In this study, we identified LIG4
rs7325927 and BTBD2
rs11670188 as predictors of STS in GBM and CCDC26
rs10464870 and rs891835, HMGA2
rs1563834, and RTEL1
rs2297440 as predictors of LTS. Consistent with most other studies,23
we found age at diagnosis to be a very important predictive factor in GBM survival. In the Surveillance, Epidemiology and End Results data, 5-year survival rates are approximately 13% for 15 to 45 year olds and only 1% for those ≥ 75 years old.24
Our LTS (mean age, 48.1 years) was significantly younger than the STS (mean age, 55.9 years). Moreover, age is the initial split on the survival tree, confirming that age is the most important risk factor. Further, different genes play roles in different age group. The older patients (> 50 years) with variant type of LIG4
rs7325927 showed the worst prognosis; whereas young patients (≤ 50 years) with combined variant type of RTEL1
rs2297440 and HMGA2
rs1563834 had the best prognosis.
A major finding in this study was the consistent association of LIG4
rs7325927 with STS. LIG4
rs7325927 was the only noteworthy SNP at the FPRP low prior probability of .05. Survival tree analysis showed that older patients with LIG4
rs7325927 (V) exhibited the highest risk of death, which strongly suggests that LIG4
rs7325927 or a genetic variant in LD with this SNP is associated with STS. LIG4
(DNA ligase IV; OMIM 601837) is essential for V(D)J recombination and DNA double-strand break repair (DSBR) through nonhomologous end joining (NHEJ). As a critical protein involved in NHEJ, LIG4
forms a heterodimer with XRCC4
to execute the final rejoining step of NHEJ. Polymorphisms of this gene have been related to risk of glioma25
and multiple myeloma,26
as well as to the survival of breast cancer.27
Another promising finding is the association of BTBD2
rs11670188 with STS. Coincidentally, BTBD2
(blood-tumor barrier modification; OMIM 608531) has also been implicated in DSBR pathway because of its tight interaction with Top1. Top1
can cause double-stranded breaks28
and plays a central role in regulating cell survival.29,30
Bredel et al31
have shown that high expression of DNA Top II alpha is associated with prolonged survival in patients with GBM. The CCDC26
, and RTEL1
genes are also of interest. CCDC26
modulates retinoic acid, which in turn increases programmed cell death in GBM cells and reduces telomerase activity32–34
; whereas both HMGA2
(high-mobility group protein family, member 2; OMIM 600698)35,36
(regulator of telomere elongation helicase 1; OMIM 608833)37
were recently proved to be directly involved in DSBR pathway.
It is interesting to note that of the five important genes noted in this study, four (LIG4
, and RTEL1
) are directly or indirectly involved in the DSBR pathway, suggesting a very strong genetic interaction network. This is particularly intriguing because DSBR plays a prominent role in cell survival, maintenance of genomic integrity, and prevention of tumorigenesis. DSBs can be generated by endogenous reactive oxygen species or destabilization of stalled replication forks as well as by exposure to a variety of exogenous agents, including ionizing radiation (IR) and chemotherapeutic agents. Furthermore, IR is the only established risk factor for glioma,38–40
and IR is also used in cancer radiation therapy. Germline variation in DSBR capabilities may affect survival because of altered response to radiation or chemotherapy. Therefore, understanding how these DSBR genes and SNPs function epistatically in the same biologic pathway that influences survival of patients with GBM, whether as enhancers or inhibitors of response to radiation or chemotherapy would be very interesting in future studies.
Although our study demonstrates a strong association of certain SNPs with outcome, and there is biologic plausibility for the associations observed in our study. However, there are limitations of our study. The main one is that most of our findings did not reach statistically significant level in the replication series. However, replication failure should not be surprising or be interpreted as necessarily refuting the initial findings because of the potential problems such as population stratification. Our study and the UCSF study are clinic based while the Swedish study was mainly population based. When comparing the characteristics of the three groups, we observed large discrepancies in age at diagnosis (median 52 years in ours v 56 in Swedish and UCSF cases), and survival time (MST 22.6 months in our data set v 16.3 in the UCSF, 13.9 months in Swedish set).
Genetic heterogeneity—this heterogeneity may exist even in populations of the same ethnic group such as whites with European ancestry.41,42
We compared the distribution of four SNPs in the three groups, and found one (LIG4
rs7325927) was significant in our data set versus Swedish and UCSF, one (RTEL1
rs2297440) was marginal significant in ours and Swedish versus UCSF (Appendix Table A4, online only), indicating differences in the case populations. Different environmental exposures and different patterns of care would also lead to greater difficulties in replication. As raised by Morgan et al,43
replication studies face the risk of nonvalidation because of the lack of generalizability, especially in a heterogeneous treatment population for a pathologically and clinically heterogeneous tumor such as GBMs (GBM, the glioblastoma “multiforme,” the latter term now taken away, was a very useful reminder of this fact). Gorroochurn et al44
point out, “even if these problems were to be remedied, trying to replicate many initial findings, even if they are quite significant, maybe be predisposed to failure and should not be interpret as necessarily contradiction the initial association.” The same feeling has been echoed elsewhere,45
and Liu et al46
have provided a theoretical justification. Thus, on a second look, replication, while important and valuable, is difficult to achieve and may not be sufficient or necessary. Additional information from other lines of evidence, such as detailed molecular mechanistic studies, may be useful for validating and illuminating the functional relevance of genes identified in our study.
Another limitation is the possibility of effect change in treatment on survival because first-line GBM treatment has been changed from involved-field cranial irradiation to chemoirradiation with daily temozolomide in 2005. However, among 590 patients with GBM, 93.6% (552) were diagnosed before 2005; only 6.4%38
were diagnosed after 2005, and none of them were LTS. Therefore, we do not think that there will be any bias toward genes that are important for DNA repairs. The third limitation is the potential limited applicability of the findings to other ethnic groups, such as African Americans, Asians, and Hispanics.
In conclusion, we have provided evidence to implicate that LIG4 rs7325927 and BTBD2 rs11670188 are predictors of GBM STS and that CCDC26 rs10464870 and rs891835, RTEL1 rs2297440 and rs6010620, and HMGA2 rs1563834 are predictors of LTS. It is highly likely that the polymorphisms of these genes, which are directly or indirectly involved in the DSBR pathway, will be novel and potentially prognostic biomarkers for GBM survival, and this could be the beginning of a risk assessment model that could predict which patients would be LTS or STS.