一句话亮点

这篇文章发现USP14通过和E3泛素连接酶FBXW7“抢座位”来稳定MTDH蛋白，进而激活NF-κB信号，同时推动头颈鳞癌的转移和化疗耐药——而且这个机制在临床样本里确实和淋巴结转移及不良预后挂钩。

背景/痛点

头颈鳞癌（HNSCC）的预后一直不怎么好，主要麻烦在于它特别容易转移，而且对化疗（顺铂、紫杉醇）很容易耐。MTDH这个蛋白在很多癌种里都被证实能促进EMT、维持肿瘤干细胞特性，还和耐药密切相关。Wang团队之前已经发过好几篇文章证实MTDH在头颈鳞癌里通过多种机制（比如调控VEGF、AKT、NF-κB）促癌。

但问题来了：MTDH这个蛋白本身没有典型的配体结合口袋或者酶活性中心，传统的“直接抑制蛋白功能”的策略很难走通。那不如换个思路——既然肿瘤里MTDH蛋白水平异常高，能不能从“它为什么不会被降解”这个角度切入？这就是整篇文章的逻辑起点。

推理链分步拆解

1. MTDH是靠蛋白酶体降解的，那谁在保护它？

他们首先用一个很基础但必须做的实验确定了MTDH的降解途径：MG132（蛋白酶体抑制剂）处理能让MTDH蛋白明显累积，而CQ（溶酶体抑制剂）没什么效果。这就把方向锁死在蛋白酶体-泛素系统上了。

然后他们用Flag-MTDH做亲和纯化+质谱，筛到了好几个可能和MTDH结合的去泛素化酶（DUB），其中USP14的评分最高。Co-IP和免疫荧光共定位进一步确认了它俩确实能拉到一起，在细胞质里有明显的共定位（Pearson系数>0.76，挺高的了）。

@方法论点评：先确定降解途径（蛋白酶体vs溶酶体），再用质谱无偏筛选互作蛋白，最后用Co-IP和共定位交叉验证——这是鉴定新底物-酶关系的经典“三步走”，每一步都在排除假阳性。

Fig. 1：MTDH is degraded via the proteasome and interacts with USP14. (A–D). Fadu and HN8 cells were exposed to MG132 or CQ under concentration- gradient conditions (A and C) or time-course conditions (B and D) to evaluate changes in MTDH protein abundance after different treatments. The accompanying quantification shows the densitometric ratio of each target band relative to GAPDH at the indicated concentrations or time points. (E) Schematic overview of the strategy used to screen MTDH-interacting proteins in Fadu cells is shown at the Top. Candidate deubiquitinases potentially associated with MTDH were identified by mass spectrometry analysis, as shown at the bottom. (F) HEK293T cells were transfected with Flag–MTDH or HA-USP14 for 48 h, followed by collection of cell lysates for immunoprecipitation analysis. (G and H) Coimmunoprecipitation assays were performed to examine the endogenous interaction between MTDH and USP14 in Fadu and HN8 cells. (I) Immunofluorescence staining was used to visualize the subcellular colocalization of MTDH and USP14. (J) Line- scan profiling and Pearson’s correlation coefficient analysis were conducted using ImageJ along randomly selected lines from panel I. (K) HEK293T cells were cotransfected with HA-MTDH together with vectors encoding Flag-tagged full-length USP14 or the indicated USP14 truncation mutants. Cell lysates were immunoprecipitated with anti-Flag antibody and analyzed by immunoblotting with the indicated antibodies. (L) HEK293T cells were cotransfected with HA-USP14 and vectors expressing Flag-tagged full-length MTDH or the indicated MTDH truncation mutants. Lysates were subjected to anti-Flag immunoprecipitation, followed by immunoblotting with the indicated antibodies.

Fig. 1. MTDH is degraded via the proteasome and interacts with USP14. (A–D). Fadu and HN8 cells were exposed to MG132 or CQ under concentration- gradient conditions (A and C) or time-course conditions (B and D) to evaluate changes in MTDH protein abundance after different treatments. The accompanying quantification shows the densitometric ratio of each target band relative to GAPDH at the indicated concentrations or time points. (E) Schematic overview of the strategy used to screen MTDH-interacting proteins in Fadu cells is shown at the Top. Candidate deubiquitinases potentially associated with MTDH were identified by mass spectrometry analysis, as shown at the bottom. (F) HEK293T cells were transfected with Flag–MTDH or HA-USP14 for 48 h, followed by collection of cell lysates for immunoprecipitation analysis. (G and H) Coimmunoprecipitation assays were performed to examine the endogenous interaction between MTDH and USP14 in Fadu and HN8 cells. (I) Immunofluorescence staining was used to visualize the subcellular colocalization of MTDH and USP14. (J) Line- scan profiling and Pearson’s correlation coefficient analysis were conducted using ImageJ along randomly selected lines from panel I. (K) HEK293T cells were cotransfected with HA-MTDH together with vectors encoding Flag-tagged full-length USP14 or the indicated USP14 truncation mutants. Cell lysates were immunoprecipitated with anti-Flag antibody and analyzed by immunoblotting with the indicated antibodies. (L) HEK293T cells were cotransfected with HA-USP14 and vectors expressing Flag-tagged full-length MTDH or the indicated MTDH truncation mutants. Lysates were subjected to anti-Flag immunoprecipitation, followed by immunoblotting with the indicated antibodies.（图注取自PDF文本层，来源：PNAS, 2026）

2. USP14确实去泛素化MTDH，而且只切K48链

接下来他们要证明的不是“结合”，而是“功能”。用USP14特异性抑制剂IU-1处理、CRISPR敲除USP14，都让MTDH蛋白水平往下掉；反过来过表达USP14，MTDH就稳住了。关键是mRNA水平没有变化——说明是翻译后调控。

他们做了几个关键实验来夯实这个结论：①CHX放线菌酮追踪实验，证实USP14确实延长了MTDH的半衰期；②泛素化实验，证实USP14过表达减少了MTDH的泛素化修饰；③他们进一步区分了泛素链类型——USP14只减少K48链（降解信号），不影响K63链（信号转导相关）。

@方法论点评：这里很值得学习的是他们对“去泛素化酶”这个身份的验证梯度——从抑制剂到敲除到过表达，从蛋白水平到半衰期到泛素化修饰，最后还区分了泛素链类型。每一步都在回答“是”或“不是”，避免停留在关联性结论上。

Fig. 2：USP14 deubiquitinates and upregulates MTDH. (A) HN8 and Tu686 cells were exposed to increasing concentrations of IU1, a USP14 inhibitor, for 24 h, followed by detection of USP14 and MTDH protein expression. (B and C) MTDH protein levels were examined in HN8 and Tu686 cells after USP14 knockout, and in Fadu cells after USP14 overexpression. (D and E) Total RNA was extracted from the cells shown in panels B and C. MTDH and USP14 mRNA expression was analyzed by quantitative PCR, normalized to GAPDH, and expressed relative to the control group. Data are presented as mean ± SD from three independent experiments. (F) Fadu cells with stable expression of HA- vector, HA- USP14, or HA- USP14-CA were subjected to USP14 3′- UTR siRNA transfection, followed by western blot analysis of protein expression with the indicated antibodies. (G) USP14- knockout HN8 cells were treated with MG132 at 10 μM for 12 h, after which MTDH protein levels were assessed. (H) USP14-knockout HN8 cells were treated with cycloheximide (CHX; 10 μg/mL) for the indicated durations to monitor MTDH protein stability. The Right panel presents densitometric quantification of MTDH protein levels using ImageJ. (I) Fadu cells overexpressing HA- vector or HA- USP14 were treated with CHX at 10 μg/mL for different time intervals, and MTDH protein abundance was then analyzed. (J) HEK293FT cells were transfected with HA- USP14, MYC- ubiquitin, and/or Flag–MTDH as indicated. Cells were treated with MG132 at 10 μM for 12 h before collection. MTDH polyubiquitination was evaluated by immunoprecipitation with Flag beads followed by western blot analysis. (K and L) HN8 cells were treated with IU1 at 50 μM for 24 h, or HN8 cells stably expressing control sgRNA or USP14 sgRNA were exposed to MG132 at 10 μM for 12 h before harvest. Cell lysates were immunoprecipitated with anti- MTDH antibody and subsequently analyzed. (M) HEK293FT cells were transfected with MYC- K48-Ub and Flag–MTDH, with or without HA- USP14, as indicated. After 48 h of culture, cell lysates were subjected to immunoprecipitation with anti- Flag antibody and then analyzed by western blotting.

Fig. 2. USP14 deubiquitinates and upregulates MTDH. (A) HN8 and Tu686 cells were exposed to increasing concentrations of IU1, a USP14 inhibitor, for 24 h, followed by detection of USP14 and MTDH protein expression. (B and C) MTDH protein levels were examined in HN8 and Tu686 cells after USP14 knockout, and in Fadu cells after USP14 overexpression. (D and E) Total RNA was extracted from the cells shown in panels B and C. MTDH and USP14 mRNA expression was analyzed by quantitative PCR, normalized to GAPDH, and expressed relative to the control group. Data are presented as mean ± SD from three independent experiments. (F) Fadu cells with stable expression of HA- vector, HA- USP14, or HA- USP14-CA were subjected to USP14 3′- UTR siRNA transfection, followed by western blot analysis of protein expression with the indicated antibodies. (G) USP14- knockout HN8 cells were treated with MG132 at 10 μM for 12 h, after which MTDH protein levels were assessed. (H) USP14-knockout HN8 cells were treated with cycloheximide (CHX; 10 μg/mL) for the indicated durations to monitor MTDH protein stability. The Right panel presents densitometric quantification of MTDH protein levels using ImageJ. (I) Fadu cells overexpressing HA- vector or HA- USP14 were treated with CHX at 10 μg/mL for different time intervals, and MTDH protein abundance was then analyzed. (J) HEK293FT cells were transfected with HA- USP14, MYC- ubiquitin, and/or Flag–MTDH as indicated. Cells were treated with MG132 at 10 μM for 12 h before collection. MTDH polyubiquitination was evaluated by immunoprecipitation with Flag beads followed by western blot analysis. (K and L) HN8 cells were treated with IU1 at 50 μM for 24 h, or HN8 cells stably expressing control sgRNA or USP14 sgRNA were exposed to MG132 at 10 μM for 12 h before harvest. Cell lysates were immunoprecipitated with anti- MTDH antibody and subsequently analyzed. (M) HEK293FT cells were transfected with MYC- K48-Ub and Flag–MTDH, with or without HA- USP14, as indicated. After 48 h of culture, cell lysates were subjected to immunoprecipitation with anti- Flag antibody and then analyzed by western blotting.（图注取自PDF文本层，来源：PNAS, 2026）

3. USP14和FBXW7在“抢”同一个结合位点

前面他们确认了USP14是MTDH的“保护者”，但MTDH本身已经被报道过可以被E3连接酶FBXW7降解。那么问题就变成了：USP14和FBXW7之间是什么关系？

他们设计了一系列“竞争”实验：敲低FBXW7，MTDH上调，符合预期；过表达FBXW7，MTDH下调，但这时候再同时过表达USP14，MTDH又被“捞”回来了——而且USP14的酶活突变体C114A做不到这一点，说明需要去泛素化酶活性。

更直接的是Co-IP结果：敲除USP14后，MTDH和FBXW7的结合明显增强；过表达USP14后，这个结合被抑制。这就是典型的“竞争性结合”模式——USP14和FBXW7抢MTDH上的同一个“座位”，谁坐上去了谁决定MTDH的命运。

@方法论点评：这个逻辑链很漂亮——先确定两个对手（E3和DUB），然后用“三方共表达+互作检测”来证明确实存在竞争关系，而不是间接调控。这种实验设计在鉴定“竞争性去泛素化”机制时很有参考价值。

4. USP14通过MTDH调控EMT、CSC和化疗敏感性

既然USP14能稳定MTDH，那么USP14的促癌表型是不是真的依赖于MTDH？他们做了经典的“挽救实验”：USP14敲除后细胞出现了上皮样表型（E-cadherin上升、vimentin和N-cadherin下降），迁移和增殖能力也减弱；而再表达MTDH就能把这些变化逆转回去。

他们还在多个维度验证了“USP14→MTDH→表型”这个因果链条：成球实验、CD44+/CD24-流式分析、体内极限稀释成瘤实验，全都指向USP14敲除削弱了肿瘤干性，MTDH挽救能把干性部分恢复。化疗药敏感性的结果也是——USP14敲除后细胞对顺铂和紫杉醇更敏感，MTDH再表达后耐药又回来了。

@方法论点评：这里的核心是“遗传挽救实验”——敲除一个基因，观察表型；再表达它的下游底物，看表型是否被恢复。这是证明“某个表型确实是通过特定底物介导”的金标准，比单纯做相关性要强好几个档次。

5. USP14抑制剂在体内能抑制转移

最后他们回到体内验证。尾静脉注射肺转移模型显示：USP14敲除显著减少了肺转移结节数和荧光信号，MTDH再表达就能逆转。用IU-1抑制剂处理也有类似效果，证实了靶向USP14的药物可行性。

临床部分他们用组织芯片做了多重免疫荧光，在Pan-CK阳性的肿瘤细胞里单细胞层面定量了USP14、MTDH和FBXW7的表达。结果和机制预期一致：转移组USP14和MTDH高、FBXW7低，且三者表达在患者样本里存在显著相关性。

@方法论点评：注意他们在临床分析里特意强调了“在Pan-CK阳性肿瘤细胞中定量”——这是在规避bulk组织测序里间质和免疫细胞的信号污染。这种“单细胞层面”的定量分析在组织芯片研究中越来越重要，能大幅提高结论的可靠性。

Fig. 4：USP14 regulates CSC self- renewal and cellular sensitivity to chemotherapy through MTDH. (A and B) Quantification of tumor sphere diameter and sphere number in the indicated cell groups. (C) The proportion of CD44+/CD24– cells in different groups was determined by FACS analysis. (D and E) HN8 cells stably expressing NC sgRNA or USP14- targeting sgRNA (#2) were subcutaneously injected into nude mice at different cell numbers: 4 × 105, 4 × 104, or 1 × 104 cells per mouse. Tumor-initiating capacity and stem cell frequency were subsequently evaluated. (F) Tumor tissues from panel D were analyzed by multiplex immunohistochemistry staining. (G and H) Cells from the indicated groups were treated with cisplatin (G) or paclitaxel (H), followed by assessment of cell survival. (I and J) Cells from different groups were subcutaneously implanted into nude mice, which were then treated with saline or cisplatin at 2 mg/kg. Data are presented as mean ± SD from three independent experiments; P < 0.05, P < 0.01, P < 0.001.

Fig. 4. USP14 regulates CSC self- renewal and cellular sensitivity to chemotherapy through MTDH. (A and B) Quantification of tumor sphere diameter and sphere number in the indicated cell groups. (C) The proportion of CD44+/CD24– cells in different groups was determined by FACS analysis. (D and E) HN8 cells stably expressing NC sgRNA or USP14- targeting sgRNA (#2) were subcutaneously injected into nude mice at different cell numbers: 4 × 105, 4 × 104, or 1 × 104 cells per mouse. Tumor-initiating capacity and stem cell frequency were subsequently evaluated. (F) Tumor tissues from panel D were analyzed by multiplex immunohistochemistry staining. (G and H) Cells from the indicated groups were treated with cisplatin (G) or paclitaxel (H), followed by assessment of cell survival. (I and J) Cells from different groups were subcutaneously implanted into nude mice, which were then treated with saline or cisplatin at 2 mg/kg. Data are presented as mean ± SD from three independent experiments; P < 0.05, P < 0.01, P < 0.001.（图注取自PDF文本层，来源：PNAS, 2026）

Fig. 5：USP14 promotes tumor lung metastasis through MTDH. (A and B) Cells were injected into the tail veins of nude mice. After 4 wk, metastatic tumor fluorescence intensity was assessed using a small animal imaging system. (C and D) Statistical analysis of metastatic tumor fluorescence intensity and the number of metastatic nodules in nude mice, corresponding to panel B. (E–I) Multiplex immunohistochemistry was performed to detect EMT- related markers in lung metastatic tumors, along with statistical graphs for each marker. (J) After tail vein injection of control (CTRL) and MTDH- overexpressing HN8 cell lines into nude mice, the mice were treated with the USP14- specific inhibitor IU1. After 4 wk, imaging was performed using the small animal imaging system. (K and L) Statistical analysis of fluorescence intensity and the number of metastatic nodules in the lung metastases of nude mice.

Fig. 5. USP14 promotes tumor lung metastasis through MTDH. (A and B) Cells were injected into the tail veins of nude mice. After 4 wk, metastatic tumor fluorescence intensity was assessed using a small animal imaging system. (C and D) Statistical analysis of metastatic tumor fluorescence intensity and the number of metastatic nodules in nude mice, corresponding to panel B. (E–I) Multiplex immunohistochemistry was performed to detect EMT- related markers in lung metastatic tumors, along with statistical graphs for each marker. (J) After tail vein injection of control (CTRL) and MTDH- overexpressing HN8 cell lines into nude mice, the mice were treated with the USP14- specific inhibitor IU1. After 4 wk, imaging was performed using the small animal imaging system. (K and L) Statistical analysis of fluorescence intensity and the number of metastatic nodules in the lung metastases of nude mice.（图注取自PDF文本层，来源：PNAS, 2026）

Fig. 6：MTDH expression is positively correlated with USP14 in HNSCC. (A) The expression of USP14, E-cad, MTDH, and N-cad was detected in HNSCC tissues with and without neck lymph node metastasis. (B) CT images showing the lesion relationship in HNSCC patients with and without neck lymph node metastasis. (C) Representative multiplex immunofluorescence images of metastatic and nonmetastatic HNSCC tissue microarray cores. Pan- CK was used to identify tumor epithelial cells. Representative channels for Pan-CK, MTDH, USP14, and FBXW7 are shown separately, together with merged images. (Scale bar, 250 μm.) (D) Heatmap showing the relative expression levels of MTDH, USP14, and FBXW7 across individual HNSCC patient samples based on automated single-cell image analysis. Expression values are shown as Z-scores. Samples are grouped according to metastatic status. (E) Quantification of the percentages of MTDH-, USP14-, and FBXW7-positive cells among total Pan-CK -positive tumor cells in metastatic and nonmetastatic HNSCC samples. Data are presented as mean ± SD. (F) Correlation analysis between MTDH and USP14 expression across HNSCC patient samples. (G) Correlation analysis between MTDH and FBXW7 expression across HNSCC patient samples. (H) Impact of USP14, FBXW7, and MTDH on patient prognosis in the Xiangya cohort.

Fig. 6. MTDH expression is positively correlated with USP14 in HNSCC. (A) The expression of USP14, E-cad, MTDH, and N-cad was detected in HNSCC tissues with and without neck lymph node metastasis. (B) CT images showing the lesion relationship in HNSCC patients with and without neck lymph node metastasis. (C) Representative multiplex immunofluorescence images of metastatic and nonmetastatic HNSCC tissue microarray cores. Pan- CK was used to identify tumor epithelial cells. Representative channels for Pan-CK, MTDH, USP14, and FBXW7 are shown separately, together with merged images. (Scale bar, 250 μm.) (D) Heatmap showing the relative expression levels of MTDH, USP14, and FBXW7 across individual HNSCC patient samples based on automated single-cell image analysis. Expression values are shown as Z-scores. Samples are grouped according to metastatic status. (E) Quantification of the percentages of MTDH-, USP14-, and FBXW7-positive cells among total Pan-CK -positive tumor cells in metastatic and nonmetastatic HNSCC samples. Data are presented as mean ± SD. (F) Correlation analysis between MTDH and USP14 expression across HNSCC patient samples. (G) Correlation analysis between MTDH and FBXW7 expression across HNSCC patient samples. (H) Impact of USP14, FBXW7, and MTDH on patient prognosis in the Xiangya cohort.（图注取自PDF文本层，来源：PNAS, 2026）

核心结论

这篇文章的核心发现可以概括为一个“三方博弈”模型：USP14和FBXW7竞争结合MTDH——USP14赢了，MTDH被去泛素化稳定下来，NF-κB信号激活，EMT和干性增强，最终促进转移和耐药；FBXW7赢了，MTDH被泛素化降解，恶性表型被抑制。

从治疗角度看，他们用IU-1抑制剂在体内验证了靶向USP14能抑制转移，这为头颈鳞癌提供了一个潜在的新策略——不直接靶向“无成药性”的MTDH，而是靶向调控它稳定性的上游去泛素化酶。

对耐药/DTP/PGCC 的启示

耐药可塑性中的“蛋白稳定性开关”：这篇文章提供的逻辑是——耐药细胞可能通过上调USP14来稳定MTDH，从而维持CSC特性和抗凋亡能力。在研究DTP细胞的耐药再激活时，关注去泛素化酶-底物轴的变化，可能比只看转录组更能抓到“快速适应”的机制。 联合用药策略的提示：既然USP14抑制剂能让细胞对顺铂和紫杉醇重新敏感，那在PGCC富集的耐药模型中，USP14抑制剂+化疗的组合是否也能打破PGCC介导的“多药耐受”值得一试。PGCC往往高表达干细胞相关标记，而MTDH恰好是干性维持的关键因子。 耐药标志物的选择：文中临床数据提示USP14/MTDH/FBXW7的表达模式可以作为转移和预后的分层依据。在耐药样本库中把这组标志物纳入检测，可能有助于区分“USP14-MTDH驱动型”和“其他机制驱动型”的耐药患者，指导后续联合用药策略。

局限

文中自己提到了一点值得注意：在CPTAC等公共蛋白质组数据库中，USP14、MTDH和FBXW7的全局组织水平相关性并不强。这很可能是因为bulk组织里混入了非肿瘤细胞的信号。他们的多重免疫荧光只在Pan-CK阳性的肿瘤细胞里做定量，解决了这个问题，但也说明在解读公共数据库时要谨慎——组织异质性会直接掩盖肿瘤细胞特异的调控关系。

另外，虽然他们证实了USP14-MTDH轴通过NF-κB发挥作用，但USP14作为一个已知的19S蛋白酶体相关DUB，它的底物远不止MTDH一个。文中质谱也筛到了HNRNPL、PSMD14等其他候选底物，这些底物是否也在USP14的促癌功能中“搭了把手”，目前还不完全清楚。USP14抑制剂治疗时可能同时影响多个通路，脱靶效应和毒性问题文中也有提及，高剂量IU-1在细胞层面就有USP14非依赖的毒性。

来源

期刊：PNAS，2026年。DOI: 10.1073/pnas.2504068123