We show that low expression of the 26S proteasome subunit PSMD9 was significantly associated with reduced incidence of LR in breast tumours after adjuvant RT. Our results for PSMD9 are, in essence, opposite to those of previous studies on the prognostic/predictive value of the expression of the proteasomal subunit p32 in cancers, in which low p32 expression was found to associate with increased LR rates after RT [
9,
10]. Two factors are worthy of discussion with respect to this apparent conflict. Firstly, and most obviously, the reports focus on different subunits of the proteasome and it is plausible that the two subunits have opposite correlations with RT response. It remains unclear whether expression of either is representative of proteasomal activity, as would be relevant when considering therapies with proteasome inhibitors. Secondly, we have used a cohort of cancer cases (n = 157) that is substantially larger than those in the previous studies (n = 47 [
9] and n = 28 [
10]) therefore our result may be more robust. A further key point related to this is that we are the first to attempt to separate the predictive and prognostic aspects of expression of a proteasomal subunit (Figure
3). The previous studies used cohorts where all patients received RT [
9,
10] therefore it was not possible to determine whether expression correlated with outcome irrespective of treatment (prognostic) or with RT response (predictive). In breast cancers, we found PSMD9 expression to correlate with outcome only in patients treated with RT (Figure
3B), indicating that PSMD9 represents a RT predictive marker rather than a simple prognostic marker for LR. While statistically significant, this result is potentially flawed by the low LR rate in patients who did not receive RT (Figure
3C). This flaw is unfortunately experimentally intractable for breast cancer since only patients at a low risk of LR are spared RT (for example, those with negative lymph nodes after mastectomy) therefore LRs are inevitably very uncommon in the no RT group. A further difficulty that follows from this is that the groups that receive and do not receive RT differ in terms of their lymph node status, although not in other pathological parameters (Additional file
3: Table S2).
The need for RT predictive markers has long been highlighted [
4]. It is well established that breast cancer subtypes (as defined by estrogen receptor, progesterone receptor, and her-2 expression) correlate with LR rates [
18,
19], with the highest rates in the estrogen receptor negative subtypes. There is evidence that these differences in LR relate to differential responses to RT [
20], although this is not well accepted enough to be taken into account when stratifying patients for RT. We do not find PSMD9 expression to correlate with estrogen receptor expression (Additional file
1: Table S1), therefore we do not believe it to be involved with the differential RT response between these groups; we can not comment on the relationships between her-2, PSMD9 and response, as her-2 data are lacking for our cohort, which largely pre-dates the start of routine her-2 testing in the UK. It is notable that development of specific molecular predictive markers for RT has lagged behind that of predictive markers for modern, targeted, systemic therapies, for which the therapeutic target itself, or molecules associated with it, have often proved good candidate markers. Recently the search for RT predictive markers has followed an analogous course, focussing mainly on proteins involved with DNA metabolism and repair, since DNA is the target of RT. Some potential markers have been identified [
5,
21,
22] but none have yet entered clinical practice. In addition, The mechanism by which PSMD9, or the 26S proteasome generally, is associated with RT response remains unclear although emerging evidence suggests that the proteasome may be an upstream regulator of some of key DNA repair pathways [
23], with low proteasome activity reducing DNA repair capacity and causing relative RT sensitivity. In particular, efficient proteasome function is potentially required for homologous recombination, a pathway for repairing RT-induced DNA damage, and for post-replication repair [
23], providing a mechanistic framework to support the emerging use of proteasome inhibitors as radio-sensitizers [
24,
25]. The 26S proteasome inhibitor, Bortezomib (Velcade; Millenium Pharmaceuticals), has shown a positive clinical benefit for inducing radio-sensitization in some cancers, yet the majority of success remains in haematological malignancies and its influence on solid tumours has been less encouraging [
26]. Clinical use of Bortezomib continues to be hampered by dose-limiting toxicities, drug-resistance and interference by some natural compounds. Our cell culture work supports the proposition that the PSMD9 subunit itself represents a specific and novel target for radio-sensitizing therapies, since transiently reduced PSMD9 expression at the time of exposure to RT was associated with reduced cell survival (Figure
4).
In conclusion, we show that PSMD9 expression may be a predictive marker for RT in breast cancer. This observation warrants further prospective evaluation, ideally in the context of prospective randomized studies of RT, to determine whether PSMD9 expression has clinical utility in targeting RT to those patients most likely to benefit, or in selecting patients with relatively RT resistant tumours for more intensive therapy. We also demonstrate that PSMD9 may itself present a therapeutic target for a further generation of subunit-specific proteasome inhibitors.