Dihydropyrimidine dehydrogenase (DPD, DYPD) is the enzyme responsible for the in vivo 80% detoxification of administered 5-FU, playing a major role as the rate-limiting step in FP metabolism. It has also been reported as the most important player in developing FP-related side effects in patients showing a deficient DPD activity and consequently more prone to the risk of toxicity. The prevalence of DPD partial deficiency in Caucasians is approximately 3%-5% (Etienne et al., 1994). A genetic background related to this inter-individual variability and its role in predicting toxicity occurrence have been described (Van Kuilenburg et al., 1999). Subjects who carry specific genetic variants related to a partial or complete DPD activity loss, may not be able to efficiently detoxify FP at normal rates, and are at greater risk of potentially life-threatening toxicity from standard doses.

The DPYD gene is highly polymorphic with 1,732 different genetic variants currently reported in GnomAD (https://gnomad.broadinstitute.org/), and 627 of those being missense variants impacting on the DPD protein sequence. Several studies investigating the role of specific DPYD polymorphisms on the risk of FP-related toxicity have been conducted in the past years. A number of them were conducted on large randomized clinical trials (Deenen et al., 2011; Lee et al., 2014; Boige et al., 2016) producing solid evidence on the predictive role of DPYD*2A, DPYD*13, DPYD-2846, and DPYD-HapB3 on treatment safety. The results were also confirmed in some large meta-analyses (Rosmarin et al., 2014; Meulendijks et al., 2015) and prospective pharmacogenetic guidelines were published providing guidance for dose adjustments based on DPYD genotype (Caudle et al., 2017). A first prospective study demonstrated in 2016 that a front-line FP dose adaptation based on the pre-emptive genotyping of DPYD*2A, was effective in preventing toxicity (Deenen et al., 2016).

A gene activity score was proposed for translating DPYD genotype into a protein phenotype, considering the diplotype allelic combination of the DPYD variants according to a four polymorphisms panel (DPYD rs3918290, DPYD rs55886062, DPYD rs67376798, and DPYD rs56038477) (Henricks et al., 2015). A gene activity score was developed to provide a global DPD metabolizing status for each patient and to draw personalized dosing guidelines based on the patients' genotype for the four variants. Currently, the most recent version of shared pharmacogenetic guidelines is based on the definition of the patient gene activity score (Amstutz et al., 2018; Lunenburg et al., 2020). In late 2018 Henricks et al., published the results of a large prospective safety analysis providing strong evidences that the implementation of four DPYD polymorphisms genotyping is feasible in clinical practice and that genotype-guided individualized dosing improved patient safety of FP treatment (Henricks et al., 2018). Eventually, this study gave the final input for revising published pharmacogenetic guidelines (suggesting a 50% dose reduction for patients carrying any of the four DPYD variants at the heterozygous status) and for recommending DPYD genotyping among the strategies for a safer use of FP (https://cpicpgx.org/guidelines/guideline-for-fluoropyrimidines-and-dpyd/; November 2018 Update). Recently, testing of patients for DPD deficiency before starting treatment has been recommended by the European Medicines Agency's safety committee (Pharmacovigilance Risk Assessment Committee) either by measuring the level of endogenous uracil in the blood, or by checking for specific variations in DPYD associated with an increased risk of severe side effects (https://www.ema.europa.eu/en/documents/referral/fluorouracil-fluorouracil-related-substances-article-31-referral-ema-recommendations-dpd-testing_en.pdf).

Despite the large amount of published data on the effect of DPYD genetic variants on the risk of toxicity both in patients receiving FP monotherapy or different combination chemotherapies (Toffoli et al., 2015; Ruzzo et al., 2017; Dalle Fratte et al., 2018), no specific evaluation of the guidelines application was conducted in patients receiving a concomitant treatment with radiotherapy, as in patients with LARC receiving chemoradiation treatment. FP are commonly administered as short course treatment with lower dosage in the context of chemoradiation treatments, raising perplexities on the opportunity to apply the same genotype-related dose reductions that are prescribed by current guidelines. We demonstrated that the carriers of DPYD rs3918290, DPYD rs55886062, DPYD rs67376798, or DPYD rs56038477 variants are still at higher risk of toxicity when treated with chemoradiotherapy. The significant correlation between DPYD genotype and toxicity risk was investigated in 828 patients (including 93 patients with LARC) receiving a FP-based chemotherapy in combination with radiotherapy. This study suggests that FP dose reductions should also be applied in DPYD variant allele carriers who will start nCRT to prevent severe FP-induced toxicity (Lunenburg et al., 2020).

Among all the pharmacogenomic markers currently available for treatment personalization of patients with LARC, DPYD test for FP dosage adaptation is the only validated and recommended. However, clinical implementation of the DPYD test prior to FP treatment is not of common practice in most of the health care systems so far. One of the major concerns preventing the translation of DPYD pharmacogenetic guidelines in the clinical practice is the lack of formal health technology assessment studies, including cost-effectiveness and cost-consequences evaluation. Our group previously reported that the costs required to manage FP-related toxicity are associated to the patient's DPYD genotype. We demonstrated that the mean toxicity management cost per patient is related to the patients DPYD gene activity score (based on the four DPYD variants) (Fragoulakis et al., 2019; Toffoli et al., 2019). More recently a large prospective trial, provided evidence of the cost-effectiveness of an upfront DPYD-guided dose individualization (Henricks et al., 2019).

Probably the ultimate evidence needed to support the introduction of DPYD testing and, more generally, of pharmacogenomics in the clinical practice will derive from the ongoing implementation projects (Krebs and Milani, 2019), one of which in Europe (Ubiquitous Pharmacogenomics, U-PGx). U-PGx (www.upgx.eu) is led by a European Consortium of pharmacogenomics experts formed in 2016 with the aim of assessing the clinical utility of implementing a panel of pharmacogenomic markers into routine care. A prospective, block‐randomized, controlled clinical study [PREemptive Pharmacogenomic testing for prevention of Adverse drug REactions (PREPARE)] was funded by the European Commission's Horizon‐2020 program. In such study, a panel of clinically relevant pharmacogenomic markers will be preemptively genotyped and implemented in healthcare institutions across seven European countries and patients' outcome investigated (Cecchin et al., 2017; Van Der Wouden et al., 2017).

The study of rare genetic variants in drug-related “pharmacogenes” represents a promising innovative approach capable to explain and justify the observed variation in overall drug response, not currently explained by common genetic polymorphisms. Furthermore, exome sequencing can accelerate pharmacogenetic discovery by assessing both common (i.e. minor allele frequency, MAF >5%) and rare (MAF < 1%) mutations in virtually all genes in an individual at relatively low cost (Gordon et al., 2014; Ingelman-Sundberg et al., 2018). In the future, additional novel and rare variants significantly impacting the DPD activity could be integrated into available pharmacogenomic algorithms to further improve the safety of FP administration.

RT exerts its cytotoxic action mainly by acting on DNA. If the DNA damage after irradiation is not completely repaired, cell death will occur either by apoptosis, mitotic catastrophe, or senescence (Maier et al., 2016) (Figure 2). At least five DNA repair molecular systems, each operating on a specific type of DNA damage, are involved in the repair of radiation-induced damage: direct reversal as the O6-methylguanine DNA methyltransferase–base repair; base excision repair (BER) including the repair of modifications generated by reactive oxygen as the 8-oxoguanine; nucleotide excision repair (NER); DNA mismatch repair; and double-strand breaks (DSB) repair that involved two main pathways, the non-homologous end joining (NHEJ) and homologous recombination (HR). A genetically determined defect in the capacity to repair RT induced damage has been reported (Lavin, 1998; Mckinnon and Caldecott, 2007; Mizutani and Takagi, 2013).

Three meta-analyses reported the role of DNA repair genetic variants as predictive markers of the risk to develop severe radiation-induced toxicity, although including patients withunselected solid cancers and not specifically LARC. including patients with unselected solid cancers and not specifically LARC. Particularly these meta-analyses demonstrated a significant involvement of polymorphisms in the ERCC excision repair 2 [ERCC2, also known as xeroderma pigmentosum group D (XPD), i.e., rs13181, Lys751Gln], x-ray repair cross-complementing 1 (XRCC1, i.e., rs25487, Arg399Gln), x-ray repair cross-complementing 3 (XRCC3, i.e., rs861539, Thr241Met; rs1799794, located in the 5'UTR) and ATM serine/threonine kinases (ATM, i.e., rs1801516, Asp1853Asn) genes in modulating the risk of radiotoxicity (Dong et al., 2015; Song et al., 2015a; Song et al., 2015b). However, the results of the meta-analyses are not fully transferable to the specific context of the radiation neo-adjuvant therapy of patients with LARC, where most promising data regard the XRCC1 rs25487 variant. A concordant association between the presence of a polymorphic rs25487-A (Gln) allele and an increased risk to develop grade ≥3 toxicity, and skin toxicity of any grade was reported by three different research papers (Duldulao et al., 2013a; Osti et al., 2017; Smith et al., 2017). XRCC1 is one of the most relevant members of the BER pathway and its missense variation rs25487 was associated with a decreased DNA repair capacity, supporting an impact on the risk of toxicity (Duldulao et al., 2013a).

Genetic polymorphisms in RAD51 recombinase (RAD51, i.e., rs1801320) and XRCC3 (i.e. rs1799794), two members of the DSB-HR repair pathway, were also associated to toxicity risk. The polymorphic RAD51 rs1801320-C allele resulted predictive of an increased risk of grade ≥3 abdominal/pelvis pain toxicity and acute skin toxicity of any grade, while the XRCC3 rs1799794-G allele was associated with increased risk of grade ≥3 urinary frequency/urgency, acute skin toxicity of any grade and higher rates of fatigue (Osti et al., 2017). The RAD51 rs1801320 G to C substitution in the untranslated (UTR) 5' region of the gene was shown to affect the mRNA stability leading to a lower protein expression and less effective DSBs repair (Gasinska et al., 2019). A genetically impaired RAD51 expression has been reported to impact the individual radiosensitivity, and the rs1801320 variant was correlated with the risk of radiation-induced toxicity also in other tumor settings (Osti et al., 2017).

The missense variation rs13181 (Lys751Gln) in ERCC2, a member of NER system, was also investigated. The ERCC2 rs13181-AA (Lys/Lys) genotype, resulted associated with an increased rate of grade ≥3 toxicity after a radiotherapy including treatment (Duldulao et al., 2013a). A following study aiming at validating the predictive role of ERCC2-rs13181 on radiotoxicity risk in patients with LARC (Smith et al., 2017) did not confirm those results.

In the context of chemoradiation treatment, inflammation-related genes have been considered as a target for the pharmacogenomic research, though the clinical setting of LARC has been poorly investigated up to date (Figure 2). The most promising results are related to the transforming growth factor beta 1 (TGF-β1), a cytokine that initiates and promotes acute and late radiation-induced side effects and is thus considered a major player in the inflammation process triggering. Exposure to radiation activates TGF-β1 via the generation of reactive oxygen species, and a sustained over-expression of TGF-β1 has been found in irradiated tissue. It was also investigated the role of rs1800471 in TGF-β1. The study of Schirmer et al., (2012), including two independent cohorts (discovery and replication) of Caucasian patients with LARC receiving neo-adjuvant radiation combined with 5-FU-based chemotherapy, evaluated the effect of TGFB1 rs1800471 on the acute organ toxicity (quality of life-impairing acute organ toxicity, QAOT). In both cohorts, TGFB1 rs1800471 resulted a significant predictor of QAOT, with all the patients carrying the polymorphic TGFB1 rs1800471-C (Pro25) allele experiencing QAOT. These data were confirmed by a subsequent study (Smith et al., 2017), performed in a similar cohort of Caucasian patients with LARC treated with FP-based CRT. The study reported that TGFB1 rs1800471-C (Pro25) allele was associated with an increased risk of severe (grade ≥3) toxicity. The functional effect of the TGFB1 rs1800471 polymorphism on the encoded protein has not been fully clarified yet. The polymorphism is in a genetic region encoding for the signaling portion of the peptide involved in the transport of the protein across the membrane of the endoplasmic reticulum. Preliminary in silico analysis has suggested that the presence of a proline at position 25, caused by rs1800471-C allele could affect the cleavage of the peptide, increasing TGFβ1 secretion and signaling (Schirmer et al., 2012). Another study including Chinese patients with LARC receiving nCRT demonstrated an effect on the toxicity risk of some polymorphisms in genes encoding for plasminogen activator (PA) inhibitor type 1 (PAI-1) and protease-activated receptor-1 (PAR-1). These two proteins, whose expression is activated by the TGFβ1-related pathway, contribute to acute and late radiation-induced injury (Zhang et al., 2015b). By multivariate analysis, PAI-1 rs1050955-GG and PAR-1 rs32934-CT genotypes were associated with lower risk of grade ≥2 diarrhea, fecal incontinence (i.e., anal toxicity) and other toxicities, while the PAI-1 rs2227631-GG was associated with increased risk of grade ≥2 incontinence (Zhang et al., 2015b). Those polymorphisms are placed in the genes regulatory regions and may affect gene expression in different ways. The 3'UTR PAI-1 rs1050955 variant is located at a microRNA binding site, whereas PAI-1 rs2227631 and PAR-1 rs32934 are located in the 5′-flanking region of the gene and may change the binding site of specific transcriptional factor (Zhang et al., 2015b). These data support a major causative role of TGFβ1 and its downstream mediators (i.e., PAI-1, PAR-1) on the risk of radiation–related toxicity, and a better elucidation of the predictive role of the related polymorphisms is warranted. However, the specific role of the genetic polymorphisms in the pathway mediators PAI-1, PAR-1 is proved by a single study and the lack of independent validation results prevents to draw final conclusion.

A second cytokine of interest is the nuclear factor κB1 (NFKB1), a transcriptional factor that plays a key role in the activation of the inflammatory response and regulates various biological defense processes, including innate and adaptive immune responses, acute phase reaction and apoptosis (Karban et al., 2004). The NFKB1 rs28362491 variant, consisting in an insertion/deletion of four bases in the promoter region of the gene, seems to decrease NFKB1 transcription level by reducing the binding capacity of some transcription factors to the gene promoter (Karban et al., 2004; Riemann et al., 2007). Regarding the predictive role of NFKB1 rs28362491 marker, literature provides only one study. The polymorphism was associated to the risk of radiation-induced toxicity. Particularly, the NFKB1 rs28362491-DEL/DEL genotype was associated with an increased risk of grade ≥2 dermatitis and proctitis. The rs28362491-DEL containing haplotype also showed a correlation with an increased risk to develop clinically significant grade≥2 acute organ toxicity (Dzhugashvili et al., 2014).

The pro-inflammatory cytokine tumor necrosis factor-alpha (TNF-α, encoded by TNFA) is another key molecule involved in the mediation of the inflammatory signal activated by ionizing radiation. As far as the predictive role of TNFA polymorphism in LARC is concerned, only one hypothesis generating study is currently available. The T allele of the TNFA rs1799964, located in the 5′-flanking region of the gene, was associated with increased TNF-α mRNA expression in vitro. It was reported to be associated with an increased risk of total grade ≥2 acute toxicity after chemoradiotherapy in Chinese patients with LARC (Zhang et al., 2015a).

Despite the promising aforementioned data about the potential involvement of some polymorphisms in genes implicated in the inflammation process in mediating the risk of toxicity in patients with LARC treated with nCRT, it must be underlined that most of the results are only reported in single exploratory studies, as for the case of polymorphisms in PAI-1, PAR-1, NFKB1, TNFA. Therefore, they should be further tested in independent validation studies in order to be then considered as effective biomarkers for a clinical translation.

Thymidylate synthase (TS, TYMS), an enzyme with a critical role in the DNA synthesis process, is the primary target of FP and its expression level was suggested to inversely correlate with the tumor sensitivity to FP (Johnston et al., 1995; Libra et al., 2004). Some common polymorphisms with an impact on the protein expression level and, in turn, with a hypothesized effect on an FP-based treatment outcome, were described in TYMS.

The most studied TYMS polymorphism consists of regulatory tandemly repeated sequences (VNTR) of 28–base pairs (rs34743033) in the 5' untranslated region of the gene and acts as an enhancer to the TYMS promoter (TYMS enhancer region[TSER]). TYMS expression is affected by the number of tandem repeats that are directly correlated with the enzyme expression level. Alleles with two or three repeats are the most common (TSER*2 or 2R and TSER*3 or 3R). The second repeat of the 3R allele hosts a functional G > C single-nucleotide polymorphism (rs2853542). The 3R(G) allele has been correlated with a more efficient transcription activity while the transcription efficiency of the 3R(C) allele is the same as that of the 2R allele. TSER and rs2853542 are often combined to identify two groups of alleles: TYMS high-expression allele (2R/3RG, 3R/3RG and 3RG/3RG) and low-expression allele (2R/2R, 2R/3RC and 3RC/3RC) (Thomas et al., 2011). A third polymorphism that has been studied for its effect on the protein expression is a 6 bp insertion at the nucleotide 1494 in the 3′UTR of the gene. It was reported to be associated with an increased TYMS tumor expression that may decrease the chemosensitivity to FP (1494del6, rs16430, or rs34489327).

In the last 10 years, the association of TYMS polymorphisms with the tumor response to nCRT in LARC has been under investigation in literature. Although several studies showed an inconsistency of statistically significant results, in general TYMS high-expression alleles were associated with a worse response to nCRT and unfavorable prognosis. On the other side, TYMS low-expression alleles have been overall associated with better response and prognosis.

According to an abstract presented in 2001 from Villafranca et al., (2001) TSER*3 homozygotes had a lower probability of downstaging compared with TSER*2 homozygotes or *2/*3 heterozygotes (22% vs 60%) and a trend toward improved 3-year DFS was also detected in the *2/*2 and *2/*3 groups, compared with that in the *3/*3 group (81% vs 41%). The group of Spindler in 2007 demonstrated that 53% of *2/*2 patients experienced a 53% p-CR compared with 26% of *2/*3 and only 17% of patients with the TSER*3/*3 variants (Spindler et al., 2007). The contribute of the 28-bp VNTR in the TYMS 5'UTR was considered along with the G > C SNP by Hur et al., reporting a significant difference in tumor response between low and high-expressing groups in 44 South Korean patients with LARC. Thirteen out of 14 patients in the low-expression genotype group exhibited a significantly greater tumor downstaging rate, as compared with 12 of 30 patients in the high-expression group (Hur et al., 2011). Based on such consistent results a prospective tandem phase II study was designed in 2011 by Tan and colleagues. TYMS genotyping was used to guide nCRT where carriers of three or more repeats in the TYMS 5'UTR VNTR were treated with a chemotherapy intensification program (Tan et al., 2011). TSER*2/*2 or TSER*2/*3, were treated with the standard of care hypothesizing a favorable response to 5-FU, while TSER*3/*3 or TSER*3/*4 genotypes who were unlikely to derive significant benefit from a FP-based CT were treated with irinotecan in addition to standard 5-FU/CRT. This genotype-based strategy appeared to be successful, increasing treatment efficacy in poor-responder patients. Downstaging and complete tumor response rates reached 64.4% and 20% for standard of care and 64.5% and 42% for irinotecan-treated patients, respectively. However, this increase in efficacy came at the cost of higher rates of grade 3 to 4 toxicities, delay rates and hospitalization (Tan et al., 2011).

Only a few studies deriving from one research group reported contrasting results, highlighting an association between *3/*3 genotype and a higher response rate. The first study identified the *3/*3 genotype as an independent prognostic factor for improved survival in 2010 (Paez et al., 2010). A successive study conducted in 2011 on a larger group of patients reported as well a higher response rate (pathologic complete remission and microfoci residual tumor) for TYMS *3/*3 genotype compared to *2/*2 or *2/*3 in the amount of 59% vs 35%, and longer median progression-free survival (PFS) and OS (Paez et al., 2011). A significantly improved pathologic response among carriers of TYMS high-expressing alleles as compared with low-expressing alleles was also found by Lamas et al., (2012).

Regardless, two systematic reviews and metanalyses were published in 2016 (Salnikova and Kolobkov, 2016) and 2017 (Yang et al., 2017) assessing that TYMS *2/*3 genotype is associated with the response to nCRT in rectal cancer and that patients with a *2/*2 or *2/*3 genotype might benefit more from nCRT than others. The initial evidence of results heterogeneity for TYMS rs34743033 resolved after the exclusion of the data of Paez et al., (Paez et al., 2011; Salnikova and Kolobkov, 2016). However, the authors suggest that such results are not conclusive due to discrepancies among the studies analyzed in the nCRT administration and surgery strategies, sample size and ethnicities. A significant bias could also have derived from the source tissue (either tumor or normal) used for germinal DNA extraction.

Eventually, 3'-UTR 6-bp rs16430 was significantly associated with DFS by Arrazubi et al., (2013), while only a trend towards tumor response in patients receiving 5-FU-based nCRT, was described by Stoehlmacher et al., (2008). The metanalysis from Yang et al., (2017) did not found 1494del6 to be associated with the response to nCRT.

Methylenetetrahydrofolate reductase (encoded by MTHFR gene) is a fundamental enzyme in the folate cycle, important for DNA synthesis and repair (Toffoli et al., 2003) and hypothesized to be another important target of the FP pharmacological activity. MTHFR catalyzes the reduction of 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate, the methyl donor for methionine synthesis from homocysteine. Two common non-synonymous variants in MTHFR, rs1801133 (677C > T, Ala222Val) and rs1801131 (1298A > C, Glu429Ala), alone or in combination, were reported to affect the activity of the enzyme and hence the folate distribution (De Mattia and Toffoli, 2009; De Re et al., 2010). MTHFR rs1801133-CC wild genotype was associated with a major response to nCRT compared with MTHFR rs1801133-TT genotype (Salnikova and Kolobkov, 2016). The radiotherapy effect was also hypothesized to be affected by the action of MTHFR. Specifically, the reduced MTHFR activity, deriving from the aforementioned polymorphisms, could lead to an enhanced availability of non-methylated folate substrates for de novo synthesis of nucleotides, which could preserve DNA integrity from ionizing radiation, reducing genetic instability and thus preserving its efficacy. A number of studies were published with pretty consistent results about the detrimental role of MTHFR rs1801133 on the outcome of a CRT treatment. In particular, the MTHFR rs1801133-CC genotype was demonstrated to be more sensitive to the therapy, whereas the effect of MTHFR rs1801131 is still controversial.

In two studies on patients with LARC, MTHFR rs1801133-T allele was significantly associated with a lower chance to get a TRG ≤ 2 (Cecchin et al., 2011) and with non-pCR when in homozygosity (Garcia-Aguilar et al., 2011). In 2014 Urlich et al., documented a non-significant trend toward worse DFS and OS for MTHFR rs1801133-TT compared to patients with CC, although not supported by a significant association with OS or DFS. Such trend became significant for both endpoints in patients receiving 5-FU via protracted venous infusion 42 days before and 56 days after radiotherapy (Ulrich et al., 2014). Nikas et al., in 2015 reported that patients with MTHFR rs1801133-CC were 2.91 times more likely to benefit from nCRT, and 3.25 times more likely not to experience recurrence of the disease compared to heterozygous (CT) or homozygous (TT) genotypes (Nikas et al., 2015). The study from Terrazzino et al., identified the haplotype MTHFR rs1801133-T/rs1801131-A as an independent predictor of tumor regression at univariate and later confirmed at multivariate analysis. Patients not carrying the MTHFR rs1801133-T/rs1801131-A haplotype displayed a higher response rate (Terrazzino et al., 2006). Eventually the systematic review and metanalysis from Zhao et al., draw the line suggesting that MTHFR rs1801133 might be correlated with the tumor response under the recessive model (CC vs. CT/TT) in overall analysis, rectal cancer, and TRG1–2 vs. 3–5 group. MTHFR rs1801131 showed no significant association with the tumor response to nCRT (Zhao et al., 2015).

It has also been hypothesized that the effect of MTHFR SNPs could change according to specific combination drug therapies given with radiotherapy. A study by Dreussi et al., (2016a) compared the effect of a set of SNPs, including MTHFR rs1801133, in patients with LARC receiving an FP-based nCRT with or without oxaliplatin. They demonstrated a differential effect of the polymorphisms on the tumor response in the two subgroups, thus suggesting an effect of drug-drug interactions on the pharmacogenomic of LARC treatment.

MTHFR polymorphisms were also anecdotally associated with FP-related toxicity. A study published in 2011 by Thomas et al., investigated the effect of MTHFR rs1801133 and rs1801131 genetic markers with the outcome of a chemoradiation treatment. MTHFR haplotypes (rs1801133-C/rs1801131-C) and diplotypes (CA–TA and TA–TA) showed a protective or a detrimental effect on the incidence of severe diarrhea or mucositis, respectively (Thomas et al., 2011).

The major role of germline variation in the DNA repair pathway on the response to radiotherapy has been already mentioned above for its association with a differential risk to develop adverse side effects after a chemoradiation treatment. The same genetic polymorphisms were investigated for their potential role in identifying good or poor responders to radiotherapy-based treatments. However, the results obtained up to date are far from their final conclusions. The missense XRCC1-rs25487 variant was investigated in a number of studies that generated heterogeneous results (Balboa et al., 2010; Grimminger et al., 2010; Cecchin et al., 2011; Paez et al., 2011; Lamas et al., 2012). The heterozygous rs25487-AG genotype was reported to be associated with both a higher (Grimminger et al., 2010) and a lower (Lamas et al., 2012) response rate. The XRCC1 rs25487-AA genotype was associated either with an improved tumor response (Balboa et al., 2010) or with a decreased PFS by other authors (Paez et al., 2011). Interestingly, the work of Balboa et al., (2010) raised the problem of the source of genomic DNA used in the different studies for XRCC1 genotyping (i.e. blood or tumor tissue). In their investigation, the authors demonstrated a discrepancy between the XRCC1 rs25487-AA genotype as determined in healthy or tumor cells. The authors found an elevated rate of tumor specific genomic alterations (i.e., a loss of heterozygosity, gain of allele), in the chromosomal region containing DNA repair encoding genes (i.e., XRCC1, ERCC1, ERCC2), with the XRCC1 locus presenting the highest percentage of aberrations. These findings highlighted a potential limitation in interpreting the results of the pharmacogenomic studies currently available.

The gene encoding polymorphisms for ERCC1 (a NER member) were also analyzed by a number of studies, generating conflicting results. The TT genotype of the synonymous rs11615 (Asn118Asn) polymorphism, impacting the mRNA expression level of ERCC1, (Yu et al., 2000), resulted associated with a higher rate of R1-R2 circumferential rectal margin resection, and higher recurrence rate (Paez et al., 2011). Opposite results were produced by another study, where the same ERCC1 rs11615-TT genotype resulted associated with higher rate of pathological complete response by multivariate analysis (Sebio et al., 2015). ERCC1 polymorphism (i.e. rs3212986, located in the 3'UTR) was also highlighted for its association with complete tumor response to treatment (Cecchin et al., 2011; Dreussi et al., 2016a) but the results were not concordant with the findings of another published study (Sebio et al., 2015).

The impact of a set of DNA repair genetic markers (i.e., XRCC1-rs25487, XRCC1-rs179978, XRCC3-rs861539, ERCC1-rs11615, ERCC2-rs13181) on the tumor pathological response was analyzed by a meta-analysis (Guo et al., 2015), including five studies for a total of 265 Caucasian patients with LARC. The meta-analysis found no association between the polymorphisms and the response to radiotherapy in the context of LARC multimodality treatment. However, it should be considered that the small sample size of each single study, together with their significant methodological heterogeneity (genotyping strategy, study design, treatment schedule and dosage, combination with different chemotherapeutics drugs, clinical monitoring), could have affected the possibility to draw a common conclusion.

Additional studies were published exploring the predictive/prognostic effect of polymorphisms in other genes involved in the DNA repair pathways (i.e., hOGG1, APE1, RAD51, and XPA). Despite the encouraging results of the investigations on their potential impact on the tumor response phenotype, a final consensus was not achieved, and the clinical value of the markers should be independently validated in further studies.

The variant G allele of the missense rs1052133 (Ser326Cys) polymorphism, impairing the functionality of hOGG1 (Vodicka et al., 2007), was indicated to be a detrimental factor for the pathological tumor response to radiation therapy (Cecchin et al., 2011). The minor G allele of the missense rs1130409 (Asp148Glu) variation in APE-1, was instead associated with an increased response rate to radiotherapy at high dose (Dreussi et al., 2016a).

The functionally defective RAD51 rs1801320-CC genotype was recently reported to be associated with a better outcome after short-course radiotherapy in term of both a longer OS and a lower risk to develop local recurrence and distant metastasis (Gasinska et al., 2019). This evidence supports the hypothesis that a defective repair of the radiation induced damage on DNA, determined by a genetic polymorphism, could increase the treatment efficacy with an improved patients' prognosis. Interestingly, patients with a RAD51 rs1801320-CC genotype presented also a different tumor tissue phenotype with lower glucose transporter 1 expression, moderate differentiation, lower Ku70 expression, lower aneuploidy, and higher P53 protein expression. Moreover, patients' gender, in association with RAD51 genotype and Ku70 expression, was shown to have an impact on OS. The findings of this study represent one of the first evidence of an association between a specific genetic background related to DNA repair (i.e., RAD51 genotype) and a peculiar rectal cancer molecular phenotype and suggest a gender-related difference in the therapy outcome.

Another recent study by Boige et al., (2019) suggested that the xeroderma pigmentosum group A (XPA) rs3176683 polymorphism may help identifying patients who could benefit from adding oxaliplatin to a capecitabine-based nCRT due to its predictive effect for tumor response to oxaliplatin-based nCRT. The probability of tumor response was seven times higher in patient with XPA rs3176683-TT genotype treated with CAPOX (i.e., capecitabine and oxaliplatin). However, the polymorphism failed to be associated with the patients' prognosis.

Since the radiotherapy exerts part of its cytotoxic effect by generating reactive oxygen species (ROS), genetic variations in oxidative stress-related enzymes have been also investigated as potential predictors of tumor response to treatment. However, the studies published up to date in LARC are sporadic and do not allow to draw final conclusion on the actual predictive or prognostic role of the markers investigated in NOS3, MPO, SOD2, and GST.

The missense rs1799983 (Glu298Asp) polymorphism in the gene encoding endothelial nitric oxide synthase (eNOS, NOS3), was related to ROS generation and patients OS. The same study also highlighted an association between the low ROS producing A-allele of the rs2333227 variant, located in the 5'flanking of the myeloperoxidase (MPO) gene, and a longer OS (Funke et al., 2009). Another potential predictive marker pointed out by literature is the missense rs4880 (Ala16Val) variant in SOD2 encoding gene. SOD2 plays a central role in the detoxification of ROS and its higher expression was reported to inhibit ROS-induced activation of NF-κB and to potentially increase the sensitivity to ionizing radiations (Ho-Pun-Cheung et al., 2011). In agreement with this hypothesis, the rs4880-T (Val), associated with lower SOD2 activity, was linked to a worse response to radiation therapy (Ho-Pun-Cheung et al., 2011).

Glutathione‐S‐transferases (GSTs) is a multigene family of phase II metabolic enzymes employed in the detoxification of reactive oxygen intermediates produced by a wide variety of potentially toxic and carcinogenic compounds, by conjugation with glutathione (Nicosia et al., 2018). The missense polymorphism 313A > G (Ile105Val, rs1695) in GSTP1 was proposed, in a study from Nicosia et al., (2018) to be a predictive and prognostic marker. Among a population of 80 patients with LARC, those with a 313AA genotype for GSTP1 rs1695presented a rate of 26.6% of pCR compared to 8.5% of the 313AG/GG population. Patients harboring at least a 313A allele presented a 5- and 8-year cancer-specific survival longer than those with 313GG genotype (87.7% and 83.3% vs. 44.4% and 44.4%, respectively). Only a trend for an improved OS in the rs1695-A allele carriers was reported.

EGFR was demonstrated to have a pivotal role in LARC as well as in the colorectal cancer carcinogenesis (Toffoli et al., 2007; Kim and Eng, 2012). Moreover, EGFR and related pathways, are activated by ionizing radiation, and have been suggested to be a druggable target for the neo-adjuvant treatment of LARC. However, the association of FP with cetuximab in the chemoradiation treatment was not successful. Nonetheless, germline polymorphisms in EGFR were investigated for their potential predictive or prognostic effect. The most studied functionally relevant EGFR polymorphisms in LARC are the following: 1. rs45608036, a (CA)n repeat in the intron 1 of the gene that alters EGFR expression in vitro and in vivo; 2. rs2227983, a missense substitution at codon 497 (Arg497Lys) that leads to attenuation in ligand binding, growth stimulation, tyrosine kinase activation, and induction of proto-oncogenes myc, fos, and jun; and 3. rs712830, a single nucleotide change located in the Sp1 binding site of the regulatory EGFR promoter region that impact EGFR transcription (Zhang et al., 2005; Spindler et al., 2006; De Mattia et al., 2015). The rs712830-GG genotype resulted associated with a poorer response to chemoradiotherapy (Spindler et al., 2006). In the same study, the rs712830-GG genotype was also associated with higher EGFR expression by immunohistochemistry.

In another exploratory study, both the EGFR rs2227983-GG (Arg/Arg) genotype and the rs45608036-CA repeats (genotype with both alleles <20 CA repeats) showed a tendency towards an increased risk of local recurrence after radiotherapy. The combination in the same patient of an rs2227983-G (Arg) and an rs45608036 < 20 CA repeats alleles was associated with the highest risk of local recurrence (Zhang et al., 2005). A subsequent meta-analysis, including about 350 Caucasian patients with LARC and evaluating the role of EGFR rs2227983 and rs45608036 on the response to chemoradiation therapy, showed a trend for a detrimental effect on the response to treatment of the EGFR shorter (S) alleles (Zhao et al., 2015). In another meta-analysis, including more than a thousand Caucasian patients with LARC, a different frequency of distribution of the EGFR rs2227983-A allele among responder and non-responder has been detected (Salnikova and Kolobkov, 2016). Further analyses are required to clarify whether specific EGFR polymorphisms can have a clinical utility in predicting the outcome of a chemoradiation therapy in LARC.

Within the EGFR-pathway, preliminary results have also been generated for some germline genetic markers in the downstream effector KRAS proto-oncogene (KRAS) and the EGFR-ligand amphiregulin (AREG). A 3'UTR KRAS variant, named LCS6 (rs61764370) was reported to alter the epigenetic transcriptional control of KRAS. The polymorphism was associated with an impaired capacity of the mature let-7 miRNA to bind the KRAS-encoding mRNA, potentially leading to KRAS up-regulation and downstream cellular pathway signaling interference (De Mattia et al., 2015). In patients with LARC, the results of the retrospective EXPERT-C trial evidenced that the minor KRAS rs61764370-G allele was associated with a higher rate of complete response after nCRT independently from cetuximab administration (Sclafani et al., 2015). The same polymorphic rs61764370-G-allele showed also an association trend for an improved 5-years PFS and OS. By a subgroup analysis, the favorable prognostic effect of the rs61764370 variant seemed to be limited only to patients with a KRAS mutated tumor, opening the possibility of identifying subgroups of good responder patients within that category of patients (KRAS mutated) presenting an overall unfavorable prognosis.

Some hypothesis-generating data were also published about the role of genetic polymorphisms in some downstream effectors and ligands cooperating in the EGFR pathway. These data are reported in single studies lacking independent validation and therefore requiring further investigation to assess their real clinical value.

Specifically, the intergenic rs11942466 polymorphisms located in the AREG gene region was associated with a higher rate of pCR. Through a classification and regression tree analysis, including some polymorphisms in the EGFR pathway and DNA repair genes, the AREG rs11942466 variant represented the most important marker for identifying the complete responder patients (Sebio et al., 2015). The phosphoinositide-3-kinase (PI3K)/phosphatase and tensin homolog (PTEN)/v-akt murine thymoma viral oncogene homolog (AKT)/mammalian target of rapamycin (mTOR) cascade is another downstream signaling pathway potentially activated by EGFR. The role of genetic polymorphisms in five genes belonging to the PI3K/PTEN/AKT/mTOR pathway [i.e, PI3K catalytic subunit alpha (PIK3CA), PTEN, AKT1, AKT2, FRAP1 encoding for mTOR] on the response to nCRT in LARC was recently investigated for the first time. The PTEN rs12569998-G allele was associated with an increased tumor response rate, while the AKT2 rs8100018-C allele with a decreased recurrence risk and a higher 5-years DFS rate (Peng et al., 2018). PTEN plays a crucial role in the DNA damage repair and in the cellular response to DNA damage, whereas AKT2 modulates the cell survival signals, important mechanisms for the response to nCRT in LARC.

Additional genetic markers affecting the cell cycle control, apoptosis and proliferation rate were evaluated in single exploratory studies. The rs9344 polymorphism in cyclin D1 (CCND1), controlling the G₁/S checkpoint of the cell cycle, is located at codon 242 in the exon-4/intron boundary of CCND1 gene and is responsible for alternate splicing of transcripts with different half-lives (Bitterman et al., 2015). The CCDN1 rs9344 marker was found to be an independent predictor of response to neo-adjuvant radiotherapy, with the A-allele being associated with an increased response rate and a lower risk of local failure (Ho-Pun-Cheung et al., 2007). Moreover, the combination of CCDN1 rs9344 with specific clinical-pathological parameters (i.e., post-therapeutic lymph node status) generated a prognostic index that accurately distinguished subgroups of patients with different recurrence-free survival and OS. Other data suggested that the p73 G4C14 → A4T14 marker could be an additional factor influencing the outcome of nCRT in LARC. G4C14 → A4T14 is a dinucleotide polymorphism consisting of two linked variants (rs2273953 and rs1801173, respectively) located at position 4 (G to A) and 14 (C to T) in the 5′-UTR of exon 2, just upstream the initial start codon of the tumor protein p73 (TP73, encoding P73) gene. The G4C14 → A4T14 polymorphism was supposed to impact the gene expression of p73, implicated in the regulation of the balance between pro- and anti-apoptotic signals (Loof et al., 2009). The GC/GC genotype was showed to be associated with a lower expression of p53 and Survivin, another regulator of the apoptosis and cell-cycle, and to be tendentially related to a longer DFS. The combination of a GC/GC genotype with a negative p53 expression and a weak expression of Survivin, predicted a longer DFS as compared to other genotype/phenotype combinations (Loof et al., 2009). All of these markers could be considered of interest, but only future validation work will eventually assess their clinical value.

The role of the tumor microenvironment in the response to nCRT in LARC was largely explored (Perez-Ruiz and Berraondo, 2016; Zhang et al., 2019). In this context, germline polymorphisms of proteins mediating inflammatory, angiogenic, hypoxia and cell adhesion phenomena were considered (Garziera et al., 2015; Garziera et al., 2017; Labriet et al., 2017; De Mattia et al., 2018a; De Mattia et al., 2018b; De Mattia et al., 2019; Labriet et al., 2019) (Figure 3). This is certainly a promising field of investigation in pharmacogenomics and this is proved by the number of studies published up to date analyzing different players and mediators of the complex inter-play between the microenvironment and the tumor tissue in determining the tumor response phenotype. However, despite many interesting results, no final consensus was reached up to date on the predictive role of each polymorphism. A lot of markers discussed below were investigated by single studies and lack of formal independent replication of the results prevents the possibility to draw final conclusion.

The promoter polymorphism rs4073 of interleukin 8 (IL-8), an interleukin with an important role in the angiogenic process, was linked to an increased IL-8 expression in vitro and was associated with a higher risk of LARC recurrence (Gordon et al., 2006). Another angiogenesis-related marker, the rs2010963 variant, located in the 5'UTR of vascular endothelial growth factor (VEGFA) gene and predictive of the protein serum level, was suggested to influence the probability to get a complete pathological response (Dreussi et al., 2016a). The missense variant rs351855 (Gly388Arg) in the gene encoding for fibroblast growth factor receptor 4 (FGFR4), a receptor tyrosine kinases involved in several cellular activities as angiogenesis, cell motility and inflammation, was also reported to contribute in determining the recurrence risk after radiotherapy by interacting with other genetic variants as IL-8-rs4073, intercellular adhesion molecule 1 (ICAM-1)-rs5498, and TGFB1-rs1800470 (Gordon et al., 2006). Among these markers, rs5498 is a missense variation (Glu469Lys) of ICAM-1, a cell adhesion molecule involved in the cell-to-cell interactions, another biological action under the control of the tumor microenvironment that deserves to be further investigated in the context of nCRT response in LARC.

Genetic polymorphisms in NFKB1, a key transcriptional factor for the activation of the inflammatory signaling cascades, has been also investigated. The NFKB1 rs28362491-DEL allele, alone or in haplotype combination, was reported to be predictive of an increased rate of pathological complete response. An association trend was also observed between the rs28362491-DEL allele and a longer DFS and OS (Dzhugashvili et al., 2014). This preliminary finding suggests how an altered triggering of the inflammatory response could be related to the resistance to treatment. TGFB1-rs1800470 and prostaglandin-endoperoxide synthase 2 (PTGS2)-rs20417 are other inflammatory-related genetic markers that have been suggested to contribute in determining the response to chemoradiation treatment in term of recurrence rate (Dreussi et al., 2016a). IL-13 is involved in the modulation of the immune system and the tumor immunosurveillance, which polymorphisms were related to the outcome of LARC patients treated with nCRT. The T-allele of the rs1800925 variant, located in the promoter region of IL13, was shown to be associated with a poorer response to nCRT in a cohort of Caucasian LARC patients (Ho-Pun-Cheung et al., 2011). The polymorphic T-allele was described to increase the transcription of IL-13 that is involved in a down-regulation of tumor immunosurveillance. This is in line with the hypothesis that an impaired chemoradiation-induced tumor immunosurveillance could decrease the efficacy of the chemoradiotherapy. However, the impact of rs1800925 variant on response to radiation therapy was not confirmed in another investigation performed in a cohort of patients with LARC of Chinese ethnicity (Xiao et al., 2016). Since the genotype frequency of rs1800925 varies significantly by ethnicity, further studies are required to better clarify the predictive value of this marker in the different ethnic groups.

Our group recently produced promising data for some polymorphisms (i.e., 3'UTR rs641701, 5' UTR rs9463772) in the IL-17F encoding gene. IL-17 is an effector of the immune system that displays anti-tumor proprieties by acting on tumor angiogenesis and by improving the host inflammatory response against neoplastic cells. IL-17F rs641701-C and rs9463772-A alleles were associated with a poor prognosis in term of higher risk of disease recurrence after surgery, of distant failure of the treatment, and of death (Cecchin et al., 2020). Within a subgroup analysis, the two polymorphisms seemed to identify subcohorts of patients with a poorer long-term prognosis within homogeneous TRG strata, making them suitable to be integrated with already available prognostic clinical parameters. A hypoxic microenvironment represents another factor that was suggested to cause both a deficiency in DNA repair and a genetic instability.

The hypoxia inducible factor 1 subunit alpha (HIF-1α, encoded by HIF1A) is an important mediator of hypoxia-induced radio-resistance. The association between hypoxia-related markers (i.e., HIF1A rs11549465, rs11549467, and rs2057482 polymorphism) and response to nCRT was analyzed but no significant association were highlighted (Havelund et al., 2012).

Germline variants impacting the activity of microRNAs, with a potential epigenetic control on the overall cellular gene expression pattern were also investigated. A single variation affecting the microRNAs activity could have a downstream down-regulation effect on a large number of genes, including those involved in crucial pathways for the response to radiotherapy as DNA repair, angiogenesis, and inflammation (Dreussi et al., 2012). The work of Dreussi et al., (2016b) evaluated a set of miRNA-related TagSNPs, potentially affecting miRNA maturation and activity, in a cohort of 270 Caucasian patients with LARC stratified in two subgroup according to the radiation dose (50.4Gy or 55.0Gy). SMAD family member 3 (SMAD3)-rs744910, SMAD3-rs745103, and trans-activation-responsive RNA-binding protein (TRBP)-rs6088619 were associated to an increased chance of pCR, while double-stranded RNA-specific endoribonuclease (DROSHA)-rs10719 and SMAD3-rs17228212 had an opposite detrimental effect on pathological tumour response. A classification and regression tree analysis highlighted that specific combination of SMAD3-rs744910 and TRBP-rs6088619 genotypes together with clinical features (i.e., longer interval time between the end of radiotherapy and surgery) increases the chance of pCR. The finding of three independent variants (i.e., located in different haploblocks) in SMAD3 provides a strong support for the involvement of this protein in the response to chemoradiotherapy. Another study (Sclafani et al., 2016) focused instead on a specific polymorphism, the rs4919510 C to G substitution that affects the mature microRNA 608 previously associated with response to treatment in patients with CRC. Within a retrospective analysis of the EXPERT-C phase II trial, the rs4919510-CC genotype was associated with worse 5 years PFS and OS after neo-adjuvant CAPOX followed by capecitabine-based chemoradiotherapy. The polymorphism probably impacts the interaction between miR-608 and its target mRNAs in a tissue-specific manner.

Most of the pharmacogenomic studies that investigated the response to nCRT in LARC have adopted a candidate gene or a pathway-based approach. Alternatively, two works opted for an unbiased strategy, performing a genome-wide analysis including thousands of polymorphisms. The study of Kim and colleagues (Kim et al., 2013), including 113 Korean patients with LARC receiving FP-based CRT, implemented a 3-step genome-wide strategy, based on genome-wide screening, clinical association, and biological validation of predictive polymorphisms. At the end, two novel markers were identified as potential predictors of response to treatment, coronin 2A (CORO2A) rs1985859 and refilin A (FAM101A) rs7955740. The reference CORO2A rs1985859-C allele was associated with higher rate of positive response. Moreover, an in vitro assay highlighted that the downregulation of CORO2A, linked to the variant rs1985859-T allele, was associated with reduced early apoptosis, increased cell survival or viability, and lower radiosensitivity. Even if FAM101A rs7955740 was not related with therapy outcome in the clinical association study, its downregulation, linked to the minor G-allele, was associated with a reduced early apoptosis and lower radiosensitivity similarly to CORO2A rs1985859. Another study (Lee et al., 2018), including a similar cohort of Korean patients with LARC and adopting a discovery/validation design, performed a whole-exome sequencing analysis identifying some further potential novel markers. Overall, five candidate variants emerged, Bcl-2-like protein 10 (BCL2L10) rs2231292, DLC1 Rho GTPase activating protein (DLC1) rs3816748, dynein axonemal heavy chain 14 (DNAH14) rs3105571, inter-alpha-trypsin inhibitor heavy chain 5 (ITIH5) rs3824658, and retinoic acid early transcript 1L (RAET1L) rs912565. Particularly, DLC1-rs3816748-C allele, DNAH14 rs3105571-C allele, and RAET1 rs912565-TT genotypes were associated with a higher rate of pCR according to a dominant model, while BCL2L10 rs2231292-CC genotype and ITIH5 rs3824658-T allele according to recessive model. In the co-dominant model, four candidate variants (all except BCL2L10 rs2231292) were significantly correlated with pCR. The identified polymorphisms potentially impact the functionality of the encoded proteins that are involved in crucial biological pathway as tumor suppression, transport along microtubules, and regulation of cell apoptosis, extracellular matrix stability, tumor invasion and metastasis. The markers of radiosensitivity emerged from genome-wide analyses, could represents an interesting candidate that require to be validated by larger independent studies.