Genomic variation as a marker of response to neoadjuvant therapy in locally advanced rectal cancer

ABSTRACT

There is variation in the responsiveness of locally advanced rectal cancer to neoadjuvant chemoradiation, from complete response to total resistance. This study compared genetic variation in rectal cancer patients who had a complete response to chemoradiation versus poor response, using tumor tissue samples sequenced with genomics analysis software. Rectal cancer patients treated with chemoradiation and proctectomy June 2006-March 2017 were grouped based on response to chemoradiation: those with no residual tumor after surgery (CR, complete responders, AJCC-CPR tumor grade 0, n = 8), and those with poor response (PR, AJCC-CPR tumor grade two or three on surgical resection, n = 8). We identified 195 variants in 83 genes in tissue specimens implicated in colorectal cancer biopathways. PR patients showed mutations in four genes not mutated in complete responders: KDM6A, ABL1, DAXX-ZBTB22, and KRAS. Ten genes were mutated only in the CR group, including ARID1A, PMS2, JAK1, CREBBP, MTOR, RB1, PRKAR1A, FBXW7, ATM C11orf65, and KMT2D, with specific discriminating variants noted in DMNT3A, KDM6A, MTOR, APC, and TP53. Although conclusions may be limited by small sample size in this pilot study, we identified multiple genetic variations in tumor DNA from rectal cancer patients who are poor responders to neoadjuvant chemoradiation, compared to complete responders.

Introduction

Chemotherapy and radiation therapy (CRT) prior to total mesorectal excision (TME) has become the standard treatment for patients with locally advanced rectal cancer. Randomized trials have demonstrated that neoadjuvant chemoradiation (nCRT) significantly improves local control compared with radiation alone or with postoperative CRT.¹ However, the response of rectal cancers to nCRT is quite variable, and thus treatment options with respect to optimal radiotherapy fractionation, dosing of chemotherapy, and timing of surgery are continuously being scrutinized. Up to 20% of rectal tumors exhibit complete resistance to nCRT, and patients with minimal or no response have been found to have worsened overall survival and disease-free survival.^2,3 Identifying a method to predict the response of rectal tumors to nCRT prior to surgery would help select appropriate treatment options for rectal cancer patients. Tumor response to nCRT is likely related to multiple variables, but tumor biology appears to play an important role and molecular biomarkers may help predict tumor responses to nCRT.⁴

There have been multiple studies reporting on various genetic and molecular biomarkers as predictors of response to nCRT in rectal cancer patients, but most of these reports have been inconclusive and non-reproducible to date.^5–9 Articles focusing on microarray gene expression have similarly failed to identify the optimal gene signature to predict tumor response.^6,7,10 In the current study, we chose to investigate potential mutations in the genetic profiles of rectal cancer patients before and after nCRT. We analyzed biomarkers in both the endoscopic and surgical specimens to compare patients who had a poor response to nCRT to those who had a complete response.

Results

Eight of seventeen specimens examined were identified to have a pCR and nine specimens were deemed PRs. Of the patients with pCR, age was a median 54.5 (range 36–63); there were three female patients and 5 were male. Of the patients with PR, age was a median 56 (range 38–70); there were two female patients and seven were male. There were no significant differences between CR and NR groups in age (p = .6672 by Mann Whitney U test) or in gender ratio (p = .6199 by Fisher’s Exact test).

The data were initially filtered to remove variants common in the general population (>1%) and include only potentially deleterious variants. One of the PRs contained 381 variants, and 201 of these variants (53.2%) were found exclusively in this patient. This PR was excluded from further analysis as an outlier representing either a hypermutated specimen,¹¹ or was potentially a result of processing or sequencing errors. This left eight patients in each group for comparison.

A total of 83 genes were found to be mutated with 195 unique variations seen in analysis of all specimens from all 16 patients. The mean number of mutations (standard deviation) per patient was 16.9 (11.2) variants found in 14.8 (8.1) genes (Table 1). In the endoscopic specimens from all patients, 81 mutated genes with 178 variants were identified, with a mean number of 13.9 (11.2) variants in 12.1 (8.1) genes per patient. In surgical resection specimens from all patients, 62 mutated genes with 116 unique variants were identified, with a mean of 9.4 (4.2) variants in 8.5 (3.9) genes per patient. Mean numbers of mutated genes and variants were higher in endoscopic specimens than surgical resection specimens in both PR and CR groups, although there was no statistically significant difference (numbers of genes, p = .4070 for all patients, p = .5635 for PR patients, and p = .2936 for CR patients; numbers of variants, p = .5095 for all patients, p = .6365 for PR patients, and p = .7003 for CR patients by Mann Whitney U test).

Table 1.

Number of variants/number of genes with a variant per patient. “Pretreatment Specimen” is the number of variants/genes with a variant found in the pre-treatment endoscopy sample. “Post-treatment Specimen” is number of variants/genes with a variant found in the final pathologic specimen after total mesorectal excision. Some variants/genes were found in both specimens.

Group/Patient Identifier	Pretreatment Specimen	Post-treatment Specimen	Total Variants per Patient
PR/1	5/5	12/10	14/12
PR/2	19/17	13/12	19/17
PR/3	11/11	6/6	11/11
PR/4	10/10	11/11	13/12
PR/5	15/15	10/10	23/23
PR/6	16/15	10/10	16/15
PR/7	10/9	9/9	11/10
PR/8	8/7	9/8	9/8
CR/1	7/6	16/16	20/18
CR/2	8/7	7/5	10/8
CR/3	7/7	3/3	8/8
CR/4	29/23	2/1	29/23
CR/5	10/9	9/8	13/12
CR/6	7/7	4/4	9/9
CR/7	9/8	11/9	13/11
CR/8	52/38	16/13	53/39
All Patients Average	13.9/12.1	9.4/8.5	16.9/14.8
PR Average	11.8/11.1	10.3/9.6	14.5/13.5
CR Average	16.1/13.1	8.5/7.4	19.4/16.0

PR = poor responders, CR = complete responders

PR patients had 91 variants in 55 mutated genes; of these, 64 variants in 42 genes were found in PR specimens only and were not seen in any CR patient sample. CR patients had 132 variants in 66 mutated genes; of these, 104 variants in 60 genes were found in CR specimens only. In endoscopic biopsy samples, PR patients had a mean (standard deviation) of 4.3 (4.0) variants in 3.9 (4.1) genes per patient that were unique to their endoscopic specimen, and CR patients had an average of 10.9 (13.4) variants in 8.8 (9.9) genes per patient that were unique to their endoscopic specimen.

Genes and variants in common between CR and PR patients

Genes with mutations in at least two endoscopic samples in both PR and CR groups are listed in Table 2, while specific gene variants that were present in both PR and CR patients are listed in Table 3. Mutations in PTEN (phosphatase and tensin homolog gene) were found in 15/16 patients, with all 15 having the same variant located at 10:89725294; one CR patient had an additional PTEN variant as well. Mutations in NF1 (neurofibromin gene) were found in 13 patients, with 12 of these having the same variant located at 17:29559062; two PR patients and two CR patients had one additional variation each as well. Mutations in ATM (ataxia-telangiectasia mutated gene) were found in 10 patients, with a variant at 11:108114784 found in five (two PR patients and three CR patients), a variant at 11:108114787 common to one PR and one CR patient, and six other variants also noted. Mutations in APC (adenomatous polyposis coli) were found in nine patients (five PR and four CR); 10 total variants were found with two CR patients having the same variant at 5:112151184. Mutations in GNAQ (guanine nucleotide-binding protein G(q) subunit alpha gene) were found in eight patients (five PR and three CR); all eight patients had the same variant at 9:80537223. Mutations in FGFR3 (fibroblast growth factor receptor 3) were found in six patients, with five patients (three PR and two CR) having the same variant at 4:1808017. Mutations in TP53 (tumor protein 53) were found in six patients (three PR and three CR); six total variants were found with two CR patients having the same variant at 17:7578406. Mutations in PIK3CA (p110α subunit of class I PI3-kinase) were found in five patients, with two patients (one PR and one CR) having the same variant at location 3:178942491 and four other variants found. Mutations in BRCA2 (breast cancer tumor suppressor-2) were found in four patients, with three patients (two PR and one CR) having the same variant at 2:48030632. Mutations in EP300 (p300 histone acetyltransferase) were found in four patients (two PR and two CR), with each patient having a unique variant. Mutations in MSH6 (mutS homolog 6) were found in four patients, with three patients (two PR and one CR) having the same variant at 2:48030632 and three other variations found. Seven other genes [SMARCA4 (SWI/SNF related, matrix associated, actin-dependent regulator of chromatin, subfamily a, member 4a), TSC2 (tuberous sclerosis complex subunit 2, tuberin), FLCN (folliculin), MET (mesenchymal–epithelial transition factor), ATRX (ATP-dependent helicase, X-linked helicase II, or X-linked nuclear protein), BRIP1 (BRCA1-interacting protein C-terminal helicase 1), PTCH1 (drosophila patched protein homolog 1), TSC1 (tuberous sclerosis complex subunit1, hamartin), NOTCH (neurogenic locus notch homolog protein 1), and SETD2 (SET domain containing, histone H3 lysine methyltransferase 2)] had mutations in four to seven patients, but did not show variants in at least two endoscopic specimens in each group. Fourteen additional genes had mutations in both groups in fewer numbers of patients; some of these also had the same variant in common in more than one patient (Supplemental Table 1).

Table 2.

Genes with variations found in endoscopy samples from at least two poor responder (PR) and two complete responder(CR) patients. Data is given as number of patients with variants, with the number having variants in endoscopic specimens in parentheses.

Gene	Number of PRs with variants	Number of CRs with variants	Total number of patients with variants
PTEN	7(7)	8(5)	15(12)†
NF1	6(6)	7(6)	13(12)
ATM	5(3)	5(3)	10(6)
APC	5(5)	4(4)*	9(9)
GNAQ	5(4)	3(3)	8(7)†
FGFR3	3(3)	3(3)	6(6)
TP53	3(3)	3(3)	6(6)
PIK3CA	2(2)	3(2)	5(4)
BRCA2	2(2)	2(2)*	4(4)
EP300	2(2)	2(2)	4(4)
MSH6	2(2)	2(2)*	4(4)

CR = complete responder, PR = poor responder,

* Patients had multiple variants of this gene. † All patients had same variant.

Table 3.

Variants found in both CR patients and PR patients.

Gene symbol	Chr:Position	Ref. allele	Samp. allele	Variation type	Gene region	Transcript variant	Protein variant	Translation impact	Classification	Number PR/CR patients
PTEN	10:89725294	T		Del	3ʹUTR	c.*65delT			Likely Benign	7/8
NF1	17:29559062	TT		Del	Intronic	c.3198-29_3198-28delTT			Likely Benign	6/6
GNAQ	9:80537223	T	G	SNV	Exonic	c.175A>C	p.M59L	MS	Uncertain Significance	5/3
ATM	11:108114784	C	A	SNV	Exonic	c.601C>A	p.Q201K	MS	Uncertain Significance	2/3
FGFR3	4:1808017	G	T	SNV	Exonic	c.1999G>T; c.1993G>T; c.1657G>T	p.A667S; p.A553S; p.A665S	MS	Uncertain Significance	3/2
FLCN	17:17118396	T	G	SNV	Exonic	c.1441A>C	p.T481P	MS	Uncertain Significance	1/3
TSC2	16:2138198	AGTG		Del	Intronic	c.4816-30_4816-27delAGTG; c.5092-30_5092-27delAGTG; c.4852-30_4852-27delAGTG; c.4429-30_4429-27delAGTG; c.4960-30_4960-27delAGTG; c.5161-30_5161-27delAGTG; c.4993-30_4993-27delAGTG			Uncertain Significance	1/2
MSH6	2:48030632	G	T	SNV	Exonic	c.2340G>T; c.3246G>T; c.2856G>T	p.P780P; p.P1082P; p.P952P	Syn	Likely Benign	2/1
TSC1	9:135786948	A	G	SNV	Exonic	c.768T>C; c.921T>C	p.A256A; p.A307A	Syn	Likely Benign	1/1
JAK2	9:5054843	T	C	SNV	Exonic; Intronic	c.448T>C; c.-280 + 54T>C; c.895T>C	p.S150P; p.S299P	MS	Uncertain Significance	1/1
BRCA2	13:32972626	A	T	SNV	Exonic	c.9976A>T	p.K3326*	SG	Uncertain Significance	1/1
TSHR	14:81422130	G	C	SNV	Exonic	c.106G>C	p.D36H	MS	Likely Benign	1/1
NF1	17:29508729	C	G	SNV	Exonic	c.656C>G	p.A219G	MS	Likely Pathogenic	1/1

CR = complete responder, PR = poor responder, Chr = chromosome, Ref = reference; Samp = sample; SNV = single nucleotide variation; MS = missence; Del = deletion; Ins = insertion; Sub = substitution; Syn = synonymous; SG = stop gain; FS = frame shift; IF = in-frame deletion.

Genes and variants discriminating between CR and PR patients

Three genes were noted to have variants in at least two endoscopic specimens of PR patients but were not found to be mutated in CR samples. These genes were KDM6A (lysine-specific demethylase 6A, four patients), ABL1 (proto-oncogene 1, non-receptor tyrosine kinase, three patients), DAXX-ZBTB22 (death domain associated protein-zinc finger – BTB domain-containing protein 22 complex, two patients), and KRAS (Kirsten rat sarcoma viral oncogene homolog, two patients) (Table 4). Six of the eight PR patients had a gene mutation in the pre-treatment specimen of one or more of four genes that can be found in the Wnt/β-catenin pathway: CTNNB1 (β-Catenin), APC (adenomatous polyposis coli), AKT (RAC-α serine/threonine-protein kinase, protein kinase B) or TP53. Six of the eight PR pre-treatment specimens had a mutation in at least one of seven genes known to be involved in the regulation of the Epithelial–Mesenchymal Transition (EMT) pathway: AKT1, APC, ATM (ataxia telangiectasia), CTNNB1, KRAS, MET, and PI3KCA. Of note, the APC, TP53, and MET genes had the highest number of mutations, in four, three and three patients, respectively, and KRAS had two variants identified as likely pathogenic; however, mutations in these genes were not exclusive to the PR group.

Table 4.

Genes with variations found in more than one endoscopic biopsy sample discriminating between PR and CR groups.

Gene	Number of PRs with variants (number in endoscopic specimens)	Number of CRs with variants (number in endoscopic specimens)
KDM6A	4(2)	0
ABL1	3(2)	0
DNMT3A	2(2)	0
DAXX; ZBTB22	2(2)	0
KRAS	2(2)	0
TSC1	2(1)	0
ARID1A	0	0
PMS2	0	3(3)
JAK1	0	3(3)
CREBBP	0	3(3)
MTOR	0	3(3)
RB1	0	3(3)
PRKAR1A	0	2(2)
FBXW7	0	2(2)
C11orf65; ATM	0	2(2)
KMT2D	0	2(2)

CR = complete responder, PR = poor responder.

When data was analyzed on the variant level by filtering to remove all variants present in the CR group, two variants were present in more than one endoscopic sample in the PR group that were not found in the CR group. A mutation in KDMA6A at X:44920542 was found in two of the PR patients. This variant is a deletion of T in an intronic region and has been classified as “uncertain significance for Kabuki Syndrome” by ACMG guidelines. A mutation in DNMT3A (DNA (cytosine-5)-methyltransferase 3-alpha) at location 2:25468946 was also found in two PR patients and none of the CR patient samples. This variant is a substitution of C for T variant in an intronic region and has been classified as “uncertain significance for T-cell lymphoproliferative disorder” by ACMG. In addition, eleven more genes had a mutation in the endoscopic specimen of a single patient each in the PR group (Supplemental Table 2).

Similarly, when variants present in the PR group patients were removed from the cohort by filtering, 10 genes with 24 variants were identified as mutated in two or more endoscopic specimens only in the CR group. These genes included ARID1A (three patients), PMS2 (two patients), JAK1 (three patients), CREBBP (three patients), MTOR (three patients), RB1 (two patients), PRKAR1A (two patients), FBXW7 (two patients), c11orf ATM (two patients), and KMT2D (two patients). Table 4 shows the number of PR patients and CR patients along with the number of patients who had variants found in pretreatment specimens for these discriminating genes.

On the variant level, three genes had specific variants present in CR patient endoscopic samples that were not found in PR samples. The variant of MTOR with substitution of C for G at 1:11316091 was found in two CR patients. This synonymous substitution is likely benign. The variant of APC with substitution of G for A at 5:112151184, seen in two CR patients, results in a missense translation and possible splice site loss; however, since the mutation is in an intron, it is classified as of uncertain significance. The variant of TP53 with substitution of T for C at 17:7578406, seen in two CR patients, also results in a missense translation, four possible protein variants, and is classified as pathogenic and damaging (Table 5). An additional 17 genes had a unique variation in the endoscopic specimen of a single patient each in the CR group (Supplemental Table 2).

Table 5.

Unique variants discriminating between PR and CR patients.

Gene symbol	Chr:Position	Ref. allele	Samp. allele	Variation type	Gene region	Transcript variant	Protein variant	Translation impact	Classification	Number PR/CR patients
DNMT3A	2:25468946	G	A	SNV	Intronic	c.974-13C>T; c.863-13C>T; c.1430-13C>T			Uncertain Significance	2/0
KDM6A	X:44920542	T		Deletion	Intronic	c.1195-27delT; c.1330-27delT; c.442-27delT; c.1351-27delT; c.1486-27delT		FS	Uncertain Significance	4/0
MTOR	1:11316091	G	C	SNV	Exonic	c.663C>G	p.T221T	Syn	Likely Benign	0/2
APC	5:112151184	A	G	SNV	Intronic	c.781-8A>G; c.835-8A>G		MS, splice site loss	Uncertain Significance	0/2
TP53	17:7578406	C	T	SNV	Exonic	c.47G>A; c.524G>A; c.128G>A; c.407G>A	p.R16H; p.R43H; p.R136H; p.R175H	MS	Pathogenic, Damaging	0/2

CR = complete responder, PR = poor responder, Chr = chromosome, Ref = reference; Samp = sample; SNV = single nucleotide variation; MS = missence; Del = deletion; Ins = insertion; Sub = substitution; Syn = synonymous; SG = stop gain; FS = frame shift; IF = in-frame deletion

Discussion

It is now well established that preoperative chemotherapy and radiation treatment reduces local recurrence of rectal adenocarcinoma when compared to postoperative treatment alone (6% vs. 13% after 5 y).¹ The standard of care for locally advanced rectal cancer now consists of preoperative CRT followed by surgical TME. However, there is great variability in the down-staging of rectal cancers after pretreatment. Importantly, a good or complete response to nCRT has been shown to be associated with an excellent oncologic outcome.^12,13 It would be clinically advantageous to identify a predictive measure of a patient’s response to preoperative treatment using gene variants identified in endoscopic biopsy tissue. Such a predictor could help treating physicians avoid the expense and morbidity of nCRT in rectal cancer patients deemed to be poor responders. In addition, manipulation of treatment options could be beneficial in assuring the best opportunity for tumor down-staging when patients are predicted to be inadequate responders.¹⁴ In our study, we examined genetic variations noted in endoscopic biopsy tissue of rectal cancer patients to see if there were patterns of genetic variance between those who did, and did not, respond to nCRT.

Genes and variants in common between CR and PR patients

The mutations most common to both CRs and PRs were in the PTEN and NF1 genes. Fifteen of the sixteen samples were noted to have a mutation in the PTEN gene in the surgical specimen, the endoscopy specimen, or both. Interestingly, all of these patients had the same variant. PTEN has been noted to be altered in several other cancers, including breast, prostate, thyroid, and endometrial cancers.¹⁵ Other studies similarly have found this gene to be an important contributor in colorectal cancer. Goel et al. demonstrated that hypermethylation (and thus silencing) of the PTEN gene has been noted in microsatellite-high colorectal cancer.¹⁶ The absence of PTEN in colorectal metastases has been noted to be predictive of a poor response to cetuximab therapy.¹⁷ As a result, methods of restoring PTEN activity are presently being investigated to improve response to chemotherapy. Recent research has indicated that overexpression of miR-22 or down-regulation of miR-21 can restore PTEN expression in colon cancer cells.^18,19 Our data seems to indicate that alterations in PTEN are present in both cancers that do and do not respond to chemotherapy. Nearly all of the specimens contained the same genetic variation in the PTEN gene. This may confirm an intrinsic role of PTEN in the formation of colorectal cancer.

Another gene that was noted to be highly altered was the NF1 gene, found in 6/8 PRs and 7/8 CRs. The NF1 gene encodes for neurofibromin, and loss of NF1 activity results in upregulation of the RAS/RAF/MAPK pathway, ultimately resulting in loss of control of growth and increased cellular proliferation.²⁰ While a recent study indicated that 5-6% of colorectal cancers contain alterations in this gene, it was unclear whether mutation of this gene is related to chemotherapeutic response.²⁰ Seventy-five percent (12/16) of specimens in the current study demonstrated a deletion at site 29559062 on chromosome 17 in the NF1 gene. This is an intron area of the gene and the deletion has therefore been classified as having no effect on gene function. As a result, the significance of this finding remains uncertain; however, the prevalence of this variant in our cohort may be indicative for a pathogenic role for NF1 in rectal cancer formation, which should be a target for future research.

Mutations in two genes that have not been described in colorectal cancer, GNAQ and ATM, were found in a high proportion of patients from both groups. GNAQ codes for a protein that couples transmembrane receptors to intracellular pathways.²¹ ATM, a serine/threonine kinase, is involved in cell division and repairing damaged DNA.²² ATM mutations have been described in several cancers, including breast, gastric, and pancreatic cancer.²³ Three of eight CR and 5/8 PR patients (50% overall) had GNAQ mutations, with all eight patients harboring the same variant of the mutation. That the same variant was present in each patient with a mutation strengthens the likelihood that this GNAQ mutation is clinically relevant and could play an important role in carcinogenesis. ATM mutations were present in 5/8 CR and 5/8 PR patients (62.5% overall), and 3/5 CRs and 2/5 PRs had the same variant. Similarly, given that 50% of all patients with ATM mutations had the same variant, there is increased likelihood that this ATM mutation could have clinical significance.

Variation in other genes such as APC (adenomatous polyposis coli)²⁴ and TP53 tumor suppressor protein²⁵ are well established in rectal cancers, and we did not find it surprising that a large number of mutations in these genes were found in our cohort. TP53 and APC have been identified as the first and third most common hotspots for mutation in a widespread lineage of cancers.²⁶ These genes, along with β-catenin, are involved in the well-known Wnt signaling pathway, dysregulation of which is often seen in cancers.²⁶ Alterations in the β-catenin degradation complex, in particular in APC, result in accumulation of β-catenin and constitutive activation of target genes such as Myc and cyclinD1 that can promote cancer formation.²⁷ Four of the eight PR endoscopic specimens exhibited genetic variants in the APC protein, and two in β-catenin. Three of these APC variants were the result of an insertion or a deletion, which likely resulted in a frameshift and a nonfunctional protein. The two variants in the β-catenin gene were of unknown significance. Mutations in the APC gene resulting in a nonfunctioning protein and prolonged activation of the Wnt pathway through β-catenin transcriptional activity endow the cells with a stem cell-like phenotype, arresting differentiation and suppressing migration. In this state, colorectal cancer cells exhibit an increased resistance against external inhibitory signals, including cytotoxic drugs.²⁸ Presence of an embryonic stem cell-like phenotype has been predictive of initial response to cisplatin and fluorouracil combination chemotherapy in gastric cancer.²⁹

FGFR3 (fibroblast growth factor receptor 3) mutations were found in endoscopic samples from three CR patients and three PR patients. FGFR3 proteins are involved in activation of the RAS-MAPK pathway and the PI3K-AKT pathway, among others that can lead to cancer.^30,31 Coding mutations and deletions have been seen in many cancers,^32,33 and FGFR fusions have recently been demonstrated to have oncogenic potential.^34,35 Over-active FGFR signaling has been successfully treated with the targeted therapeutics nintedanib, palifermin, fostamatinib, lenvatinib, ponatinib, dovitinib, and pazopanib,³⁶ prompting diagnostic sequencing of FGFR genes, especially in lung cancer patients.³⁷ Activation of PI3K-AKT is involved in the epithelial the mesenchymal transition pathway, where it correlates with tumor progression and is also associated with poor response to neoadjuvant 5FU-based chemoradiation.³⁸

All of the genes listed in this section as well as others [SMARCA4 (also known as BRG1), PIK3Ca (phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha), BRCA3 (breast cancer 3, also known as BRACX), EP300 (E1A binding protein p300), MSH6 (mutS homolog 6DNA mismatch repair protein), which is implicated in Lynch syndrome, and FLCN (folliculin tumor suppressor, also called BHD)] were found to be prevalent in both the complete responders and poor responders. We believe this signifies that these genes play an intrinsic role in rectal cancer development and are less likely to predict whether a rectal cancer will respond to neoadjuvant treatment.

Genes and variants discriminating between CR and PR patients

We identified four genes that were found to be mutated in at least two endoscopic specimens of patients who were PRs to neoadjuvant treatment and were not found to be mutated in any of the specimens from CRs. These genes are KDM6A, ABL1, DAXX-ZBTB22, and KRAS (Table 4).

Mutations in KDM6A (lysine demethylase 6A, also known as UTX) were found in four PR patients with an allele fraction of ~10%. All four patients had the same variant at X:44920542. This gene codes for a histone demethylase, which is believed to act as a tumor suppressor. Homozygous and hemizygous somatic mutations in KDM6A have been identified in various cancer types, including multiple myeloma, esophageal squamous cell carcinoma, renal cell carcinoma, myeloid leukemia, breast cancer, pancreatic cancer, and glioblastoma.^39,40 In acute myeloid leukemia, KDM-A depletion has been associated with decreased sensitivity to AraC (Cytosine arabinoside), an antimetabolic agent that interferes with DNA do novo synthesis, as does the fluoropyrimidines family of chemotherapeutics used in our rectal cancer patients.⁴¹ KDM6A-deficient pancreatic cancer has been found to be sensitive to BET inhibitors, a class of drugs that reversibly bind the bromodomains of the Bromodomain and Extra-Terminal motif (BET) proteins BRD2, BRD3, BRD4, and BRDT, resulting in suppression of downstream activation of RNA polymerase II.⁴² BET inhibitors have also been used to successfully treat NUT midline carcinoma⁴³ and some forms of acute myeloid leukemia, multiple myeloma, and acute lymphoblastic leukemia.^44–46 This pathway may present a potential therapeutic target for otherwise chemoresistant rectal cancer.

ABL1 (ABL proto-oncogene 1, non-receptor tyrosine kinase) has been shown to be involved in multiple oncogenic pathways for leukemia and solid tumors, including in gastric and colorectal cancer. Mutations in ABL1 were found in three PR patients in our study. ABL1 is a proto-oncogene that encodes a tyrosine kinase normally involved in cell differentiation, division, adhesion, and response to stress. Targeted therapy directed at the ABL1-BCR fusion protein that is found in chronic lymphocytic leukemia (CLL) and acute myeloid leukemia (AML) has revolutionized the treatment of these cancers.⁴⁷ In addition, ABL1 inhibitor therapy has been used to overcome drug resistance in CLL patients resulting from the up-regulation of antiapoptotic Bcl-xL, A1/Bfl-1, and Mcl-1 proteins through activation of CD40, including in patients with defective p53.⁴⁸ Further mutations in ABL1 confer resistance to these agents;⁴⁹ however, the vulnerability of specific isoforms of ABL1 to inhibition offers another avenue for drug development for rectal cancer patients with a common mutation who may be resistant to other treatment regimens.

Two PR patients had mutations in the DAXX (death domain-associated protein 6, BING2) – ZBTB22 (Zinc Finger and BTB domain-containing 22, BING1) complex. The two genes reside adjacent to each other on chromosome 6p21.32, 180 kbp centromeric of the HLA-DP locus of the major histocompatibility complex in a region linked to several diseases.⁵⁰ ZBTB22 is a transcriptional regulator of DAXX. A second copy of each gene with alternative splicing for the DAXX transcript appears to reside in an otherwise uncharacterized region approximately 32 kpb upstream from the GeneLoc location for DAXX-ZBTB22. Mutations seen in our subjects are within the DAXX coding region of the alternative site. DAXX interacts with the apoptosis antigen FAS and the centromeres in the G2 phase of the cell cycle, activating apoptosis signal-regulating kinase 1 through the Jun N-terminal kinase pathway.⁵¹ DAXX also interacts with AXIN1; the interaction stimulates the interaction of DAXX with TP53, the phosphorylation of TP53, and induces cell death on UV irradiation.⁵² DAXX-depleted non-cancer cells are resistant to UV irradiation or oxidative stress-induced cell death.⁵³ In HCT116 cells, DAXX confers resistance to 5-fluorouracil-induced apoptosis mediated through p53.⁵⁴ As patients with DAXX-mutations were resistant to 5-FU in our study, the particular mutations seen may have increased DAXX activity, affected its interaction with p53, or interfered with protein translocation across cell compartments.

KRAS was found to be mutated in two PR patients. KRAS is a ubiquitous GTPase proto-oncogene that is an early member of the RAS/MAPK signal transduction pathway and is involved with cellular proliferation. KRAS transcriptionally upregulates the mRNA levels of NRF2 through a TRE enhancer located in the proximal promoter of NRF2. Somatic mutations in KRAS are found at high rates in a variety of cancers types⁵⁵ including colorectal cancers,⁵⁶ but the specifics of the mutations vary by organ systems. The A146 mutation in KRAS is found only in colorectal cancer, while pancreatic, lung adenocarcinoma, and gastric cancer contain mutations in different codons of KRAS.²⁶ KRAS mutation in colorectal cancer has previously been reported to be predictive of a poor response to therapy with the epidermal growth factor receptor (EGFR) inhibitors panitumumab and cetuximab.⁵⁷ In addition, KRAS gene amplification can affect response to treatment. Inhibition of the NRF2 pathway by brusatol was able to overcome KRAS-mediated chemoresistance and enhanced the efficacy of cisplatin in non-small cell lung cancer.⁵⁸ Mutant allele-specific imbalance arising through chromosome 12 hyperploidy or KRAS amplification that affects KRAS codon 13 is associated with worse overall survival in colorectal carcinoma.⁵⁶ These tumors are not responsive to treatment with EGFR inhibitors. Testing for wild-type KRAS is currently used in clinical decision-making when EGFR inhibitor treatment is being considered, as cetuximab given with chemotherapy is up to 59% more effective in colorectal cancer patients with no KRAS mutations.⁵⁹

Two PR patients had heterozygous mutations in DNMT3A (DNA methyltransferase 3 alpha), both at the same locus (2:11316091) with an allele fraction of ~34% and 50%, respectively. The protein coded is an arginine methyltransferase, an epigenetic modifier of CpG sites important for embryonic development and X-chromosome inactivation. DNMT3A mutation has been associated with poor outcomes in cytogenetically normal acute myeloid leukemia⁶⁰ and has also been associated with lung adenocarcinoma.²⁶ Guryanova et al. showed that DNMT3A-mutated cells did not exhibit nucleosome and chromatin remodeling in response to anthracycline treatment, which resulted from attenuated recruitment of histone chaperone SPT-16. This defect led to an inability to sense and repair DNA torsional stress, which resulted in failure of the expected production of DNA breaks and increased mutagenesis after anthracycline exposure.⁶¹ Shen et al. reported that DNMT3A mutation may not only associated with increased cell proliferation in acute myeloid leukemia but also with reduced apoptosis of the leukemic cells; either of these could results from the same chemoresistance mechanism.⁶²

When looking at the complete responder group, mutations were noted in endoscopic biopsy specimens from at least two patients in genes including ARID1A, JAK1, CREBBP, and MTOR (three patients each), and PMS2, RB1, PRKARIA, FBXW7, c11orf65, and KMT2DA (two patients each) that were not seen to be mutated in PR specimens. The ARID1A gene encodes proteins with helicase and ATPase activities of the SWI/SNF family. These proteins regulate transcription by altering the chromatin structure around target genes.⁶³ PMS2 (post-meiotic segregation homolog 2, mismatch repair protein) is an endonuclease that induces single-strand DNA breaks near a mismatch for exonuclease cleavage and degradation of damaged DNA. PMS2 mutations are well known in inherited cancers, including mismatch repair cancer syndrome and hereditary colorectal cancer including hereditary nonpolyposis colorectal cancer (Lynch syndrome).^64,65 JAK1 mutation resulting from concordant mismatch repair deficiency has been implicated in microsatellite instability in colorectal tumors, immune evasion, and resistance to anti-PD-1 treatment; however, clinical follow-up indicated that patients with JAK1 mutations had a higher overall survival rate than patients wild type for JAK1.⁶⁶ CREBBP (CREB binding protein) is a transcriptional coactivator and molecular scaffold.⁶⁷ Somatic mutations of CREBBP have been identified in ovarian cancer.⁶⁸ Chemotherapy treatment-related translocations in CREBBP can result in AML, CML, and myelodysplastic syndrome.⁶⁹ MTOR (mammalian target of rapamycin) is a serine/threonine kinase that is part of the P13K/AKT pathway, a central regulator of cellular metabolism, growth, and survival. This pathway has been heavily implicated in many cancer types,⁷⁰ and several inhibitors of MTOR are in clinical use, including everolimus, temsirolimus, and zotarolimus, with varying degrees of success.^71,72 PMS2 (PMS1 homolog 2, mismatch repair system endonuclease) interacts with MUT1 to form the MutL-α heterodimer. Along with MutS-α and MutS-β, this complex corrects DNA mismatches and small insertion/deletion loops. Mutations in PMS2 are associated with Lynch Syndrome and Turcot syndrome;^73,74 however, studies have not determined if mutational status confers any association with chemoradiation resistance in these cancers. RB1 (retinoblastoma transcriptional corepressor 1) is a negative regulator of the cell cycle and was the first tumor suppressor gene identified.⁷⁵ Defects in RB1 have been associated with sensitivity to neoadjuvant treatment with radiotherapy and cytotoxic agents such as cisplatin in bladder cancer.⁷⁶ PRKARIA (protein kinase, cAMP-dependent, regulatory subunit type I alpha, also known as tissue-specific extinguisher 1 and CNC1) is mutated in Carney complex syndrome type I, characterized by multiple neoplasia including endocrine tumors;⁷⁷ in sporadic adrenocortical tumors with presentation of ACTH-independent Cushing’s syndrome;⁷⁸ in intracardiac myxoma, a benign neoplasm that results in decreased lifespan due to stroke and obstructive heart failure;⁷⁷ and in papillary thyroid carcinoma as a result of translocation between chromosomes 10 and 17, resulting in activation of the RET tyrosine kinase domain in thyroid cells.⁷⁹ FBXW7 (F-box and WD repeat domain-containing 7) is a member of the ubiquitin-proteasome system involved in the degradation of proto-oncogene products such as Notch1, c-Myc, cyclin E, Mcl-1, MTOR, and Jun, thus acting as a tumor suppressor;⁸⁰ mutations in FBXW7 may also be oncogenic in themselves under specific circumstances.⁸¹ It is located at chromosome 4q32, a region deleted or mutated in up to 30% of human cancers.⁸² FBXW7 has been shown to be involved in repair of radiation damage by the non-homologous end-joining mechanism, thus mutation may make patients more sensitive to neoadjuvant chemoradiation.⁸³ The c11orf65 mutation is an open reading frame on chromosome 11 at location q22.3, near the gene for ATM. The protein product is largely uncharacterized but is expressed in multiple tissue types and is associated with T-cell prolymphocytic leukemia, as are mutations in ATM and KMT2DA. KMT2DA is a member of the KMT2 (MLL, mixed-lineage leukemia protein) family of histone-lysine N methyltransferases.⁸⁴ It has been identified as mutated in a significant frequency of breast and colorectal cancers,⁸⁵ and regulates cell cycle progression and viability in colon cancer.⁸⁶ Several small-molecule inhibitors that disrupt the interaction of the histone methyltransferases with WDR5 (WD40 repeat protein 5) are under development as targeted therapeutic agents in MLL, including vorinostat and romidepsin in cutaneous T-cell lymphoma.⁸⁷

On the variant level, specific variants seen in more than two endoscopic samples from patients in the CR group and not in the PR group were MTOR, APC, and p53, which were discussed above. These genes and others noted to be unique to the patients in this study should be further investigated as to their role in colorectal cancer and chemoradiation sensitivity.

Heterogeneity

We analyzed the surgical specimens and endoscopic specimens of CR and PR patients to determine if there was a reduction in genetic heterogeneity following neoadjuvant treatment. When these samples were compared, there was a slight reduction in the numbers of variants found following neoadjuvant treatment in all categories (Table 1), and a greater reduction in CR samples than PR samples. There was a 13% reduction in the total number of deleterious variants/genes in the PRs, and nearly a 50% reduction in the total number of variants/genes in the CRs, although this reduction did not achieve statistical significance, likely due to the small sample size. Our results are consistent with others, and indicates that there is less heterogeneity in cancer cells following treatment.⁸⁸ However, it is interesting to note that our specimens demonstrated greater heterogeneity in the pretreatment specimens of subjects who would subsequently respond to chemoradiation. The reason for this is unclear. Although the variants that were most common in post-treatment specimens obviously cannot be used as pre-treatment predictors of chemoradiation response, analyses of the gene variants seen in post-treatment PR samples may eventually provide insight into pathways that might be responsible for chemoradiation tolerance. A greater numbers of variants unique to the surgical sample in CRs were also noted, with uncertain significance, as minimal or no tumor cells were present in the post-treatment samples analyzed for this group.

Limitations

The authors recognize some notable limitations of this study. First of all, the sample size in this pilot study was small, with 16 total patients (after exclusion of ineligible patients). A larger number of patients would be needed to fully power the study and demonstrate patterns of molecular biomarkers. With a larger sample number, a multi-variant biomarker panel could be determined from a subset of samples and then validated using tissue from additional subjects. Further experiments could be performed in vitro using fresh biopsy tissue that could confirm the hypothesis that cells containing identified variants had greater or lesser susceptibility to chemotherapy. In addition, this was a prospective analysis of retrospective data over a long time-period; therefore, there could have been minor degradation of tissue used for DNA analysis in some of the samples, although this has not been observed in other studies using similarly archived tissue at this institution. Another limitation was that genetic analysis was performed on CR post-treatment surgical samples. This means that genetic analysis was performed on specimens that theoretically did not contain cancer. If the surgical specimen did not contain cancer, then the genetic variants found may either be variants that predispose to cancer, or potentially may be variants related to the scar tissue that forms as the cancer regresses. If they are in fact variants within cancer, then the patients who met pathologic criteria for CR may not have actually achieved a complete response to therapy on the molecular level, which would possibly indicate that all patients should undergo surgical resection. Another limitation is that the study only considered variants within 160 cancer panel genes, and there may be many other potential markers that were not examined in this study. As tools for genetic analysis become more efficient and cost-effective, future studies should include analysis of additional markers, particularly including genes that may be associated with drug and radiation resistance.

Conclusions

In comparing rectal cancer patients who were complete responders or poor responders to preoperative CRT, multiple genetic variations were identified. Several genes and variants were common to both groups, including PTEN, NF1, GNAQ, ATM, SMARCA4, APC, and p53. We believe that these represent genes that are altered in most rectal cancers, although some of them have not been previously associated with rectal cancer. There were several genes and variants that were noted to be unique to each group. Mutations in genes that may identify cancers refractory to neoadjuvant therapy include KDM6A, ABL1, DNMT3A, DAXX, and KRAS. Gene mutations in other genes may indicate a predisposition to respond to neoadjuvant therapy. These genes include ARID1A, PMS2, JAK1, CREBBP, RB1, PRKAR1A, FBXW7, c11oft65 ATM, and KMT2D, as well as particular specific variants in MTOR, APC, and TP53. These genes may form the basis for a panel to identify rectal cancers which may or may not respond to neoadjuvant treatment based on the genetic profile of an endoscopic biopsy specimen. Further confirmatory research using a larger number of patient specimens is required.

Patients and methods

This study was approved by an Institutional Review Board and incorporated the institution’s rectal cancer database and Biobank repository. A waiver of consent was granted to use previously stored tissue. Seventeen patients diagnosed with locally advanced rectal cancer who were treated with nCRT followed by radical surgery were identified between June 2006 and March 2017. These patients had confirmed rectal adenocarcinoma on endoscopic biopsy, and underwent staging with endorectal ultrasound and CT imaging of the chest, abdomen, and pelvis. All patients then received preoperative fluorouracil (5-FU) or capecitabine (Xeloda), accompanied by a total dose of 45–54 Gy radiation. Patients underwent TME between 6 and 16 weeks after nCRT. As a pilot study, the 17 patients chosen for analysis were rectal cancer patients for whom both suitable endoscopic biopsy and post-treatment surgical resection tissue specimens were available, as well as CT imaging and complete records of neoadjuvant treatment.

Preoperative endoscopic biopsy specimens and postoperative surgical specimens were zinc formalin fixed and processed at the time of surgery to paraffin-embedded blocks. Histology was assessed by one pathologist (WL). Each surgical specimen was given a tumor regression grade (TRG) in concordance with the American Joint Committee on Cancer (AJCC) criteria.⁸⁹ The specimens were classified into four TRG groups: 0 – complete response (no viable cancer cells), 1 – moderate response (single cells or small groups of cancer cells), 2 – minimal response (residual cancer outgrown by fibrosis), 3 – poor response (minimal to no tumor kill). TRG2 and TRG3 tumors were grouped and analyzed as “poor responders” (PR) and TRG0 tumors were deemed pCR. For clarity, TRG1 tumors were not included in the analysis cohort.

DNA was isolated from pre-treatment tumor biopsy and post-treatment surgical specimen formalin-fixed paraffin-embedded tissue blocks. Multiple tissue core samples were selected from areas of malignancy as identified by the pathologist. For some samples, laser capture microscopy was utilized to select malignant cells. DNA extraction and quantification were performed using the Qiagen’s GeneRead DNA FFPE Kit and GeneRead DNA Quantimize Kit according to manufacturer’s instructions. DNA libraries were prepared using Qiagen’s Comprehensive Cancer Panel that focuses on 160 cancer-related genes. Libraries were sequenced on the Illumina NextSeq 500. Alignment was done using Biomedical Genomics Workbench with subsequent biological interpretation with Ingenuity Variant Analysis.

Filters were applied to exclude variants of lesser sequencing confidence and those common in the general population (variants observed with an allele frequency greater than or equal to 1.0% of the genomes in the 1000 genomes project, the NHLBI Exome Sequencing Project, the Allele Frequency Community Genome Aggregation Database and CGI from Complete Genomics and the Exome Aggregation Consortium). Data were then filtered to focus on variants likely to have a negative effect on gene/protein function as known or possibly pathogenic, associated with a gain of function, established as deleterious in the literature, resulting in gene fusion, inferred activation mutation by Ingenuity, or predicted gain of function by BSIFT.

The data were analyzed at both the gene and variant level to identify mutations that were found only in the PR group that may be predictive of resistance to chemoradiation, only in the CR group that may be predictive to radiation sensitivity, or in both groups that may indicate predisposition to rectal cancer. Finally, the pre-treatment biopsy specimens were compared to the final surgical specimens in each group to examine changes in tumor gene heterogeneity with nCRT.

Funding Statement

Funding from a private donor (Michael Serling) was received for this study. The funder had no role in the in design of the study, collection, analysis, and interpretation of data, or manuscript writing.

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics approval and consent to participate

This study was approved by the Beaumont Health Institutional Review Board (2011-255). A waiver of consent was granted to use previously stored tissue.

Competing Interests

The authors declare that they have no competing interests.

Presentation

Podium Presentation at the American Society of Colon and Rectal Surgeons 2018 Annual Meeting, May 19-23 2018, Nashville, TN, USA.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.