced autophagy during aging contributes to lineage infidelity and predisposition to cancer.

Remarkably, although p63 protein levels profoundly increase in autophagy-deficient cells, p63 is not a direct target of autophagic degradation. Rather, the stabilization and activation of p63 is secondary to the accumulation of complexes containing NBR1 and p62/SQSTM1. In our previous study, we demonstrated that aberrant accumulation of NBR1 was both necessary and sufficient to promote metastasis in autophagy-deficient breast cancers. In contrast, p62/SQSTM1 was not sufficient to promote metastasis, suggesting that NBR1 was unique among the ACRs in facilitating metastatic outgrowth of autophagy-deficient cells⁷. However, recent cell biological studies revealed that NBR1 catalyzes the formation of p62/SQSTM1 biomolecular condensates, also termed p62 bodies, typically observed in the setting of autophagy deficiency²³. These results motivated us to scrutinize the potential interconnections between NBR1 and p62/SQSTM1 in promoting metastasis. Here, we now demonstrate that either autophagy inhibition or NBR1 overexpression in breast cancer cells elicits a profound increase in NBR1-p62/SQSTM1 complexes. Furthermore, in both autophagy-deficient and NBR1-overexpressing breast cancer cells, we uncover that p62/SQSTM1 is crucially required for both stabilizing and activating p63, resulting in basal cytokeratin expression and enhanced lung metastatic outgrowth in vivo. In contrast, expression of a PB1 domain mutant (D50R) prevents the accumulation of NBR1-p62/SQSTM1 complexes in breast cancer cells, thereby attenuating basal differentiation and metastasis in vivo. Altogether, these results support that the autophagic clearance of NBR1-p62/SQSTM1 suppresses metastatic outgrowth and pro-metastatic differentiation programs (Extended Data Fig. 7f). Given the growing interest in selective autophagy and protein homeostasis, our results provide unique insight into the physiological significance of these autophagy-targeted complexes during metastatic progression in vivo.

In autophagy-deficient cells, the accumulation of ACR-containing biomolecular condensates has been proposed to modulate transcription factors and other signaling mediators. Most notably, these multimeric complexes can sequester molecules that normally facilitate transcription factor degradation via the ubiquitin-proteasome system (UPS)¹¹. Accordingly, our proteomic analysis of NBR1-interacting proteins identifies ITCH as the ubiquitin E3-ligase that mediates the UPS-mediated degradation of p63³⁹. We demonstrate that ITCH biochemically interacts with and localizes at NBR1-p62/SQSTM1 complexes in autophagy-deficient cells, and that ITCH loss-of-function in autophagy-competent cells is sufficient to promote p63 stability, pro-metastatic differentiation and lung metastasis in vivo. Importantly, these metastatic phenotypes are abrogated upon expression of an NBR1 mutant lacking the ZnF domain, which disrupts the ability of NBR1-p62/SQSTM1 complexes to bind and sequester ITCH. Furthermore, the chemical and dynamic properties of ITCH puncta resemble p62/SQSTM1 puncta, which has been demonstrated to form phase-separated biomolecular condensates^20–23. Based on our results, we propose that NBR1-p62/SQSTM1 complexes accumulate in autophagy deficient breast cancer cells, which are able to sequester ITCH away from p63, thereby triggering pro-metastatic basal differentiation and metastatic outgrowth (Extended Data Fig. 7f).

Remarkably, this model for p63 activation resembles how p62/SQSTM1 condensates activate NRF2, a transcription factor that mediates antioxidant defense. Upon autophagy inhibition, accumulated p62/SQSTM1 sequesters KEAP1, the substrate adaptor for the CUL3 E3-ubiquitin ligase complex that degrades NRF2³⁶. Notably, we also identified KEAP1 as an NBR1-interacting protein, consistent with recent studies demonstrating that NBR1 catalyzes p62/SQSTM1 condensate formation and triggers KEAP1-NRF2 antioxidant pathway activation²³. Although NRF2 activation is implicated in both metastasis and adaptive resistance to autophagy inhibition, our results indicate that KEAP1-NRF2 pathway activation alone is not sufficient to promote metastasis or basal differentiation in the breast cancer models we have utilized^{43, 46, 47}. Further clarifying whether and how NRF2 activation contributes to metastasis upon autophagy inhibition in breast cancer remains an important question for future study.

Our study has certain limitations. First, for our correlative studies in human breast cancers, we employ an autophagy gene signature as a surrogate for autophagic capacity in patients. While recognizing that static measurements are not ideal for evaluating the highly dynamic process of autophagy, we have found this indirect measure to be more effective than single immunohistochemical markers, such as LC3 or p62/SQSTM1, to assess autophagy in human tissues. Second, although our results are most consistent with the model that ITCH directly interacts with NBR1-p62/SQSTM1, we have uncovered numerous additional NBR1-interacting proteins through our proteomics studies. Thus, we cannot completely exclude that additional factors may contribute to the sequestration of ITCH within NBR1-p62/SQSTM1 complexes. Third, because p63 knockdown caused death of key cell lines in our studies, we have been unable to rigorously assess whether p63 is functionally required for the increase in metastasis observed during autophagy inhibition. At the same time, while ITCH downregulation is sufficient to promote metastatic outgrowth, its effects on lung metastases are not as pronounced as that obtained with ATG7 knockdown or WT NBR1 overexpression (Fig. 8g–8i). Therefore, we postulate that additional pathways contribute to the effects of autophagy inhibition or enhanced NBR1-p62/SQSTM1 condensate formation on metastasis. Additionally, the ITCH ubiquitin ligase likely plays a role in the proteasomal degradation of proteins beyond p63. Our ongoing studies are aimed at identifying other pro-metastatic pathways downstream of NBR1-p62/SQSTM1 and deciphering the broader range of ITCH substrates to identify potential therapeutic targets for impeding metastatic progression.

To date, interest in therapeutically inhibiting autophagy in the clinical oncology setting has been motivated by the fundamental tumor-supporting functions attributed to this pathway, such as nutrient scavenging, metabolic salvage and immune evasion^{3, 48}. Our research uncovers an additional consequence of inhibiting autophagy in breast cancer cells, the disruption of proteostasis, that orchestrates untoward effects on pro-metastatic basal differentiation. As new autophagy-modulating compounds are developed for the treatment of cancer, simultaneously targeting the formation of NBR1-p62/SQSTM1 condensates may be a potential therapeutic strategy to prevent life-threatening metastatic recurrence when autophagy is inhibited.

Methods

Ethical Approval:

Animal studies performed in accordance with NIH and the US Public Health Service Policy on Humane Care and Use of Laboratory Animals with protocols approved by UCSF Institutional Animal Care and Use Committee (AN201121). Institutional Review Board protocols for tissue acquisition and organoid generation were approved at the Dana Farber Cancer Institute (Protocol 93–085) and University of California San Francisco (Protocols 23–39361 and 10–01375). Donors gave informed consent to have anonymized tissues used for scientific research purposes; patients were not compensated.

Mice

Mice were housed in the specific pathogen-free PSB facility (The University of California, San Francisco) under the following conditions: 12h:12h light: dark cycle, ambient temperature between 20–26°C, and humidity between 30–70%. Host FVB female mice for experimental metastasis assays were purchased commercially at 6 weeks of age and tail vein injection was performed at 8 weeks of age. Numbers of mice for each experiment included in the corresponding figure legend. Formalin-fixed paraffin embedded (FFPE) tissues from primary mammary tumors and lung metastasis analyzed in Extended Data Fig. 1 a–c were originally generated from MMTV-PyMT; CAG-Cre ER;Atg12F/F or Atg5F/F compound transgenic mice in⁷. For all cohorts analyzed, compound transgenic offspring were backcrossed to C57BL/6 mice for at least two generations, bred for homozygosity of Atg12 and Atg5, and tested for C57BL/6 congenicity (Jackson Laboratory). In these studies, the primary tumor size of orthotopically transplanted tumors did not exceed the maximum allowable tumor size (2.0 cm³) in the IACUC protocol. See Supplementary Table 2 for further details of mouse strains.

Cell Lines and Culture Conditions

Human cell lines used in this study and their source are described in Supplementary Table 2. The R221A cell line was isolated from a spontaneous mammary tumor in MMTV-PyMT transgenic mouse in the FVB/N background⁴⁹. All other cell lines were obtained from ATCC and maintained at 37°C in a humidified incubator with 5% carbon dioxide. R221A (from Barbara Fingleton⁴⁹, female), 4T1 (ATCC, CRL-2539, female), MDA-MB-231 (HTB26, female), HEK293T (CRL-3216, female) and Phoenix (CRL-3213, female) cells were maintained in DMEM supplemented with 10% FBS and 1% penicillin/streptomycin. MCF10A (female) cells were maintained in DMEM/F12 supplemented with 5% horse serum, 20 ng/ml EGF, 0.5 mg/ml hydrocortisone, 100 ng/ml cholera toxin, 10 mg/ml insulin, and 1% penicillin/streptomycin.⁵⁰. Human and mouse cell lines were authenticated annually using Short Tandem Repeat (STR) profiling and mouse primary cells using PCR genotyping. All cells were routinely tested via PCR (Sigma, MP0025) for Mycoplasma contamination.

For drug treatments, the following concentrations were used unless otherwise specified: Ulk kinase inhibitor (MRT 68921, Ulk-i) at 10nM in MDA-MB-231 cells and 1μM in R221A cells, hydroxychloroquine (HCQ) at 20 μM for all cell lines, and MG132 at 10 μM for all cell lines. Unless indicated, drug treatments were performed overnight (18h). For cycloheximide chase assays, CTRL, shATG7 or drug co-treated MDA-MB-231 cells (HCQ, MG132) were cultured in either DMSO or 50mg/ml cycloheximide for indicated times before harvesting for immunoblotting. To assess starvation-induced autophagic degradation of p62/SQSTM1 and ITCH, MDA-MB-231 were cultured in low serum media (DMEM with 0.2% fetal bovine serum) for 18 hours, either in the presence or absence of 20 μM HCQ for the final 6 hours to inhibit lysosomal function. See Supplementary Table 2 for further details of the cell lines and chemicals.

Human Organoid and Culture Conditions

Specimens were obtained from the UCSF Bakar Cancer Hospital (San Francisco, USA) or Brigham & Women’s Hospital (Boston, USA) on the day of surgery, viably frozen as tissue pieces, or used to generate organoid cultures. Malignant effusions were taken from paracentesis or thoracentesis on the day of procedure. Each tissue was minced using scalpels and digested in a solution containing DMEM/F12 (Gibco; 11330), 2 mM GlutaMax (Gibco; 35050), 10 mM HEPES (Gibco; 15630), 50 U/mL Penicillin-Streptomycin (Gibco; 15070), and 1 mg/ml collagenase XI (Sigma-Aldrich; C9407). Tissue digestion was performed at 37°C with constant shaking for 1–2 hours. Malignant effusions did not undergo collagenase digestion but were treated with RBC Lysis Buffer (BioLegend; 420301) if red blood cells were visible, following manufacturer’s instructions. Cells pelleted by centrifugation at 110 × g for 3 min, further dissociated by sequentially pipetting with 10, 5, and 1-ml pipette tips, and recentrifuged. The resulting cell pellet was used directly to establish organoid cultures by embedding in basement membrane extract type II (Cultrex; 3533-005-02), allowing this to harden at 37°C for 20 minutes to form a hydrogel dome, and then overlaying this dome with Type 1 Organoid Medium²⁷.

Organoids were split when previously cultured organoids become confluent, taking up ~80% of the hydrogel dome, approximately every 2–4 weeks. The medium was aspirated and using a bent pipet tip, 500 μl of TryplE Express (Gibco; 12604–013) was added to the organoid well, and vigorously triturated until the hydrogel dome was completely broken apart. Organoids were then incubated at 37°C for 2–15 minutes until visible disassociation and contents of the organoid well were transferred to a 15 ml conical. The well was washed 3 times with solution containing DMEM/F12 (Gibco; 11330), 2 mM GlutaMax (Gibco; 35050), 10 mM HEPES (Gibco; 15630), 50 U/mL Penicillin-Streptomycin (Gibco; 15070) and the washed contents were combined in the same 15ml conical. 200μl of fetal bovine serum (Cytiva; SH30910.02) was added and the culture was centrifuged at 110 × g for 3 minutes to create a cell pellet. For re-culturing, the cell pellet was resuspended in 50 μl of basement membrane extract type II (Cultrex; 3533-005-02) and transferred into a 24 well non-TC treated plate (CytoOne; CC7672–7524), to create a hydrogel dome. Subsequently, the dome was hardened at 37°C for 20 minutes, and overlaid with Organoid Medium²⁷. Alternatively, TryplE Express was used to disperse organoid cells immediately prior to treatment overnight (18h) with the pharmacological autophagy inhibitors MRT 68921 (Ulk-I, 1μM) or hydroxychloroquine (HCQ, 20 μM). Following treatment, disassociated organoid cells were cyto-spun onto glass slides at 110 × g for 3 min on a Cytospin 4 apparatus (Thermo Scientific) and subject to immunofluorescence staining.

Experimental Metastasis Assays

For tail-vein injection, polyoma middle T mammary tumor-derived R221A cells (1 × 10⁶) stably transduced with the indicated lentiviral ectopic expression constructs or shRNAs were resuspended in 150 μl PBS and injected into 7–8 weeks-old female FVB/N mice via lateral tail-vein. Unless otherwise indicated, animals receiving tail vein injections of R221A cells were euthanized at 2 weeks post-injection.

Tissue processing, Ex Vivo Lung Imaging and Quantification of Metastasis

Whole lungs of experimental mice were harvested and incubated in 3.7% aqueous buffered zinc formalin for 24h at RT before being transferred to 70% ethanol, processed, paraffin-embedded, and cut at 5μM for mounting onto slides. Slides were stained with H&E by the UCSF Helen Diller Family Comprehensive Cancer Center Pathology core (San Francisco, CA). For immunofluorescence (IF) staining of tissues, paraffin-embedded sections were subject to antigen heat retrieval with citrate buffer (pH = 6.0) in a decloaking chamber, blocked in 3% hydrogen peroxide for 30 minutes. These slides were subject for immunofluorescence staining and imaging according to the methods described below. For histological quantification of pulmonary metastasis, hematoxylin and eosin (H&E) stained sections of tumor bearing lungs from all mice in each experiment were quantified. H&E stained sections were scanned using an Aperio XT Whole Slide Scanner (Leica) and all metastases from 2–3 lung lobes were manually annotated using Aperio ImageScope software. Metastasis number and size were enumerated in individual mice. The average number of metastases was quantified from lung sections containing all the lobes of an individual mouse lung. To calculate the size of individual metastases, the average diameter of each macro-metastatic focus was measured and used to calculate lesional area. At least ten metastases were evaluated per mice per H&E section.

Quantitative RT-PCR

Quantitative RT-PCR was performed using New England BioLabs Luna^® Universal qPCR Master Mix on the StepOnePlus Real-Time PCR system (Applied Biosystems) in accordance with manufacturers’ instructions. Fold change was determined using the ΔΔCT method normalized to housekeeping gene, β-Actin. Mean of fold value was determined from 3–4 individual experiments per condition for each of the genes. Primers used for RT-qPCR detained in Supplementary Table 2.

Lentiviral shRNA Constructs, Virus Preparation and Infection of Cell Lines

All predesigned lentiviral shRNA expression plasmids (Millipore-Sigma) originated from The RNAi Consortium (TRC) in the pLKO.1-Puro lentiviral expression vector. The TRC number of all shRNAs are listed in Supplementary Table 2. Two independent shRNAs were used for each gene target. shRNAs were received as stabs in Luria agar and cultured on Luria agar plates containing 100 μg/mL ampicillin. Single colonies were isolated for plasmid preparation (Qiagen). Sanger sequencing confirmation was performed over the hairpin sequence for all lentiviral shRNA expression constructs. Empty pLKO.1 vector or gene specific pLKO.1-shRNA vectors were co-transfected into Phoenix cells with pVSV-G and pCMV 8.2-deltaR helper plasmids using FuGENE HD according to the manufacturer’s protocol. Virus-containing conditioned medium was harvested 48 h after transfection, filtered, and used to infect recipient cells. 1:5 (for MCF10A) or 1:10 (for all other cell lines) dilutions of lentivirus-containing medium were added to experimental cell lines in the presence of 5 μg/mL polybrene (Millipore). To enhance the infection efficiency, cells were spin-infected at 687 × g for 2.5 h followed by incubation at 37C in a 5% CO₂ tissue culture incubator. The virus-containing medium was replaced with normal culture media at 24h post-infection. When selection was utilized, cells were incubated for another 24–48h prior to addition of puromycin (Sigma; 1–4 μg /ml final concentration).

Site Directed Mutagenesis and Cloning

All plasmids and vectors are listed in Supplementary Table 2. Cloning of pLenti-Blast-CMV-3xFLAG empty and wild type NBR1 (WT NBR1) constructs were performed using pLenti-Blast-CMV empty vector (Addgene). NBR1 mutants were generated by site directed mutagenesis of pLenti-Blast-CMV-3xFLAG-WT-NBR1 and ITCH deletion mutants were generated by site-directed mutagenesis of pLV-ITCH-GFP (Sino Biologicals, HG11131-ACGLN). To perform mutagenesis, Agilent QuikChange XL Site-Directed Mutagenesis Kit was used according to the manufacturer’s protocol. XL-10 gold cells were used for the transformation and selection of mutagenesis products to ensure high efficiency and integrity of the lentiviral constructs. Invitrogen One Shot^™ Stbl3^™ Chemically Competent E. coli was used for all other bacterial transformation to ensure integrity of the lentiviral constructs. Primers used for site directed mutagenesis and sequencing can be found in Supplementary Table 2. The remaining constructs were obtained from Addgene or Sino Biologicals as specified in the Supplementary Table 2.

Flow cytometric analysis

MDA-MB-231 and R221A cells expressing shATG7 or treated with pharmacological autophagy inhibitor, either hydroxychloroquine (HCQ) or MRT68921 (Ulk-i), were immunostained for p62/SQSTM1, either alone or in combination with TP63 using the protocols described for immunofluorescence staining below. 15,000–20,000 stained cells per experimental condition were analyzed using a BD LSR Fortessa cell analyzer and BD FACSDiva software (BD Biosciences). The following antibody concentrations were used: p62/SQSTM1 primary (1:1000), TP63 (1:1000), Alexa Fluor tagged secondary antibodies (1:10,000). Flow cytometric analysis was repeated three times per cell line.

Immunoprecipitation and Immunoblotting

Experimental cells were either treated, transfected or transduced as specified in each individual experiment followed by selection with 1–4ug/ml puromycin (Sigma). Total cell lysate was harvested from indicated experimental cells in RIPA buffer (Pierce) supplemented with protease or protease/ phosphatase inhibitor cocktails (Pierce) at 4°C. Lysate was centrifuged at 16,000 × g to remove insoluble material. Protein concentrations of the total soluble lysates were measured using BCA protein assay kit (Pierce) according to the manufacturers protocol. For immunoprecipitation, cells were harvested in IP Lysis buffer (Pierce) supplemented with protease and phosphatase inhibitor cocktail (Pierce) at 4 °C. For immunoprecipitation with GFP or other protein specific antibodies, Protein-A (for rabbit antibodies) or Protein-G (for mouse antibodies) conjugated Dynabeads (Thermo Fisher) were blocked in 0.1% BSA for 1 hour and then incubated with the indicated primary antibodies for 1h at 4°C with gentle mixing. α-FLAG M2 magnetic beads (Sigma) were used for immunoprecipitation of FLAG-tagged proteins, including mass spectrometric analysis of 3XFLAG-NBR1 (further detailed in Mass Spectrometric Analysis section below). Antibody-conjugated beads were then centrifuged and washed with cell lysis buffer before adding to the cell lysate. Immunoprecipitation was performed overnight with gentle mixing at 4°C, followed by centrifugation and washing of the beads with lysis buffer. Equal amounts of protein from total cell lysates and immune-precipitates were denatured with Bolt Sample Reducing Agent in NuPAGE LDS Sample Buffer at 94°C for 10 min. Polyacrylamide gel electrophoresis was performed using Bolt 4–12% Bis-Tris Plus gradient gels. Resolved proteins were transferred onto PVDF membranes using 1X Tris-Glycine buffer containing 20% methanol for 90 minutes at 100 constant volts. Membranes were blocked in 5% milk or bovine serum albumin in PBS with 0.1% Tween-20 (PBST) for 30 minutes at 20°C prior to overnight incubation with indicated primary antibody at 4°C in blocking solution. Membranes were washed in PBST prior to incubation with HRP-conjugated secondary antibodies (1:5,000) in blocking solution for 2 hours at room temperature, washed again with PBST and developed via Immobilon Forte Western HRP substrate. A complete list of antibodies used for immunoprecipitation and immunoblotting and their source is provided in the Supplementary Table 2.

Immunofluorescence, Microscopy and Live Cell Imaging

For immunofluorescence, cells were seeded on glass coverslips and transfected or transduced with lentiviral shRNAs or treated with drugs as specified in each individual experiment. After 48–72 hours following transduction, 15–18 hours following drug treatment or 10–60 seconds following either 5% 2,5-hexanediol or 5% 1,6-hexanediol treatment, cells were fixed for immunofluorescence staining. Cells were fixed with 10% neutral-buffered formalin, permeabilized with 0.2% Triton X-100, pre-incubated with 5% BSA in PBS, and then incubated with indicated primary antibodies overnight at 4°C. Cover slips were washed, incubated with fluorescent-conjugated secondary antibodies (Invitrogen) for 2 hours at room temperature, followed by counterstaining and mounting on glass slides with Vectashield Mounting Medium containing DAPI (Vector Labs). Antibodies used for immunofluorescence and their source listed in Supplementary Table 2.

Fluorescent imaging was performed on a Leica TCS SP8 laser scanning confocal microscope (DMi8 platform) using Leica Application Suite X software and Leica PCO Edge 5.5 sCMOS camera. Imaging was performed using either 100×/1.44 oil immersion HC Plan Apo objective lens or 63×/1.40 oil-immersion HC Plan Apo CS2 objective lens or 25x water-immersion HC Plan Apo objective lens. Excitation used Leica Express UI white light lasers, and multidimensional image acquisition was performed using a combination of PMT and HyD detectors for multiple fluorescent dyes. Sequential scans were performed between frames to minimize fluorescence bleed-through between fluorophores. A pinhole value of 1.0–1.6 was used for imaging. Images were acquired in 1024 × 1024-pixel format, and 16–32 line-averages or 3–8-frame averages were recorded at moderate speed (200–400). The number of cells, puncta or events quantified are specified in each individual figure legend. All immunofluorescence experiments were performed with at least three independent times, unless otherwise specified in the figure legends.

For live cell imaging to detect fission/fusion events or fluorescence recovery after photobleaching (FRAP), MDA-MB-231 cells were infected or transfected with mCherry-p62/SQSTM1 and or EGFP-ITCH. 24–28 hours following transfection or transduction, cells were trypsinized and re-seeded on glass-bottom dishes for microscopy. To detect ITCH or p62 fission/fusion events, time-lapse images were acquired using a Nikon Ti motorized inverted microscope with Perfect Focus System and a Yokagawa CSU-W1 Spinning Disk confocal. Images were magnified using a Nikon 63x oil immersion objective with a numerical aperture of 1.49 and working distance of 0.12mm at super resolution (4x). GFP-ITCH was excited with a 488nm (8% of 100mW) laser and collected with a 510nm filter. mCherry-p62 was excited with a 561nm (6% of 100mW) laser line. Both fluorophores were excited and collected simultaneously collected with a Hamamatsu ORCA-FusionBT CMOS dual camera system. A single plane image was collected every second for 3 minutes with an exposure time of 300ms and 1×1 camera binning. During image acquisition, all samples were in an enclosure kept at 37C and 5% CO₂. All images were registered and deconvolved using Nikon Elements (AR 5.30.03). The brightness and contrast of the images were adjusted using Fiji (ImageJ).

For studying kinetics of FRAP recovery of ITCH or p62/SQSMT1 puncta, imaging was performed on a Nikon Ti microscope equipped with a CSU-W1 spinning disk confocal (Yokogawa), Andor solid-state lasers, an Andor DU-888 EMCCD camera, and a Plan Apo VC 100× 1.4 NA objective (Nikon), controlled by micro-manager (https://micro-manager.org/). Cells were kept at 37°C in a temperature- and humidity- controlled chamber (Okolab, San Bruno, CA) during imaging. A Rapp Optoelectronic UGA-40 photobleaching system was used to photo-bleach the fluorescence protein in point-shaped ROIs using a 405nm laser (Vortran, Sacramento, CA). Intensity of each puncta in each of the time frames were quantified by using Fiji (ImageJ). Rate of recovery of fluorescence of ITCH (green) and p62/SQSTM1 (red) puncta in each time point following photobleaching were calculated by comparison to the fluorescence intensity of the puncta before photo-bleach. Recovery of fluorescence was scored at 10-seconds interval. (N=26 for GFP-ITCH and N=22 for mCherry-p62).

Mass Spectrometric Analysis

For proteomic identification of NBR1 interacting proteins, HEK-293T cells were transduced with 3XFLAG constructs (control, WT-NBR1 or D50R-NBR1) followed by immunoprecipitation as described above. Following the last lysis buffer wash, anti-FLAG affinity purification beads (Sigma) were washed twice on ice with 1 mL of ice-cold buffer (20mM TrisHCl pH8 + 2mM CaCl₂) without protease inhibitors. A small portion of the beads (<5% total sample) was set aside for immunoblot and Coomassie staining. For direct visualization of the proteins in the NBR1 immunoprecipitates and total cell lysates samples were separated by SDS-PAGE followed by 0.1% Coomassie blue (Sigma) staining according to manufacturer’s protocol. A buffer of 50% methanol, 10% glacial acetic acids were used for staining and washing medium.

The remaining beads were briefly centrifuged to remove the residual wash buffer with a fine pipette. To elute the 3XFLAG tagged proteins, beads were first washed twice in washing buffer (0.05M Tris-HCl pH 7.5 + 0.15M NaCl + 0.001M EDTA + 0.05% NP40) followed by twice in the same wash buffer without NP40. Next, beads were eluted with 20uL of 100 mg/ml FLAG-peptide (Sigma) dissolved in elution buffer (0.05M Tris-HCl pH 7.5 + 0.15M NaCl + 0.001M EDTA + 0.05% Rapigest, Waters). The elution step was repeated twice and all the eluted proteins were combined. Next, 9.5 μL of 20 mM Tris-HCl, pH 8.0 and 0.5 μL of 100 mM DTT was added to the eluted volume and incubated at room temperature for 30 min with gentle agitation. Following the reduction step, samples were alkylated with 3mM iodoacetamide at 30°C with gentle agitation for 45 minutes in the dark. The alkylation reaction was quenched by adding 3mM DTT before proceeding to the digestion step. To perform digestion of the isolated proteins, 0.4 μg of MS grade trypsin and 0.4 μg of MS grade trypsin/Lys-C protease mixture (Thermo Fisher) was incubated overnight at 37°C on a shaker. To ensure efficient digestion, an additional volume (0.2 μg of MS grade trypsin and 0. μg of MS grade trypsin/Lys-C) of protease mixture was added in the morning and incubated further for 2h at 37°C on a shaker. Digestion was stopped by adding formic acid to a final concentration of 2%. The resulting samples were stored at −80°C until processed for mass spectrometric analysis as described below.

The resulting digested peptide samples were desalted using PreOmics iST desalting columns and associated wash buffers. The desalted peptides were then dried overnight in a Labconco CentriVap Benchtop Concentrator. For tandem mass tag (TMT) labelling, the dried peptide samples were resuspended in 15 μl of 50mM HEPES buffer (pH 8.5). 5 μl of 20 μg/μl TMT label (0.8 mg resuspended in 40 μl of acetonitrile) was then added to each sample. Samples were incubated with TMT label at room temperature for 1 h. Unreacted TMT label was then quenched by adding 2.5 μl of 5% hydroxylamine solution diluted in 50 mM HEPES buffer to each sample. Labelled sampled were then dried overnight to acetonitrile prior to high-pH fractionation. Labelled peptides were then resuspended and fractionated using Pierce High pH Reverse Phase Fractionation Kit (Pierce 84868) according to manufacturer instruction. The resulting fractions were then dried overnight. The fractions were then resuspended in 0.1% Formic Acid/3% Acetonitrile solution for analysis on the LC-MS. The peptide concentration in each fraction was quantified using the Pierce Quantitative Colorimetric Peptide Assay (Pierce 23275) and the concentration of the peptides was adjusted to 200 ng/μl. For LC-MS/MS analysis, 5 μl of each fraction (1 μg of peptides) was injected into a Dionex Ultimate 3000 NanoRSLC instrument with 15-cm Acclaim PEPMAP C18 (Thermo) reverse phase column coupled to a Thermo Q Exactive Plus mass spectrometer. Analysis of the mass spectrometry data was done using MaxQuant and the resulting data was further processed using Perseus. Complete mass spectrometric data is provided in Supplementary Table 1.

Bioinformatic Analysis of Single Cell RNA-Seq Data

Single cell RNAseq data from primary³² and metastatic breast cancer tumors³³ obtained from Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) using series GSE161529 and GSE186344, respectively, was used for the correlative analysis of autophagy biogenesis gene signatures with p63 target genes in single breast cancer cells. A 28 gene set described in⁷ was utilized to define the autophagy biogenesis gene signature. The Seurat pipeline was applied to each sample⁵¹. Raw counts of annotated cancer cells were normalized (Seurat NormalizeData function) and scaled (ScaleData function). Briefly, the average expression of autophagy gene sets was calculated on single cells, subtracted by the aggregated expression of control feature sets randomly selected using the AddModuleScore function with default parameters. The expression of selected p63 targets and autophagy biogenesis signature scores was then visualized in heatmaps using the R package ComplexHeatmap³⁴.

ColabFold (AlphaFold2) Protein Structure Prediction of the NBR1-p62/SQSTM1-ITCH Trimeric Complex

The structural prediction of the trimer complex comprising ITCH (N-terminus to WW4 domain), NBR1, and p62/SQSTM1 (N-terminus to ZnF domains) was conducted using the AlphaFold2 algorithm via ColabFold (v1.5.5), an open-access notebook supported by Google Colaboratory, which enables efficient protein structure predictions using pre-configured AlphaFold2 parameters^{40, 41}.

(https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb).

The job can be located at the following notebook address:

https://colab.research.google.com/drive/1py_oJrL9LQw4jKZda1cow21qA1sTQ6i8?authuser=1#scrollTo=ADDuaolKmjGW

Briefly, amino acid sequences obtained from UniProt were formatted as a single FASTA file, with each sequence separated by colons (:) to indicate their multimeric configuration, as recommended by ColabFold. Structural predictions were performed using the AlphaFold2-Multimer algorithm with the following settings:

• MSA Mode: mmseqs2_uniref_env for generating alignments from UniRef and environmental sequence databases.

• Template Detection: Enabled, using the pdb100 structural database.

• Model Type: alphafold2_multimer_v3, optimized for multimeric complexes.

• Number of Recycles: 3, with early stopping (recycle_early_stop_tolerance=0.5).

• Pairing Strategy: greedy, pairing taxonomically matching subsets from the same species while including unpaired MSAs.

• Relaxation: Atomic relaxation of the model was disabled (num_relax=0).

• Max MSA: Automatically determined by ColabFold.

• Number of Seeds: 1.

The algorithm automatically performed multiple sequence alignments (MSA) using MMseqs2 and modeled the trimeric protein complex, optimizing protein-protein interactions via Monte Carlo Tree Search (MCTS). Structural models were ranked based on the predicted Local Distance Difference Test (pLDDT) scores, a metric indicating model confidence. The top-ranked model Rank 1, exhibiting the highest pLDDT score, was selected for further evaluation. Predicted aligned error (PAE) matrices were used to assess the confidence of inter-chain interface predictions and binding sites. Structural visualization and analyses, including inter-chain interaction assessments and overall topology evaluations, were performed and visualized using ChimeraX (UCSF)⁵². All predictions were conducted on Google’s cloud infrastructure using a Tesla T4 GPU allocated within the Colaboratory environment.

Statistics and Reproducibility:

All experiments were performed with at least three independent biological replicates unless otherwise specified. Each replicate was derived from separate cell culture preparations or animals to ensure biological variability. Details of quantification of each experiment has been provided throughout the methods. Statistical analysis was performed using GraphPad Prism or Microsoft Excel. All error bars shown represent standard error of the mean. Data distribution was assumed to be normal but this was not formally tested. All statistical information including the statistical methods, the P-values and number (N) of events, puncta, condensates, or cells scored or tissues or animals analyzed for each individual experiment are detailed in the figures and corresponding figure legends. No data were excluded from analyses. Sample size and statistical analysis performed in each and every graphical representation throughout the manuscript is detailed in the respective figure legends for Figures and Extended Data Figures. Source data is provided for the numeric values used to derive each graphical representation.

For experimental metastasis assays, host mice were randomized into cohorts for tail vein injection with the indicated cell types. For animal studies, the experiments were designed to detect differences in means of greater than 50%, with the standard deviations of the means typically at 30–40%. For power calculations, the alpha value was set to 0.05 and assuming a normal data distribution, sample sizes of 8–12 usually allowed for statistical power of 0.8 to detect differences of at least 1.3*SD. For binary outcomes, the minimum detectable difference in proportions is 0.45. Effects of such magnitude were observed in previously published studies from our lab⁷.

Extended Data

Extended Data Fig. 1. p62/SQSTM1 Promotes Basal Differentiation Upon Autophagy Inhibition In Vivo.

a, Primary or metastatic tumor tissues generated from MMTV-PyMT;CAG-Cre^ER;Atg12^F/F or MMTV-PyMT;CAG-Cre^ER;Atg5^F/F tumors in Marsh et al.⁷ and stained for p63 (green), p62/SQSTM1 (red) and nuclei (DAPI, blue). Representative individual cells in boxed areas enlarged in Extended Data Fig. 1a-i (p62 high, p63 negative), 1a-ii (p62 high, p63 positive) and 1a-iii (p62 low, p63 positive) and enumerated in Extended Data Fig. 1b–c. Vehicle=autophagy-competent primary tumors or metastases, OHT=autophagy-deficient primary tumors and TAM= autophagy-deficient metastases. b, Quantification of tumor cells positive for p62. 600 or more tumor cells enumerated from each of four individual animals. Vehicle=autophagy-competent primary tumors or metastases, OHT=autophagy-deficient primary tumors and TAM= autophagy-deficient metastases. c, Quantification of p62 levels (high versus low) and p63 positivity in tumor cells from autophagy-deficient tumors. 600 or more tumor cells were enumerated from each of four individual animals. d, Number of metastases per section of lung in shCTRL (n = 10 mice), shATG7 (n = 22), shATG7 + shp62, #1 (n = 13), shATG7 + shp62, #2 (n = 13). e-f, Representative images of lung metastases stained for CK14 (green, e), p63 (green, f) and nuclei (DAPI, blue). Certain images in Extended Data Fig. 1f (e.g., shATG7 and shATG7 + shp62, #1) also shown in Fig. 1d. g, Quantification of p63 (nuclear) positive and CK14 positive cells in the lung metastases for the indicated conditions. 250 or more cells from 4–8 metastases counted per cohort. 5 individual mice from three independent experiments evaluated. h, Immunoblot for the indicated proteins in MCF10A cells expressing the indicated shRNAs.. Samples run on separate gels with equal loading of total proteins. i, PyMT R221A cells immunostained for p63 (red), p62 (green) and nuclei (DAPI, blue) for the indicated conditions. Statistics: Statistical significance was determined by either a 2-tailed t test for (b) and (c) or by one-way ANOVA followed by Dunnett’s multiple comparisons test for (d) and (g). Data are represented by mean ± S.E.M; each data point represents one mouse in all graphs. Data and unprocessed blots available as source data.

Extended Data Fig. 2. Autophagy Inhibition Promotes Basal Differentiation in Mouse Mammary Cancer and Human Breast Cancer Cells.

a, CK14 (red), CK5 (green) and nuclei (DAPI, blue) immunostaining of mouse PyMT R221A cells expressing the indicated shRNAs. b, Representative flow cytometry profile of p62/SQSTM1 positive (red) and p62, p63 double positive (yellow) MDA-MB-231 cells. c, Dual reporter autophagic flux probe described in Kaizuka et al²⁶. d, Representative immunofluorescent staining of MDA-MB-231 cells for p63 (light blue) for the indicated cohorts. Expression of GFP (green) and RFP (red) in individual cells expressing dual reporter autophagic flux probe. Arrows mark representative p63 positive cells. Right offset: Merged image with magnification corresponding to boxed area. e, Quantification of p63 positive cells in Extended Data Fig. 2d. Each data point indicates the percent of positive cells calculated from one individual experiment (n=3 independent experiments). 200 or more cells were enumerated per experiment per condition. f, Summary of human prophylactic mastectomy (PM) and breast cancer (BC) organoids used in this study. g, Representative immunofluorescent images of primary cells generated from human prophylactic mastectomy (PM) organoids (Org 164) and treated with solvent, Ulk-i or HCQ for 18h. Cells immunostained for CK14 (red) and nuclei (DAPI, blue) following treatment. h, Quantification of human BC organoid cells positive for p63 and CK5 in Extended Data Fig. 2h. 250 or more cells were counted from each of three individual organoids (N=3). i, Representative immunofluorescent images of primary cells generated from human breast cancer (BC) organoids (Torg 40) and treated with solvent, Ulk-i or HCQ. Following treatment, cells immunostained as indicated for the following markers: CK5 (red, top row), p63 (green, top row), p62 (green, bottom row), and nuclei (DAPI, blue). Scale bar equals 20 μm. Right: Magnification of p62 immunostaining in representative control or Ulk-i treated cells. i, Statistics: Statistical significance determined by a 2-tailed t test in (e) and (h). Data are represented as mean ± S.E.M. Each dot represents percentages calculated from an individual experiment (e) or a human patient organoid (h). Data and unprocessed blots available as source data.

Extended Data Fig. 3. p62/SQSTM1 is Required for The Effects of NBR1 Overexpression on Tumor Metastasis and Pro-Metastatic Basal Differentiation.

a, Immunofluorescent staining for p62 (green) and CK5 (red) in MDA-MB-231 cells expressing indicated shRNAs and, when indicated, treated with Ulk-i. Offset: Magnification of boxed area in Ulk-i+shCTRL cells. b, Immunoblot analysis of MDA-MB-231 and R221A cells from indicated cohorts. Loading control (actin) corresponds to same blot. c, Immunofluorescent staining of indicated MDA-MB-231 cell cohorts for FLAG (NBR1, green), p62 (red) and nuclei (DAPI, blue). d, Quantification of total number of metastases per section of lung for shCTRL (n = 7), WT-NBR1 (n = 6), WT-NBR1 + shp62, #1 (n = 8), wt-NBR1 + shp62, #2 (n = 9). e, Representative H&E staining of lung metastases from indicated groups. f-g, Lungs bearing metastases from indicated groups stained for p63 (green, f) or CK14 (green, g) and nuclei (DAPI, blue). h, PyMT R221A cells expressing the indicated constructs stained for p63 (green, left panel) and CK14 (green, right panel) and nuclei (DAPI, blue). i, Quantification of p63 (nuclear staining) and CK14 positive R221A cells. 400 or more cells from each of three individual experiments were counted per condition. j, Quantification of total number of metastases per section of lung in shCTRL (n = 12), WT-NBR1 (n = 11), D50R-NBR1 (n = 10). shCTRL and WT NBR1 cohorts were also included as controls in Extended Data Fig. 3d and 7a. k, Lungs bearing metastases from the indicated groups stained for p63 (green, top), CK14 (green, bottom) and DAPI (blue). l, Quantification of p63 (nuclei) and CK14 positive cells within lung metastases for the indicated conditions. Minimum of 250 cells from multiple (4–8) metastases counted per cohort. A total of five (N=5) individual mice from three independent experiments evaluated. Statistics: Statistical significance determined by either one-way (d, j and l) or two-way (i) ANOVA followed by Dunnett’s multiple comparisons test. Data are represented by mean ± S.E.M; each dot represents one animal (d, j and l) or an individual experiment (i). Data and unprocessed blots available as source data.

Extended Data Fig. 4. Autophagy Inhibition Promotes Transcription of p63 Target Genes and p63 Protein Stability.

a, qRT-PCR analysis of p63 and p63 target mRNA expression in mouse 4T1 cells for the indicated groups. b-c, qRT-PCR analysis of p63 mRNA expression in human triple-negative MDA-MB-231 cells (b) and mouse PyMT R221A (c) cells for the indicated groups. d, Quantification of p63 protein levels in MDA-MB-231 cells following cycloheximide treatment. Following ATG7 knockdown or 18h treatment with indicated agents, cells were treated with cycloheximide for indicated times, lysed, subject to immunoblot analysis and integrated band intensity was calculated via densitometry. For each condition and timepoint, integrated band intensity of p63 protein was normalized to both its respective actin control and the corresponding t=0 timepoint. e, Representative immunoblot from cycloheximide chase assays of MDA-MB-231 cells used for quantification in Extended Data Fig. 4d. MG132 employed as a control to inhibit proteosomal degradation of p63. Samples were run on separate gels with equal loading of total proteins. f, Representative immunofluorescent staining of CK14 (red) and nuclei (DAPI, blue) in lung metastasis from indicated groups. g, Phase-contrast micrograph of human MDA-MB-231 and mouse PyMT R221A cells 24-hours following treatment with CTRL or p63 shRNAs. h, Working model: Autophagy inhibition in breast cancer cells results in accumulation of the ACRs NBR1 and p62/SQSTM1, which in turn, increases protein levels of the p63 transcription factor and enhanced expression of p63-target genes. Statistics: Data (a-c) are represented by mean ± S.E.M. Each dot represents a technical replicate within one bio-replicate; one representative of three bio-replicates are shown. Each data point in (d) represents mean band intensity ± S.E.M from three (N=3) independent experiments. Data and unprocessed blots available as source data.

Extended Data Fig. 5. NBR1-p62/SQSTM1 Complexes Sequester the Ubiquitin Ligase ITCH in Autophagy Deficient Cells.

a, Coomassie staining of total cell lysates from indicated cells utilized for TMT-MS in Fig. 4c and Extended Data Fig. 5b. b, Volcano plot of proteins identified by TMT-MS from α-FLAG immunoprecipitates from HEK293T cells expressing WT NBR1 vs. D50R (both 3XFLAG-tagged). Proteins plotted according to −log₁₀ P values as determined by two-tailed t-test and log₂-fold enrichment (n=3 biological replicates). Red: proteins significantly enriched in WT NBR1, identified with P>0.05 and log₂-fold change>|0.5|. Blue: significantly enriched in D50R. Grey: not enriched in either group.. c, Immunoblot of α-FLAG immune complexes (lanes 1–3) and total cell lysates (lanes 4–6) from 293T cells expressing indicated constructs. Samples run on separate gels with equal loading; loading control (actin) corresponds to the NEDD4 blot. d-e, Immunofluorescent staining of 293T cells for FLAG (NBR1, green), nuclei (DAPI, blue) and TAX1BP1 (red) in (d) or ITCH (red) in (e) for the indicated conditions. f, Immunoblot analysis of total cell lysates of MDA-MB-231 cells used for the immunoprecipitation studies in Fig. 4d. g, Immunoblot analysis of total cell lysate of MDA-MB-231 cells for the immunoprecipitation studies in Fig. 4e. Actin loading control in (f) and (g) corresponds to same blot. h, Quantification of percent colocalization (relative to total number of puncta) of FLAG (NBR1) and ITCH puncta in MDA-MB-231 and MCF10A cells. For each condition, a minimum of 500 puncta from three individual high-power fields (N=3) representing at least 30 individual cells were evaluated. i, Representative immunofluorescent staining for ITCH (red), FLAG (NBR1, green) and nuclei (DAPI, blue) of MDA-MB-231 cells expressing the indicated constructs and shRNAs. Statistics: Each data point (h) represents percent of enumerated puncta colocalized per independent fields. Data represented by mean ± S.E.M. Statistical significance was determined by two-way ANOVA followed by Dunnett’s multiple comparisons test. Data and unprocessed blots available as source data.

Extended Data Fig. 6. Dynamic Properties of p62 and ITCH in Biomolecular Condensates.

a, Immunoblotting of MDA-MB-231 cells expressing indicated FLAG-tagged NBR1 constructs. α-FLAG immune complexes (upper) and total cell lysates (lower) blotted for the indicated proteins. Samples run on separate gels with equal loading of immune complex (upper) or total proteins (lower). For α-FLAG immune complexes, NBR1 loading corresponds to p62 gel; for total lysate, loading control (actin) corresponds to the p63 gel. b, MDA-MB-231 cells expressing indicated ITCH constructs (GFP-tagged) immunostained for ITCH (green), p62/SQSTM1 (red) or DAPI (blue). c, Representative immunofluorescent staining of MDA-MB-231 cells for p62/SQSTM1 (green) and ITCH (red) for the indicated treatment groups. d, Representative selected frames of GFP-ITCH (green) and mCherry-p62 (red) puncta in MDA-MB-231 cells for the indicated time points. Frames demonstrate course of fission event of multiple (#1– #3) green (ITCH) and p62 (red) puncta; point of fission is indicated by arrow. Selected frames from this sequence of images are also shown in Fig. 7e. Also see Supplementary Video 3. Certain images in Extended Data Fig. 6d are also shown in Fig. 7e. e, Representative selected frames of GFP-ITCH (green) and mCherry-p62 (red) puncta in MDA-MB-231 cells for the indicated time points. Frames demonstrate course of fusion event of ITCH (green, #1) and p62 (red, #2) puncta; point of fusion indicated by arrow. Selected frames of this from this sequence of images are also shown in Fig. 7f. Also see Supplementary Video 4. Certain images in Extended Data Fig. 6e also shown in Fig. 7f. Unprocessed blots available as source data.

Extended Data Fig. 7. Depletion of ITCH Triggers p63 Mediated Basal Differentiation and Metastatic Outgrowth of Mammary Tumors.

a, Quantification of total number of metastases per section of lung for CTRL (n = 12 mice), WT-NBR1 (n = 11) and del-ZnF-NBR1 (n=10). shCTRL and WT NBR1 cohorts were also used as controls in Extended Data Fig. 3d. b, Quantification of total number of metastases per section of lung for shCTRL (n = 13 mice), shATG7 (n=6), shITCH, #1 (n = 25), shITCH, #2 (n=24), shKEAP1, #1 (n=12) and shKEAP1, #2 (n=12). Mice in shCTRL and shATG7 cohorts are also used in Extended Data Fig. 1d. Related to Fig. 8g–l. c-d, Control (shCTRL) and ITCH depleted MDA-MB-231 cells immunostained for p63 (green) and CK14 (red) in (c) or for p62 (green) and CK5 (red) in (d). Nuclei counterstained with DAPI (blue). e, Immunoblot analysis of MDA-MB-231 cells for the indicated proteins (left), 48h following treatment with CTRL, NEDD4 or ITCH shRNAs. Samples were run on separate gels with equal loading of total proteins; loading control (actin) corresponds to ITCH blot. f, Model for breast cancer cells with normal autophagy (top) versus cells with autophagy inhibition (bottom). Normal autophagic degradation of ACRs reduces steady state levels of p63 due to ITCH-mediated UPS turnover. Inhibiting autophagy results in accumulation of NBR1-p62/SQSTM1 complexes, which sequester ITCH away from p63. As a result, stabilized p63 promotes the increased expression of basal cytokeratins and increased metastasis. Statistics: Each dot (a, b) represents one animal. Statistical significance (a, b) was determined by one-way ANOVA followed by Dunnett’s multiple comparisons test. Data are represented by mean ± S.E.M. Data and unprocessed blots available as source data.

Data Availability

The mass spectrometry proteomics data associated with this study have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the identifier PXD047557.

Published scRNA sequencing data^{32, 33} was obtained from Gene Expression Omnibus sources GSE161529 and GSE186344. Source data are provided with this paper. Other information related to the data reported in this paper and requests for reagents can be directed to the lead contact, Jayanta Debnath (Jayanta.Debnath@ucsf.edu).