Response to: USP26 regulates TGF‐β signalling by deubiquitinating and stabilizing SMAD7; not applicable in glioblastoma

Abstract

The response by the authors.

We were recently alerted by Dr. Thomas Ware and Dr. Hong‐Jian Zhu that the Kaplan–Meier analysis shown in Figs 5D and EV5B–E of our original study was based on a mixture of glioblastoma, oligodendroglioma and astrocytoma patients 1, 2. Since these tumours are clinically and molecularly distinct, the relevance of USP26 and the tested TGF‐β pathway components to the survival of glioblastoma patients was questioned. Moreover, we were alerted to an error in the statistical analysis in Fig 5A.

During the evaluation of the REMBRANDT (REpository for Molecular BRAin Neoplasia DaTa) database (http://www.betastasis.com/glioma/rembrandt/kaplan_meier_survival_curve/), we did not realise that the dataset used not only included glioblastoma samples, but also oligodendroglioma and astrocytoma samples. As such, our initial analysis of USP26 expression included all three tumour types and Figs 5D and EV5B–E are thus incorrectly labelled.

We have now re‐evaluated the expression of USP26 in specifically the REMBRANDT glioblastoma dataset and included further analysis of the TCGA (cBioPortal) glioblastoma datasets. In line with our previous results, we demonstrate that low USP26 expression is associated with poorer overall survival in glioblastoma, at the median threshold (high vs low), showing statistical significance with an overall P‐value of 0.0259 (REMBRANDT) (new Fig 5D in Kit Leng Lui et al 3). These findings are in agreement with those presented by Ware and colleagues, where they indicate that, “Our findings agree with the authors reported trend with low USP26 expression conferring to poorer prognosis, although the effect on survival is minor and the clinical relevance remains unclear” 2. Furthermore, we have now analysed USP26 expression in the most recent TCGA GBM cohort from cBioPortal (TCGA GBM, mRNA RNA‐seq v2 RSEM data, z score 2 threshold 1.0, n = 166). Again, we demonstrate that low USP26 expression correlates with poorer overall survival in glioblastoma patients with a significant P‐value of 0.0153 (Fig 1). We have now also included new analyses correlating the expression of USP26 in astrocytoma and oligodendroglioma. We demonstrate that USP26 does not associate with patient survival in oligodendroglioma or astrocytoma (Fig 2A and B). These data are therefore consistent with the conclusions in our original manuscript that loss of USP26 confers poor prognosis in glioblastoma.

Figure 1. USP26 expression correlates with TGF‐β activity and patient survival in glioblastoma.

Kaplan–Meier curves showing that glioblastoma patients with low levels of USP26 have a significantly lower overall survival than patients with high levels of USP26; TCGA (cBioPortal), P = 0.015, n = 158. All Kaplan–Meier curves are extrapolated from datasets based on median threshold levels (high versus low).

Figure 2. Correlation of TGF‐β pathway components with overall survival in GBM .

(A, B) Kaplan–Meier curves showing that USP26 expression does not correlate with overall survival in oligodendroglioma (A, REMBRANDT, P = 0.3984, n = 49) or astrocytoma (B, REMBRANDT, P = 0.314, n = 102). (C) Correlation of TGFB pathway components with overall survival in GBM. Kaplan–Meier curves of glioblastoma patients with TGFBRII (TCGA cBioPortal, P = 0.014, n = 158). P‐value was obtained by log‐rank test. All Kaplan–Meier curves are extrapolated from datasets based on median threshold levels (high versus low).

In a similar fashion, we attempted to correlate TGFBRI, TGFBRII, TGFBRIII, and SMAD7 expression with survival in glioblastoma. As noted, we did not realise that the REMBRANDT dataset was comprised of glioblastoma and lower‐grade gliomas. As such, results pertaining to these analyses were also mislabelled. As has been demonstrated in a number of recent publications, high TGF‐β activity is a poor prognostic factor in glioblastoma; therefore, we reasoned that deregulated TGFBRI, TGFBRII, TGFBRIII, and SMAD7 expression may also correlate with lower survival outcome. While our new analysis of the REMBRANDT database does not show a correlation between lower patient survival and high TGFBRI, TGFBRII, or TGFBRIII expression at the median threshold (new Fig EV5B‐D in Kit Lend Lui et al 3), our additional analysis of the TCGA (cBioPortal) dataset would suggest a correlation for TGFBRII, as claimed in our paper (high versus low; Fig 2C). Furthermore, analysis of the TGF‐β negative regulator, SMAD7, in the REMBRANDT GBM dataset confirms that low expression of SMAD7 correlates with decreased overall survival at the median threshold (high versus low, P = 0.039; new Fig EV5E in Kit Lend Lui et al 3). Taken together, these data are consistent with our original conclusions that deregulation of TGF‐β pathway components, resulting in the hyperactivation of the TGF‐β pathway, may act as biomarkers for poor clinical outcome in glioblastoma.

We were also alerted to an error in the statistical analysis in our original Fig 5A. We reanalysed our IHC analysis comparing USP26 expression with pSMAD2. Our original analysis indicated a negative correlation (r = −0.15, P ≤ 0.01, Dataset EV1). However, we note that analysing correlation on GraphPad Prism, we achieved a correlation coefficient of r = −0.15 with a P‐value of 0.18. To further confirm the role of USP26 in regulating TGF‐β activity, we cross‐referenced USP26 expression with the MSigDB hallmark TGF‐β signature enrichment score in glioma, a consensus signature derived from gene set enrichment analysis (GSEA, Broad Institute). Our analysis indicates that USP26 negatively correlates with TGF‐β activity (r = −0.13, P = 0.002). Furthermore, we demonstrate that USP26 expression negatively correlates with two known TGF‐β‐mediated transcriptional targets, SMAD7 (r = −0.133, P = 0.046) and LIF (ρ = −0.1859, P = 0.0165) in glioblastoma (Datasets EV2 and EV3). Collectively, these data are consistent with our claim on the potential role of USP26 and TGF‐β pathway activity in this disease.

As indicated earlier, we would like to sincerely thank the reviewers for pointing out inaccuracies in our manuscript. Importantly, the reviewers have also extended our original findings and sought to correlate TGF‐β pathway components in specific subtypes of glioblastoma (classical, mesenchymal, neural, and proneural). The reviewers correctly point out, “…that TGF‐β signalling should be examined by stratification of tumour grades and subtypes to more clearly identify its role in tumour progression”. We agree with this statement. Stratification of patient populations based on validated biomarkers and clinical histology will be critical in identifying which patients may benefit the most from TGF‐β pathway inhibitors. Although our initial analysis was performed on a relatively small patient population, our data suggest that low levels of USP26 enhance TGF‐β activity and correlate with poorer overall survival in glioblastoma. This preliminary analysis will be required to be confirmed in larger patient populations.

Supporting information

Dataset EV1

Dataset EV2

Dataset EV3

EMBO Reports (2020) 21: e47269