2.1 Patient specimens

TNBC specimens were collected from patients who had undergone taxane-based neoadjuvant or adjuvant chemotherapy. All patients were diagnosed at Sun Yat-sen University Cancer Center. The study included 26 patients treated with neoadjuvant chemotherapy and 157 patients treated with adjuvant chemotherapy. Detailed clinical and pathological characteristics are provided in the Supporting Information Tables.

2.2 Immunohistochemistry

Immunohistochemical (IHC) staining was carried out on the TNBC tissue sections included in the study, utilizing anti-ALYREF and anti-RNF31 antibodies, as previously described.³³ The integrated optical density (IOD) of the IHC staining was measured using Image-Pro Plus Version 6.2 software (Media Cybernetics Inc.), with subgroup classification based on the median IOD value. Samples with IOD values equal to or exceeding the median were classified as the high-expression group, while those with IOD values below the median were assigned to the low-expression group. The full process of IHC staining is provided in the Supporting Information Materials.

2.3 Xenograft tumour models and treatment

Immune-deficient nude mice were used to examine the in situ growth of MDA-MB-231 tumours. Mice were housed in barrier facilities on a 12 h light/dark cycle. The indicated cells (1 × 10⁶ cells/mouse) were injected into the mammary fat pads of nude mice (n = 6/group) to establish the orthotopic xenograft models. At 14 days after tumour induction, when the median tumour volumes reached approximately 200 mm³, dimethyl sulfoxide (DMSO), paclitaxel (PTX) (10 mg/kg), and Thiolutin (5 mg/kg) were intraperitoneally (i.p.) injected according to the indicated groups. Paclitaxel was added every 2 days for six times in total, and Thiolutin was added every day for five times in total. Tumour volumes were measured every 3 days and calculated using the following formula: tumour volume = 1/2 × length × width². Once tumour-bearing mice lost 10% of their weight, experienced cachexia, or the tumour volumes reached 1500 mm³, the mice were sacrificed, and the tumours were excised for further examination.

2.4 Patient-derived organoid culture

Patient-derived organoid cultures were established from histologically confirmed TNBC tissues obtained from biopsy specimens at Sun Yat-sen University Cancer Center, with written consent provided by the patients before surgery. Tissue processing and organoid culture were performed as previously described.⁴² In brief, freshly resected tumour tissues were promptly processed upon receipt and cultured. Depending on the tumour size, the untrypsinized organoids were embedded in Matrigel. Once the Matrigel-cell solution solidified in the 48-well plates, the breast cancer organoid medium was added immediately and refreshed every 2 to 3 days, as appropriate. After 2 to 4 weeks of growth, TNBC organoids were harvested for quantitative real-time reverse transcription PCR (qRT-PCR) and western blotting. Trypsinization and passaging of organoids were carried out approximately every 14 to 21 days.

2.5 Drug response test of TNBC organoids

Organoids exhibiting high RNF31 expression and optimal conditions were selected for the drug sensitivity test. Of the five organoids analyzed, two (#1 and #2) were identified as having high RNF31 levels and were chosen for further drug response testing. PTX was administered at a concentration of 10 nM, while Thiolutin was applied at a concentration of 1 µM. Organoid images were captured using a Nikon Eclipse Ti2-E microscope (Nikon). Cell viability was assessed using the CellTiter-Glo 3D cell viability assay (Promega; G9683), following the manufacturer's protocol.

2.6 Statistical analysis

The SPSS software (version 25.0, IBM Corp.) was applied to perform all statistical analyses. Differential comparisons were carried out using Student's t-test (two-tailed), the Mann–Whitney U test (two-tailed), the chi-squared test, one-way analysis of variance (ANOVA), or one-way repeated measures ANOVA test, as appropriate. Survival comparisons were evaluated using the Kaplan–Meier survival method and a log-rank test. Independent risk factors for survival were assessed using the Cox regression models. A p-value of less than .05 was considered statistically significant.

3.1 ALYREF trafficking correlates with the response to PTX in triple-negative breast cancer

Previous studies reported that ALYREF shuttles between the nucleus and the cytoplasm.^{19, 43} However, the regulatory mechanisms of ALYREF function and its subcellular distribution in humans are not fully understood. To identify the importance and biological function of ALYREF in the treatment response of patients with TNBC, we first investigated the impact of ALYREF on different chemotherapy agents in TNBC cells. First, we established small interfering RNA (siRNA)-mediated knockdown of ALYREF in TNBC cell lines, MDA-MB-231 and SUM159PT. The knockdown efficiency was validated using qRT-PCR and western blotting (Figure S1A). Cell viability assays using cell counting kit-8 (CCK8) were performed to evaluate cell viability in response to treatment with different chemotherapeutic drugs (epirubicin (EPX), cyclophosphamide (CTX) and PTX)² in ALYREF knockdown cells. The results showed that ALYREF knockdown significantly reduced cell viability upon PTX treatment in MDA-MB-231 and SUM159PT cells (Figure 1A). However, the effect of ALYREF knockdown on EPX and CTX cytotoxicity was less significant, as the cell viability was only slightly decreased in ALYREF knockdown cells in comparison with that in the scramble siRNA control cells (Figure S1B). In addition, a flow cytometry-mediated Annexin V assay showed that silencing of ALYREF significantly increased the apoptosis rate upon PTX treatment in both MDA-MB-231 and SUM159PT cells (Figure 1B; Figure S1C). To further confirm the role of ALYREF in PTX efficacy, we generated PTX-resistant MDA-MB-231 and SUM159PT cell lines (MDA-MB-231-R and SUM159PT-R; Figure 1C) and detected ALYREF expression in PTX-resistant and parental cells (MDA-MB-231-P and SUM159PT-P). Unexpectedly, both the mRNA and protein levels of ALYREF were comparable between PTX-resistant and parental TNBC cells (Figure S1D), indicating that it was not the expression level of ALYREF, but possibly post-translational processing, which might be involved in the PTX response in TNBC. We then detected ALYREF using an immunofluorescence assay in TNBC cells. As shown in Figure 1D and Figure S1E, in parental TNBC cells, ALYREF was primarily located in the nucleus, while in the presence of PTX, ALYREF was trapped in the cytoplasm. Importantly, in PTX-resistant TNBC cells, ALYREF was mainly localized in the nucleus. Consistently, the distribution changes of ALYREF were validated by subcellular fraction extraction and subsequent western blotting: the nuclear-localized ALYREF was reduced upon PTX treatment in parental cells but recovered to a level comparable to that in untreated cells in PTX-resistant cells (Figure 1E). To validate the effect of ALYREF distribution on PTX resistance, we detected ALYREF distribution in 26 tumour tissues collected from patients with early-stage TNBC treated with PTX-based neoadjuvant chemotherapy. Tumour tissues were obtained after systemic treatment. Immunohistochemical (IHC) staining showed that the ALYREF level was slightly increased in chemotherapy-resistant (progressive disease/stable disease, PD/SD) TNBC (p = .052; Figure 1F) compared with that in chemotherapy-sensitive TNBC (complete response/partial response, CR/PR, according to the RECIST 1.1 standard).⁴⁴ We observed that in the chemotherapy-resistant group, ALYREF was mainly localized in the nucleus, while ALYREF was trapped in the cytoplasm with reduced nuclear levels in the chemotherapy-sensitive group (Figure 1F). In addition, the nuclear-to-cytoplasm ratio of ALYREF was higher in the chemotherapy-resistant group (Figure 1G). Noticeably, the correlation between the total ALYREF level and poorer relapse-free survival (RFS) and overall survival (OS) was not statistically significant (p > .05; Figure 1H), while a high nuclear ALYREF level following systemic treatment correlated significantly with poorer RFS and OS in patients with TNBC (Figure 1I). Meanwhile, the nuclear ALYREF level did not correlate significantly with tumour stage or lymph node stage in TNBC (Figure S1F).

Taken together, these results indicated that ALYREF nuclear transport correlated substantially with a poor PTX response in TNBC.

3.2 RNF31-mediated ALYREF ubiquitination facilitated ALYREF nuclear transport upon PTX treatment

Given the above findings for the subcellular distribution of ALYREF and its potential role in PTX resistance in TNBC, we next investigated the molecular mechanism through which ALYREF was transported from the cytoplasm to the nucleus under the stress conditions imposed by PTX. Currently, the nuclear transport pathway of ALYREF remains mostly unknown. To this end, we first performed immunoprecipitation followed by mass spectrometry analysis (IP/MS) to identify proteins that interact with ALYREF in PTX-resistant-TNBC cells (Figure 2A; Table S6).

Emerging evidence shows that protein post-translational modification, specifically ubiquitination, plays a vital role in protein trafficking^{34, 35}; therefore, we focused on the E3 ligases in the mass spectrometry assay, and found that several E3 ligases interacted with ALYREF, including RNF31, Cbl proto-oncogene like 1 (CBLL1), Makorin ring finger protein 1 (MKRN1), and MKRN2. Assessment of protein levels in the nuclear fractions and whole cell lysates showed that siRNA-mediated depletion of RNF31, but not CBLL1, MKRN1, or MKRN2, reduced the nuclear ALYREF level in PTX-resistant TNBC cells (Figure 2B; Figure S2A). The effect of RNF31 on ALYREF nuclear trafficking and PTX resistance was then further examined.

The expression level of RNF31 was upregulated in PTX-resistant cells compared with that in parental cells (Figure 2C). Exogenous overexpression of RNF31 in parental TNBC cells showed that in the absence of PTX treatment, RNF31 scarcely influenced the nuclear localization of ALYREF. By contrast, upon PTX treatment, overexpression of RNF31 increased the nuclear level of ALYREF in parental TNBC cells (Figure 2D; Figure S2B). Conversely, in PTX-resistant cells, silencing RNF31 dramatically reduced the nuclear ALYREF level under PTX treatment (Figure 2D, Figure S2B). An immunofluorescence assay showed that ectopic expression of RNF31 relieved the cytoplasm-trapped ALYREF and increased the nuclear ALYREF level in response to PTX treatment in parental TNBC cells, while knockdown of RNF31 trapped ALYREF in the cytoplasm in PTX-resistant TNBC cells (Figure 2E; Figure S2C,D). These data suggested that RNF31 promotes ALYREF nuclear translocation upon PTX administration.

Subsequently, we confirmed the interaction between ALYREF and RNF31 using endogenous and exogenous reciprocal immunoprecipitation. The results clearly showed that ALYREF interacted with RNF31 in PTX-resistant TNBC cells (Figure 2F; Figure S2E) and HEK293FT cells (Figure 2G). RNF31 serves as the catalytic subunit of the E3 ubiquitin ligase complex LUBAC³⁶; therefore, we detected the impact of RNF31 on ALYREF and observed that overexpressing RNF31 enhanced the linear-linked ubiquitination level of ALYREF (Figure 2H). Specifically, an active site mutant of RNF31 (C699S/C702S) failed to modulate the linear ubiquitination of ALYREF (Figure 2I). Then, we demonstrated that silencing RNF31 (Figure 2J; Figure S2F) or blocking RNF31 activity using Thiolutin (an inhibitor of RNF31)⁴⁵ (Figure 2K; Figure S2G) eliminated the linear ubiquitylation of ALYREF in PTX-resistant TNBC cells. As expected, Thiolutin treatment reduced the nuclear ALYREF level dramatically in PTX-resistant TNBC cells (Figure S2H). An in vitro assay showed that overexpressing RNF31 substantially increased the cell viability and colony formation ability of parental TNBC cells under PTX treatment, while knockdown of ALYREF blocked the effect induced by RNF31 overexpression in parental TNBC cells (Figure 2L; Figure S2I). Silencing RNF31 or Thiolutin treatment both inhibited the cell viability and colony formation ability of PTX-resistant TNBC cells upon PTX treatment (Figure 2M; Figure S2J).

The above results showed that RNF31 ubiquitinated ALYREF and facilitated ALYREF nuclear import upon PTX treatment, consequently mediating PTX resistance in TNBC cells.

3.3 RNF31 enhanced ALYREF nuclear transport via importin 13

We then investigated the pathway by which RNF31 mediates ALYREF nuclear translocation upon PTX treatment. Nuclear import of proteins is predominantly carried out by the IPO superfamily, which relies on the small GTPase Ran.^{46, 47} The collapse of the RanGTP-RanGDP concentration gradient across the nuclear membrane dramatically blocks Ran-dependent IPO shuttling across the nuclear membrane.^46-51 Emerging evidence demonstrates that various types of cellular stress, including heat stress and oxidative stress, destroy the RanGTP-RanGDP gradient, which inactivates the IPO system and traps proteins in the cytoplasm. Karyopherin subunit Beta 1 (KPNB1 or importin β1) and KPNA2 (importin α1) are well-investigated karyopherins that transport a variety of cargoes in a Ran-dependent manner.⁵² Interestingly, our anti-ALYREF IP/MS assay showed that KPNB1 and KPNA2 interacted with ALYREF in PTX-resistant cells (Table S6). Besides, IPO13, a bidirectional nuclear transporter, was found recently to function efficiently under several stress conditions in a Ran-independent manner.^{50, 51} Therefore, we investigated the function of different karyopherins in ALYREF trafficking under PTX-induced stress. siRNA knockdown of IPO13 dramatically reduced the nuclear ALYREF level in PTX-resistant TNBC cells, while silencing KPNB1 or KPNA2 only slightly altered the nuclear ALYREF level in PTX-resistant TNBC cells (Figure 3A; Figure S3A). Notably, in parental TNBC cells without PTX treatment, silencing KPNB1 or KPNA2 significantly reduced the nuclear ALYREF level, while the knockdown of IPO13 barely influenced the nuclear ALYREF level (Figure 3B; Figure S3B).

In PTX-resistant cells, the knockdown of either RNF31 or IPO13 decreased the nuclear ALYREF level (Figure 3C). In parental TNBC cells, ectopic expression of RNF31 rescued the decreased nuclear ALYREF level induced by PTX treatment, while silencing IPO13 eliminated the effect of ectopic RNF31 expression (Figure 3D). Subsequently, we confirmed the interaction between ALYREF and IPO13 exogenously: reciprocal immunoprecipitation-western blotting showed that ectopic expression of RNF31 triggered the interaction between ALYREF and IPO13, while the absence of RNF31 attenuated the interaction between ALYREF and IPO13 in HEK293FT cells (Figure S3C). RanGTP is mainly localized in the nucleus, whereas RanGDP is predominantly located in the cytoplasm.^{46, 47} The collapse of the RanGTP-RanGDP concentration gradient by stress reduced the RanGTP level and trapped Ran in the cytoplasm.^{49, 51} As expected, upon PTX treatment, the nuclear Ran level was dramatically reduced (Figure 3E), and the RanGTP level was dramatically decreased. RAN binding protein 1 (RanBP1) was found to bind to RanGTP but not RanGDP^{53, 54} (Figure S3D). Notably, both the mRNA and protein levels of IPO13 were comparable between PTX-resistant and parental TNBC cells (Figure 3F). The above results indicated that in the absence of PTX treatment, KPNB1 and KPNA2 function as the predominant transporters of ALYREF in TNBC, while upon PTX stress, IPO13, but not KPNB1 or KPNA2, serves to efficiently transport ALYREF into the nucleus in TNBC cells.

Individual inhibition of either IPO13 or RNF31 dramatically reduced the nuclear ALYREF level in PTX-resistant TNBC cells, while dual inhibition did not produce an enhanced effect on the nuclear ALYREF level, indicating that RNF31 and IPO13 regulate the nuclear ALYREF level in the same pathway (Figure 3G). We also noticed that the interaction of ALYREF with mRNA export complex components (NXF1, NXT1, THOC5, and UAP56)^{43, 55, 56} was reduced by silencing RNF31 or adding Thiolutin in PTX-resistant TNBC cells (Figure 3H,I). Besides, silencing IPO13, and IPO13 silencing in combination with Thiolutin, reduced the nuclear ALYREF level in PTX-resistant TNBC cells (Figure 3J; Figure S3E). However, combining Thiolutin with IPO13 silencing did not show an additive effect on reducing the nuclear ALYREF level, further demonstrating that RNF31 and IPO13 function in the same pathway to regulate ALYREF nuclear translocation upon PTX treatment (Figure 3J; Figure S3E). Consistently, silencing IPO13, or IPO13 silencing combined with Thiolutin, dramatically inhibited cell viability and colony formation ability upon PTX treatment in PTX-resistant TNBC cells, while there was no significant difference between the effects silencing IPO13 and the combination of silencing IPO13 and Thiolutin (Figure 3K,L; Figure S3F).

Taken together, the above data indicated that RNF31 facilitated ALYREF nuclear transport in response to PTX treatment in an IPO13-dependent manner.

3.4 RNF31 facilitated ALYREF-mediated PTX resistance in TNBC in vitro

To investigate whether RNF31 is involved in ALYREF-mediated PTX-resistance in TNBC, we first examined the mRNA abundance of PTX-resistance-related factors upon PTX treatment⁸ using qRT-PCR in PTX-resistant and parental MDA-MB-231 cells. In parental cells, PTX treatment dramatically reduced the mRNA levels of multiple factors related to PTX resistance. Overexpressing RNF31 rescued the reduced level of PTX-resistance-related factors in parental cells, specifically upregulating the expression level of microtubule protein TUBB3, microtubule-binding protein STMN1, and TAU. Blocking ALYREF expression eliminated the effect of ectopic RNF31 expression on parental cells upon PTX treatment (Figure 4A). In addition, in PTX-resistant cells, silencing of RNF31 decreased the expression of PTX-resistance-related factors (Figure 4B). Three PTX-resistance-related factors, TUBB3, STMN1, and TAU, were selected for further exploration because their expression changes were most significant in response to ectopic RNF31 expression.

Compared with those in parental TNBC cells, TUBB3, STMN1, and TAU levels were substantially increased in PTX-resistant TNBC cells (Figure S4A,B). Consistently, upregulated TUBB3, STMN1, and TAU levels were observed in patients with TNBC who were resistant to PTX-based chemotherapy (Figure S4C). The expression levels of TUBB3, STMN1, and TAU correlated significantly with nuclear ALYREF levels in patients with TNBC (Figure 4C). Furthermore, blocking RNF31 with an siRNA or Thiolutin, or silencing IPO13, decreased both the total and cytoplasmic mRNA levels of TUBB3, STMN1, and TAU, as validated by nuclear-cytoplasmic fractionation and subsequent qRT-PCR (Figure 4D). ALYREF acts as an mRNA binding and mRNA export adaptor; therefore, we next explored the effect of RNF31 and IPO13 on ALYREF binding to mRNA. Anti-ALYREF antibody-mediated RNA immunoprecipitation (RIP)-PCR showed that blocking RNF31 and silencing IPO13 significantly decreased the level of ALYREF binding to TUBB3, STMN1, and TAU mRNAs in PTX-resistant MDA-MB-231 cells (Figure 4E). By contrast, blocking RNF31 and silencing IPO13 promoted the binding of TUBB3, STMN1, and TAU mRNA by Mtr4 exosome RNA helicase (MTREX, also known as hMTR4), a key marker of the mRNA degradation complex within the nucleus⁵⁷ (Figure 4E). In addition, blocking RNF31 and silencing IPO13 dramatically shortened the half-life of TUBB3, STMN1, and TAU mRNA in PTX-resistant MDA-MB-231 cells (Figure 4F). Besides, in parental cells, PTX treatment dramatically reduced the protein levels of TUBB3, STMN1, and TAU, and overexpressing RNF31 rescued the reduced levels of TUBB3, STMN1, and TAU in parental cells (Figure 4G). Meanwhile, blocking ALYREF expression eliminated the effect of ectopic RNF31 expression on parental cells upon PTX treatment (Figure 4G). The above results showed that RNF31 maintained ALYREF's mRNA binding and exporting functions; therefore, we further examined whether ALYREF affected RNF31 expression in response to PTX treatment. As shown in Figure S4D, in PTX-resistant cells, silencing ALYREF and IPO13 decreased RNF31 mRNA and protein levels, indicating that RNF31 and ALYREF form a positive feedback loop in PTX-resistant TNBC cells.

Taken together, the above results demonstrated that RNF31 facilitated ALYREF-induced PTX resistance by upregulating the expression of PTX-resistance-related factors.

3.5 RNF31 deficiency reversed PTX resistance in TNBC in vivo

To further determine whether RNF31 is involved in PTX resistance in TNBC, we examined the effect of RNF31 on the PTX response in a TNBC xenograft model. To this end, RNF31 knockdown and IPO13 knockdown PTX-resistant MDA-MB-231 cells were injected into the breast fat pad of nude mice for a tumour growth assay (Figure 5A). Notably, deficiency of RNF31 and IPO13 greatly impaired tumour progression upon PTX administration, as indicated by significantly reduced increases in the size and weight of the xenograft tumours (Figure 5B–E). The tumours in the IPO13 knockdown plus paclitaxel group were slightly smaller than those in the RNF31 knockdown plus paclitaxel group, while the difference was not statistically significant (p = .676) (Figure 5B). These observations were further confirmed by a terminal deoxynucleotidyl transferase nick-end-labelling (TUNEL) assay and cleaved caspase-3 detection, which showed much higher levels of TUNEL staining and cleaved caspase 3 levels in tumour tissues of the RNF31 knockdown and IPO13 knockdown groups, compared with those in the shVector plus PTX group (Figure 5F,G). The remaining tumours also showed strongly reduced nuclear ALYREF levels in tumour tissues of the RNF31 knockdown and IPO13 knockdown groups, as detected by western blotting and IHC assays (Figure 5H,I). Consistently, the mRNA and protein levels of TUBB3, STMN1, and TAU were dramatically reduced in the tumour tissues of the RNF31 knockdown and IPO13 knockdown groups treated with PTX (Figure 5J,K).

This in vivo model provided evidence that RNF31 deficiency significantly overcame PTX resistance in TNBC.

3.6 Inhibition of RNF31 activity increased PTX sensitivity in TNBC

We further sought to translate these observations in vivo and in patient-derived organoids. PTX-resistant MDA-MB-231 cells were injected into the breast fat pad of nude mice to induce tumour formation. Then the mice were separated randomly into four groups: control, PTX treatment, Thiolutin treatment, and PTX plus Thiolutin treatment, respectively (Figure 6A). Notably, PTX plus Thiolutin dramatically inhibited tumour growth, as shown by the smaller tumour size and delayed tumour growth, while Thiolutin monotherapy also showed some tumour inhibition (Figure 6B–E). These observations were confirmed by a TUNEL assay for cell death, which showed a much higher TUNEL signal in tumour group treated with PTX plus Thiolutin than in the other groups (Figure 6F).

To further determine the synergistic effect of Thiolutin and PTX in TNBC, we treated TNBC patient-derived organoids with PTX and Thiolutin. Two TNBC-derived organoids expressing a high level of RNF31 were selected for drug evaluation (Figure 6G; Figure S5A). Compared with PTX single agent treatment, administration of PTX plus Thiolutin dramatically inhibited tumour growth in both organoids, while Thiolutin monotherapy slightly reduced organoid tumour growth (Figure 6H,I).

These results indicated that Thiolutin enhanced the sensitivity of TNBC to PTX therapy.

3.7 Clinical relevance of RNF31 in patients with TNBC

Finally, we assessed RNF31 expression in patients with TNBC. Immunohistochemical staining was performed on 183 patients with TNBC who received PTX-based neoadjuvant (n = 26) or adjuvant (n = 157) chemotherapy. The IHC staining and correlation analysis revealed that RNF31 expression was higher in patients with TNBC who were resistant (relapse-free survival < 5 years) to PTX-based chemotherapy than in patients with TNBC who were sensitive (relapse-free survival ≥5 years) to PTX-based chemotherapy (Figure 7A). Patients with TNBC with high RNF31 levels who received PTX-based neoadjuvant or adjuvant chemotherapy experienced poorer RFS and OS (Figure 7B). The clinical characteristics of the 183 patients included in the survival analysis are listed in Tables S1–S3. Univariate and multivariate analyses of Cox regression models showed that the RNF31 level was an independent predictive factor for OS for patients with TNBC who received PTX-based neoadjuvant or adjuvant chemotherapy (hazard ratio [HR] = 2.790; 95% confidence interval [CI] 1.467–5.305; Tables S4 and S5).