*3.2. Overall Survival*

The median OS was, respectively, 12.4 and 20.5 months in the mutated and wild-type KRAS groups (adjusted HR mut vs. wt: 1.27, 95% CI: 0.70–2.27, *p* = 0.441).

Polβ, analyzed as a continuous variable, had no impact on survival in the multivariable models including KRAS status or not (HR: 0.99, 95% CI: 0.96–1.01, *p* = 0.39; HR = 0.99, 95% CI: 0.95–1.02, *p*-value = 0.388).

Patients who were negative for DNA polymerase β staining had a median OS of 11.6 months compared to 20.6 months in the positive group. The absence of DNA polymerase β caused a worse but not statistically significant OS compared to DNA polymerase β-expressing patients (HR pos vs. neg: 1.43, 95% CI: 0.57–3.57, *p* = 0.439). With the inclusion of KRAS mutational status in the statistical model, the effect on survival with Polβ was stronger (HR pos vs. neg: 1.67, 95% CI: 0.52–5.56, *p* = 0.386). The results of the multivariate analyses for OS are reported in Table 3. Kaplan–Meier curves for OS, reported in Figure 3A,B, show the effect of KRAS status on the relationship between Polβ and OS.


**Table 1.** Baseline characteristics of patients (*n* = 109) and their relation with Polβ as a continuous or dichotomous variable. Pos, positive; neg, negative.

At a median follow-up of 18.8 months (Q1–Q3: 8.3–48.9), there were 90 progressions, 62 deaths, and 100 deaths or progressions. Q1–Q3: first–third quartile, pos: positive, neg: negative, †: Kruskal–Wallis test, \*: Spearman correlation, \*\*: Fisher test.

**Table 2.** Multivariable Cox models adjusted for ECOG-PS, age, histology, smoking, and therapy for progression-free survival, considering Polβ continuous or Polβ positive vs. negative. Pos, positive; neg, negative.


**Table 3.** Multivariable Cox models adjusted for ECOG-PS, age, histology, smoking, therapy, and immunotherapy for OS, considering Polβ continuous and Polβ positive or negative. Pos: positive, neg: negative.


**Figure 2.** (**A**) Kaplan–Meier curves for PFS according to the positive or negative DNA polymerase β staining. (**B**) Effect of KRAS status on the relationship between Polβ and PFS adjusted for ECOG-PS, age, histology, smoking, and therapy.

**Figure 3.** (**A**) Kaplan–Meier curves for OS, according to the positive or negative DNA polymerase β staining. (**B**) Effect of KRAS status on the relationship between Polβ and PFS adjusted for ECOG-PS, age, histology, smoking, therapy, and immunotherapy.

#### *3.3. Overall Response Rate*

There were no differences between the DNA polymerase β negative and positive staining groups, or among different Polβ as a continuous variable in ORR to platinum-based first-line therapy (Table 4).


**Table 4.** Objective response rates by DNA Polymerase β H-score (Polβ). CR, complete response; PR, partial response; SD, stable disease; PD, progressive disease.

#### **4. Discussion**

KRAS mutations have often been investigated as possible biomarkers for selecting chemotherapy, but results have varied, casting doubt on the true utility of this protein. In a previously published randomized prospective trial from our group, an analysis of 247 patients showed that those carrying KRAS mutations and treated with a first-line platinum-based regimen had worse PFS than patients with wild-type KRAS [5]. The present study detected a not-statistically-significant effect for OS, KRAS-mutated patients having a worse prognosis than KRAS wild-type patients. A possible explanation, although the trend is in line with previous observations, is that the statistical power of this cohort of patients was half that in our earlier study, where KRAS status was significantly associated with survival. On the other hand, the LACE-Bio pooled analysis, including data of 1543 patients participating in four clinical trials, showed that there is no difference in terms of outcomes in early-stage lung cancer patients with either wild-type or mutated KRAS [12]. Our different result may suggest that KRAS mutations could play different roles in early and advanced disease. In advanced stages, KRAS could be a condition necessary, but not sufficient, to explain a more aggressive phenotype.

There is preclinical evidence that KRAS and its mutated versions modulate DNA repair, hence the cellular response to genotoxic agents. Oncogenic RAS can inactivate BRCA-1 dependent homologous recombination (HR) by favoring the dissociation of BRCA-1 from chromatin [13]. Moreover, activated KRAS can suppress the expression of DNA repair genes (including BRCA1, BRCA2, EXO-1, and TP53) [14]. In leukemic cells, mutant KRAS promoted the upregulation of components of the alternative nonhomologous end-joining (NHEJ) pathway, such as DNA ligase IIIα, PARP1, and XRCC1, and the inhibition of the alternative NHEJ pathway selectively sensitized KRAS-mutated cells to chemotherapy [15].

Our group also suggested KRAS-dependent specific alterations in the BER system, where we found DNA polymerase β as a possible selection factor. We demonstrated at the preclinical level that DNA polymerase β could play a role in the response to cisplatin-based chemotherapy, and the data indicated a pattern of sensitivity or resistance depending on the KRAS mutational status [9]. These findings support the hypothesis that the combination of mutant-KRAS status with DNA repair could be a predictive biomarker for response to platinum-based therapy.

On the basis of these assumptions, we planned a translational study to clinically validate KRAS and DNA polymerase β as "biomarkers" for poor response and outcome to platinum-based first-line chemotherapy. We investigated DNA polymerase β as a possible selection marker, alone or in combination with KRAS status. DNA polymerase β expression, summarized in the H-score and considered as a continuous variable, was meaningless to both PFS and OS, alone or with KRAS.

When we compared negative or positive DNA polymerase β staining patients, we detected an interesting, though not statistically significant, difference: OS patients negative for DNA polymerase β staining had worse outcomes than the positive staining group. This result was confirmed even when KRAS status was considered in the analysis.

These data, although interesting and calling for further analysis, are not supported by the literature, where DNA polymerase β upregulation was described as causing resistance to cisplatin in an ovarian cancer model [16]. In a colorectal cancer model expressing high levels of DNA polymerase β, cisplatin was ineffective compared to the same model in which DNA polymerase β was downregulated. In the same paper, 5-year OS curves showed that patients with high DNA polymerase β expression had a significantly poorer prognosis than those with low expression [17]. However, DNA polymerase β has been investigated as a selection marker in very few, only retrospective studies, and our is the first attempt to investigate it, prospectively, in NSCLC.

A recent report suggests that if cells are not able to repair DNA single-strand break lesions through BER (as should be the case here for cells negative for DNA polymerase β), these lesions are channeled to the HR system [18]. We do not know whether this is also true for cisplatin-induced DNA lesions and whether these patients have HR alterations, but it does suggest an intriguing explanation for the worse outcome observed in DNA polymerase β-negative patients.

To our knowledge, this is the first investigation of the role and value of DNA polymerase β, alone or in combination with KRAS status, as a marker of response to platinum-based therapy in NSCLC. Besides the results, this paper also stimulates the idea to further investigate the combination of biomarkers that indicate how different biological pathways coexist or work together in those scenarios, where no single biomarker has been shown to have strong value.

In conclusion, KRAS may have a negative role in platinum-based therapy responses in NSCLC, but its impact is limited. The absence of DNA polymerase β might indicate a group of patients with poor outcomes compared to patients positively staining for this protein. In addition, a mutated form of KRAS in tumors not expressing DNA polymerase β further worsens survival. Therefore, these two biomarkers together might well identify patients for whom alternatives to platinum-based chemotherapy should be used.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/8/2438/s1, Figure S1: Negative and positive DNA polymerase β staining tissues magnified 400X.

**Author Contributions:** Conceptualization, H.L., M.B., M.C.G., and M.M. (Mirko Marabese); data curation, S.M.; formal analysis, M.F.A. and E.R.; funding acquisition, M.C.G. and M.M. (Mirko Marabese); investigation, M.G., H.L., E.C., G.L.R., F.L.C., A.C.B., A.P., M.M. (Michele Milella), A.F., M.D.M., P.R., and G.N.; writing—original draft, M.M. (Mirko Marabese); writing—review and editing, M.F.A., M.G., H.L., E.C., G.L.R., F.L.C., A.C.B., A.P., M.M. (Michele Milella), E.R., A.F., M.D.M., P.R., S.M., G.N., M.B., M.C.G., and M.M. (Mirko Marabese). All authors have read and agreed to the published version of the manuscript.

**Funding:** The research leading to these results received funding from Transcan Era-Net JTC-2011 project ID 40 (M.C.G.) and AIRC under MFAG 2016 project ID 18386 (M.M. (Mirko Marabese)).

**Conflicts of Interest:** M.C.G. reports grants from AstraZeneca, Bristol-Myers Squibb, MSD Oncology, Roche, Takeda, Incyte, Lilly, Merck, Pfizer, Clovis, Merck Senoro, Bayer, Spectrum Pharmaceuticals, F. Hoffmann-La Roche, Blueprint Medicine; personal fees from AstraZeneca, Bristol-Myers Squibb, MSD Oncology, Novartis, Roche, Takeda, Tiziana Life Sciences, Celgene, Incyte, Boehringer Ingelheim, Otsuka Pharma, Bayer, Inivata, Sanofi-Aventis, SeaGen International GmbH; non-financial support from AstraZeneca, Roche, Pfizer, F. Hoffmann-La Roche. The remaining authors have nothing to disclose.
