Stockton 

et al. Radiat Oncol (2021) 16:129

https://doi.org/10.1186/s13014-021-01853-y

RESEARCH Open Access

Complete response to neoadjuvant

chemoradiotherapy in rectal cancer
is associated with RAS/AKT mutations and high
tumour mutational burden
Joanne D. Stockton1, Louise Tee1, Celina Whalley1, Jonathan James1, Mark Dilworth1,2, Rachel Wheat1,
Thomas Nieto1,2, S-CORT Consortium, Ian Geh2, João D. Barros‑Silva1† and Andrew D. Beggs1,2*†   

Abstract 
Background:  Pathological complete response (pathCR) in rectal cancer is beneficial, as up to 75% of patients do not
experience regrowth of the primary tumour, but it is poorly understood. We hypothesised that the changes seen in
the pre-treatment biopsies of pathCR but not seen in residual tumour after chemoradiotherapy were the determi-
nants of responsiveness.
Methods:  Two groups of patients with either complete response (pathCR group, N = 24) or no response (poor
response group, N = 24) were retrieved. Pre-treatment biopsies of cancers from these patients underwent high read
depth amplicon sequencing for a targeted panel, exome sequencing, methylation profiling and immunohistochemis-
try for DNA repair pathway proteins.
Results:  Twenty four patients who underwent pathCR and twenty-four who underwent poor response underwent
molecular characterisation. Patients in the pathCR group had significantly higher tumour mutational burden and
neoantigen load, frequent copy number alterations but fewer structural variants and enrichment for driver mutations
in the PI3K/AKT/mTOR signalling pathway. There were no significant differences in tumour heterogeneity as measured
by MATH score. Methylation analysis demonstrated enrichment for hypomethyation in the PI3K/AKT/mTOR signalling
pathway.
Discussion:  The phenomenon of pathCR in rectal cancer may be related to immunovisibility caused by a high
tumour mutational burden phenotype. Potential therapy resistance mechanisms involve the PI3K/AKT/mTOR signal-
ling pathway, but tumour heterogeneity does not seem to play a role in resistance.

Introduction
Rectal cancer is a common malignancy [1], with approxi-
mately 11,000 cases per year in the UK [2]. Treatment
typically consists of excisional surgery [3] with neoadju-
vant therapy if the cancer is locally advanced, consisting
*Correspondence: [email protected] of either short course radiotherapy [4] (25  Gy in 5 frac-
†
João D. Barros-Silva and Andrew D. Beggs have contributed equally to
tions over 1  week) with surgery the following week, or
the manuscript.
1
Surgical Research Laboratory, Institute of Cancer and Genomic Science, long course neo-adjuvant chemoradiotherapy (nCRT) [5]
University of Birmingham, Vincent Drive, Birmingham B15 2TT, UK (45–50 Gy in 25 fractions over 5 weeks, with synchronous
Full list of author information is available at the end of the article

© The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​
mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
Stockton et al. Radiat Oncol (2021) 16:129 Page 2 of 12

iv 5-FU or oral capecitabine) with surgery 6–10  weeks pathway is associated with a longer survival, presum-
later. The former regimen usually demonstrates little, if ably as a consequence of better response in radiotherapy
any tumour regression, but if it does occur, is associated [20], however the data concerning the role of NHEJ and
with more favourable prognosis, and the latter regimen the response to radiotherapy is unclear, with studies giv-
can lead to significant tumour shrinkage and down- ing conflicting reports of its relationship with radiation
staging, with pathological complete response (pathCR, response [21]. Lal et al. [22] also showed that the immune
defined as complete regression of tumour in the resection context of a tumour depended on genomic factors, with
specimen) observed in approximately 10–15% of patients KRAS mutation, CMS2 or CMS3 classification all being
[6]. Multiple investigators have shown higher rates of independently associated with a reduced immune infil-
response [7, 8] with higher radiotherapy doses, with tration. This has implications for responsiveness to ther-
maximum pCR rates of 25–30%. The current standard apy, as a large proportion (40%) of colorectal cancer has
of care within the United Kingdom for locally advanced KRAS mutation, and immunovisibility of the tumour is
rectal cancer is long course chemoradiotherapy, however essential for a good response to neoadjuvant therapy.
the recent development of the concept of “total” neo- Tumour heterogeneity undoubtedly also plays a role
adjuvant therapy (TNT), whereby consolidation chemo- in determining pathCR [16, 23], as does immunogenic-
therapy is given after chemoradiotherapy, increased pCR ity caused by the formation and expression of clonal
rates from 15 to 25% in the consolidation group, acting as neoantigens. Rectal cancers have little hypermutation
an additional therapeutic option in these patients which and therefore rates of immunovisibility are low. Whilst
may lead to its introduction as standard of care [9]. Short Akiyoshi et  al. [1] have recently shown that levels of
course radiotherapy with a delay to surgery [10], or with clonal neoantigens are higher in patients undergoing a
neoadjuvant chemotherapy in the RAPIDO trial [11], has good response to neoadjuvant chemoradiotherapy, sug-
also been shown to be of benefit and demonstrates the gesting that hypermutation is important, a previous inte-
critical nature of the sequencing and timing of treatment grated molecular analysis of rectal cancer [16] found no
in order to maximise response. key genomic features that correlated with resistance. This
Clinical complete response (CCR) is defined as the may be because of the significant heterogeneity of the
absence of tumour on imaging and/or clinical examina- datasets used in the analysis.
tion, but does not definitively exclude residual tumour, Owing to the uncertainty surrounding the precise
which requires a resection specimen to confirm. CCR is mechanisms of sensitivity of rectal cancer to chemoradi-
well correlated with pathCR and, this could allow routine otherapy, we aimed to study the phenomenon of pathCR
use of a watch and wait strategy. In addition to its role in rectal cancer. We hypothesised that the genomic
as an indicator of potential cure [12] (defined as > 5 years changes responsible for the phenomenon of pathCR
recurrence free), path CR could be used to delay exci- would be seen uniquely in the pre-treatment biopsies of
sional surgery [13] or allowing organ preservation sur- those undergoing a complete response, and not seen at
gery [14], such as TEMS or TAMIS [15]. all in the residual post-treatment specimens of patients
The molecular drivers of pathCR are unclear [16, 17], who had undergone neo-adjuvant chemoradiotherapy
but they are thought to relate to factors that promote and therefore would represent two divergent opposites to
radiotherapy- and chemotherapy-related tumour cell kill- identify the phenomenon.
ing. For example, rectal cancers exist within a low oxygen
tension environment [18] leading to intrinsic resistance. Methods
As radiotherapy induces the formation of oxygen derived Patients
free radicals, which cause tumour cell death by direct A prospective database of all patients undergoing neo-
DNA damage, a low oxygen environment leads to fewer adjuvant chemoradiotherapy for rectal cancer was used
available oxygen free radicals, leading to lower cell death. to identify patients. Ethical approval was obtained from
Another hypothesised mechanism underlying differences the NW Research Ethics committee (ref 15/NW/0079).
pathCR is variation in DNA repair. Ionising radiation Patients who underwent long course chemoradiother-
[19] induces a DNA double strand break (DSB), which is apy and achieved pathCR were identified, as determined
then repaired either by homologous recombination (HR), by complete regression of the tumour, with absolutely
where a sister chromatid is used to repair the defect, or no tumour cells remaining (Mandard grade 1/ TRG 1),
non-homologous end joining (NHEJ), where a complex on examination of the specimen by a Consultant His-
repair mechanism including BRCA​, ATM, ERCC5 and topathologist, as opposed to minimal residual disease
others contribute to a DNA repair complex that re-joins where several cells were allowable. All patients under-
the damaged segments of DNA. There is evidence from went long course chemoradiotherapy with either oral
previous studies that aberrant functioning of the NHEJ capecitabine (825  mg/m2) or infusional 5-flourouracil
 Stockton et al. Radiat Oncol (2021) 16:129 Page 3 of 12

during a radiotherapy course of 45  Gy in 25 fractions Genomics

over 35  days. Pre-treatment stage and post-treatment Extracted genomic DNA was used for targeted amplicon
response (at 6  weeks after finishing treatment) was resequencing using the Fluidigm 48 × 48 Access array.
assessed using magnetic resonance imaging. Resection PCR amplicons covering APC, KRAS, BRAF, NRAS,
of the primary tumour then occurred as soon as possible FBXW7 and SOX7 (primer sequences available on
after the 6  week MRI scan. The histopathology archives request) were designed using Primer3 [25]. Primer speci-
were then searched to find the pre-treatment endoscopic ficity was checked using PrimerBLAST and UCSC in-
biopsies of these patients for downstream analysis. This silico PCR. Briefly, 20 ng of FFPE DNA was injected into
was defined as the “pathCR cohort”. A randomly selected the Access Array system with PCR primers and thermal
second cohort of rectal cancer patients was identified, cycled according to manufacturer’s specifications. Ampli-
who had no response to treatment or progressed whilst cons were then ligated to Illumina sequencing indexes &
on treatment, as defined both by post-treatment MRI adapters and pooled and sequenced on an Illumina MiSeq
and Mandard grade 4 or 5. This was defined as the “non- to an average read depth of > 1000  × using a 100  bp PE
responder cohort”. For the validation cohort, patients sequencing strategy. For FFPE exome sequencing, a cus-
from work stream 3 of the Stratification in Colorectal tom modification of the Illumina TruSeq Exome hybridi-
cancer (S-CORT) were utilised, which represent a cohort sation kit was used. At least 300 ng of FFPE-derived DNA
of patients undergoing long course chemoradiotherapy was prepared with the following modifications: Firstly,
for whom pre-treatment biopsy material was available. no size selection was performed after end repair and
For germline DNA, representative normal tissue from the DNA fragments were amplified with 12 cycles of PCR.
proximal resection margin of the discovery cohort was Enrichment was performed using a bead ratio of 0.8, then
obtained for genomic analysis. samples were combined into pools of 3 plex for coding
exome (TruSeq exome, 45mb in size) probe hybridisation
Samples and subsequent clean up. 10 cycles of amplification were
Formalin fixed blocks were retrieved for these patients performed to enrich the final libraries which were then
and cut into 4uM sections on frosted glass slides for nee- pooled into a final 12 plex library. Sequencing was per-
dle macrodissection and immunohistochemistry. Pre- formed on an Illumina NextSeq. For the S-CORT WS3
treatment specimens were all endoscopically obtained samples, a custom panel consisting of 61 oncogenes was
biopsies, whereas post-treatment specimens consisted sequenced to 500 × using Agilent SureSelect bait capture
of tumour blocks selected by a histopathologist as rep- at the Wellcome Trust Sanger Institute. Methylation was
resentative of the tumour. This consisted of the block interrogated using the Illumina HumanMethylation 450
whereby the H&E section showed maximum tumour array system. Between 100 and 250 ng of FFPE DNA was
content. A representative H&E section was used to target processed using the Illumina FFPE restore kit, and then
tumour cells for macrodissection to maximise tumour hybridised to the HumanMethylation 450 array following
content for DNA extraction using a modified protocol of the manufacturer’s instructions. Slides were washed and
the Qiagen DNEasy kit (Qiagen). Eluted DNA was quan- scanned on an Illumina iScan scanner.
tified using Nanodrop spectrophotometry (for contami-
nants) and Qubit fluorimetry (for concentration). Imaging mass cytometry
Slide staining and CyTOF were performed on FFPE sec-
Immunohistochemistry (IHC) tions using methods as previously described [26]. CyTOF
This was performed on 4uM slides as previously IMC data were analysed using an image-processing
described using a Leica BondMax autostainer. IHC was pipeline as described (https://​github.​com/​Boden​mille​
performed against yH2AX (Abcam ref ab26350), ATM rGroup/​imcto​ols). Ilastik generated Probability probabil-
(ab36810), Ku70/80 (ab53126), MLH1 (ab92312), MSH2 ity maps and raw multi-channel files from each region
(ab52266) and MSH6 (ab14204) and slides were then of interest were analysed using CellProfiler and the
scanned on a Leica slide imaging platform. IHC was Cytomapper R package (v3.12) [27, 28].
scored by the QuPath system [24]. Briefly, a single section
was zoomed to 10 × view of a representative area of epi- Bioinformatics
thelium/tumour, calibration was performed according to For the amplicon resequencing, FASTQ files were
the QuPath manual, either nuclei or membranous count- trimmed (Trimgalore) and aligned to the GRCh38 refer-
ing was set depending on the antibody and auto counting ence genome using a standard pipeline using the BWA
of DAB stained positive/negative cells was carried out. (v0.7.17-r1188) aligner [29]. Mutation calling was per-
Staining was reported as a percentages of positive/nega- formed using FreeBayes (v1.0.0) [30]. For the samples
tive cells.
Stockton et al. Radiat Oncol (2021) 16:129 Page 4 of 12

from the SCORT consortium, alignment using BWA Results

was carried out to the GRCh38 reference genome and Samples
mutation calling performed with Caveman (v1.14.0) In total, there were 48 patients with samples available in
and Pindel (v1.0) [31, 32]. For the exome sequencing the study (Fig. 1). In the pathCR group, 24 pre-treatment
analysis, FASTQ files were trimmed and aligned to the biopsies were available. In the post-treatment group there
GRCh38 reference genome using an exome sequencing were 24 post-treatment samples that had a poor response
pipeline Isaac v4 aligner [33], Manta SV caller (v1.6.0) to chemoradiotherapy as defined by histological tumour
[34] and Canvas CNV (Canvas 1.40.0.1613 + master) regression grade. The S-CORT WS3 cohort had 231 pre-
[35] caller. Enrichment was determined by compari- treatment biopsies available for analysis with variable lev-
son to the Illumina TruSeq exome v1.2 BED file. Muta- els of response. Matched germline DNA was available for
tion calling was performed using Strelka 2 [36] and all samples apart from the S-CORT WS3 cohort.
MuTect2 in tumour-normal subtraction mode. For both
amplicon sequencing and exome analysis, significantly Sequencing metrics
mutated genes were identified using Intogen (v1.0) [37], All available samples successfully underwent ampli-
MutSigCV (v2.1) [38] and dNdScv (v.0.1.0) [39]. Muta- con sequencing and exome sequencing. For the ampli-
tional signatures were estimated using the Mutation- con sequencing average read depth was 1020 × (range
alPatterns (v3.0.1) [40] R package. Tumour mutational 357–5210 ×). For the whole exome sequencing the aver-
burden was estimating by annotation of the combined, age read depth was 99 × for tumour samples and 55 × for
overlapping variant calls from Mutect/Strelka2 to fil- control normal. For methylation arrays all samples
ter to non-synonymous variants (nsV). NsV were then hybridised successfully.
divided by the TruSeq Exome panel size (45.3  Mb) to
give a figure in mutations/mb.
For structural variant (SV) calls, Manta was used in Mutational profiles
tumour/normal mode on exome sequencing data and In the amplicon sequencing group, using treat-
lists of structural variants outputted to VCF file. For ment response as a classifier to select top significantly
copy number calls (CNV), Canvas was used in tumour mutated genes correlated with response as defined
normal mode and calls outputted as VCF files. by MutSigCV,were PIK3CA (p  = 2.4 × ­10−9), FBXW7
Clonality was determined by running superFreq (p = 8.68 × ­10−4) and PTEN (p  = 1.02 × ­10−3). Neither
(v1.0) [41], a cancer specific tumour exome caller and dNdScV nor Intogen can classify mutations by co-var-
river plots were producing from this package. For neo- iates therefore a Fisher exact test was used to compare
antigen calls, tumour-normal subtracted VCF files the two groups (pathCR vs. non-responder) for the out-
produced with Strelka, GT fields were added via con- put of each caller. Using dNdScV, the top significantly
version of the SGT field (using a custom script), anno- mutated genes were AKT1 ­(pmis = 2.01 × ­10−3), FBXW7
tated with the variant effect predictor (VEP), filtered ­(pmis = 5.90 × ­10−3), FAM123B ­(pmis = 2.48 × ­10−2) and
for indels and then analysed using PVacTools v2.0.0 POLE ­(pmis = 1.32 × ­10−2). Gene centric analysis with
against MHC Class I binding predictions [42]. Neoan- Intogen demonstrated recurrent mutations in PIK3CA,
tigens with a “best” median IC of < 50  nmol  ­L−1 were PTEN, FBXW7 and POLE. Pathway analysis of the
counted as binding neoantigens. HLA typing for each genes present in the pathCR group but not the non-
patient was determined with HLA-LA* (v1.0.1) on ger- responder group demonstrated 14 pathways (Table  1)
mline exome sequencing data [43]. being significantly over-represented in the dataset
For methylation, IDAT files were imported into R/Bio- including hsa05210 (Colorectal Cancer, p = 1.88 × ­10−16,
conductor and analysed on the ChAMP (v2.20.1) pipeline q = 1.62 × ­10−14), hsa05222 (Small cell lung cancer,
[44] using standard settings. iDAT files were imported, p = 8.43 × ­10−12, q = 3.63 × ­10−10), and hsa04150 (mTOR
standardised to beta methylation values, then underwent pathway, p = 4.99 × ­10−10, q = 1.43 × ­10−8), all of which
SWAN normalisation. Normalised beta-values under- contain genes involved in the mTOR/AKT pathway.
went differential methylation analysis using limma, DMR For the exome sequencing data, analysis by MutSigCV
were called using dmrLasso. Pathway analyses were con- using pathCR as a covariate revealed 1,412 genes that
ducted in GProfiler2 [45]. were significantly mutated (Additional file  1). The top
For statistical analysis, Stata 15.1 was used, and all dis- genes were HIVEP3, HS6ST3, KIAA1671, LRRC4C and
tributions plotted to check for a Gaussian distribution. ROBO2. This was due to the preponderance of hyper-
Unpaired t-testing was set with a significance threshold mutant samples within the pathway which lead to select
of 0.05, and any missing values were imputed using the of non-driver genes, and filtering would have removed
Stata impute commands. samples. Pathway analysis by GProfiler demonstrated no
 Stockton et al. Radiat Oncol (2021) 16:129 Page 5 of 12

Fully sensive Fully resistant

Rectal cancers n= 48

Pathological complete Poor response

response (n= 24) (n= 24)

Pre-treatment Pre-treatment biopsies and

biopsies only resecon specimens

High read depth Exome sequencing Methylaon arrays

amplicon sequencing

Differences between complete responders and poor

responders are those which make the tumour
sensive

Fig. 1  Flowchart of samples going through study

Table 1  Intogen significantly enriched pathways

ID Pathway Z score p value Q value

hsa05210 Colorectal cancer 8.14 1.88E−16 1.62E−14

hsa05222 Small cell lung cancer 6.73 8.43E−12 3.63E−10
hsa04150 mTOR signalling pathway 6.10 4.99E−10 1.43E−08
hsa04310 Wnt signalling pathway 5.96 1.26E−09 2.71E−08
hsa05212 Pancreatic cancer 5.77 3.91E−09 6.60E−08
hsa00562 Inositol phosphate metabolism 5.71 5.37E−09 6.60E−08
hsa04070 Phosphatidylinositol signalling system 5.71 5.37E−09 6.60E−08
hsa04115 p53 signalling pathway 5.60 1.01E−08 1.09E−07
hsa05166 Human T-cell leukemia virus 1 infection 5.50 1.87E−08 1.79E−07
hsa05169 Epstein-Barr virus infection 4.99 2.96E−07 2.12E−06
hsa05162 Measles 4.99 2.96E−07 2.12E−06
hsa04210 Apoptosis 4.99 2.96E−07 2.12E−06
hsa05217 Basal cell carcinoma 4.66 1.56E−06 1.03E−05
hsa04120 Ubiquitin mediated proteolysis 4.59 2.16E−06 1.33E−05
Stockton et al. Radiat Oncol (2021) 16:129 Page 6 of 12

enrichment of any KEGG pathways. Analysis by Intogen is due to the effects of transcription coupled nucleo-
using revealed demonstrated 1620 genes significantly tide repair and signature 30 is due to a defect in base-
mutated (Additional file 1 as determined either by Onco- excision repair due to inactivating mutations in NTHL1
DriveClust or OncoDriveFM. The top five genes were (Additional file 1).
CDC27, CTBP2, IGSF3, PABPC3 and ZNF432. Path- Tumour mutation burden (TMB) was significantly
way analysis with Intogen demonstrated over-represen- higher in pre-treatment samples from pathCR patients
tation of focal adhesion (hsa04510, p  = 2.25 × ­10−135, than non-responders (exome data, ­medianpathCR = 38.28
−133
q = 5.82 × ­10 ), which contains within it the MAPK/ muts/mb, range 15.93–86.95 v ­medianNR 7.27 muts/mb,
PIK3K and Wnt signalling pathways; as the top rated range 3.08–47.3, p = 0.02 Wilcoxon, Fig. 2).
pathway (Additional file 1). Analysis by dNdScV demon- In order to validate the finding that tumour muta-
strated 209 genes significantly mutated (p < 0.05), the top tional burden seemed to correlate with response, we
five of which were ZNF717, MUC3A, APC, OR4C5 and analysed data from the SCORT WS3 “Grampian”
KRAS. cohort consisting of pre-treatment biopsies (n = 231)
Mutational signature analysis (version 3 single base from patients undergoing long course chemoradio-
substitution) was performed on the exome sequencing therapy. Total numbers of coding mutations were sig-
samples from the pathCR group. The top ranked signa- nificantly (p = 0.036) greater in the “responder” (n = 35,
tures were signatures three, five and thirty. Signature mean 9.2 mutations) vs. “non-responder” (n  = 196,
three is postulated to be due to defective base repair mean 7.8 mutations) groups.
due to faulty homologous recombination, signature five

Fig. 2  Differences in genomic characteristics between responders and non-responders. A Tumour mutational burden in mutations/megabase;
B Structural variants per sample; C Numbers of neoantigens per sample; D Change in Tumour mutational burden in paired samples pre and post
response; E MATH score per sample
 Stockton et al. Radiat Oncol (2021) 16:129 Page 7 of 12

Structural variants and copy number variation an increase in clonal diversity (Fig. 4) in all patients, sug-
Copy number estimation was performed on all tumour: gesting that radiation therapy drove the generation of an
normal exome pairs successfully (Fig.  3). Three biopsy increase in clonal diversity. For both the pathCR group
samples showed an extremely complex pattern of copy and the non-responder group, we calculated MATH
number gain and loss (with modal CN of 67, 52 and 66 score, a measure of tumour heterogeneity for all samples
respectively), which corresponded with complete patho- in the pre-treatment biopsies of all samples, finding that
logical response to chemoradiotherapy. The median copy there was a median MATH score of 4.7 in non-respond-
number of the pathCR group was 50 (IQR 45–66) and of ers and 4.2 in pathCR group (Mann Whitney p = 0.5036),
the non-responder group was 46 (IQR 44–47, Wilcoxon suggesting that tumour heterogeneity did not play a role
Ranked sums p = 0.01). No recurrent copy number alter- in treatment resistance (Fig. 5).
ations were observed. In the poor response group, the
copy number seen in the pre-treatment biopsy (median Methylation analysis
45, IQR 44–49) did not significantly vary with the post In order to understand if there were epigenetic deter-
treatment specimen (median 46, IQR 44–47, Wilcoxon minants of response, we compared the methylome of
Ranked sums p = 0.94). the pre-treatment biopsies of the pathCR group to the
Structural variant (SV) calling was performed on all tumour of the no response group. This demonstrated
tumour: normal exome pairs successfully. The median 1853 differentially methylated positions and Go:Profiler
number of structural variants in the pre-treatment analysis of these differentially methylated genes demon-
biopsies of the pathCR group was 4 (IQR 0–10) vs. 52 strated that the WikiPathways WP306 (Focal adhesion
in non-responders group (IQR 18–80, p < 0.001, Fig.  2). p = 1.16 × ­10−3), WP185 (Integrin mediated focal adhe-
There was no significant difference in the numbers of SV sion p  = 2.42 × ­10−3) and WP3932 (PI3K-Akt-MTOR
between the pre-treatment biopsies of the non-respond- pathway p = 3.31 × ­10−2) were enriched in the pathCR
ers and the post-treatment specimens. group for demethylation, suggesting potential activation
of this pathway. Pathway WP306 contains the PI3K/Akt/
Clonal evolution analysis MTOR signalling genes as well as WP3932 and repre-
In the seven cases from whole trios of normal, pre- sents a general enrichment of this pathway.
treatment biopsy and post-treatment specimens were Analysis of differentially methylated regions (con-
available (only in the poor response group), clonal evo- tiguous blocks of differentially methylated CpGs > 10 in
lution analysis with superFreq was carried out. This number, DMR) 347 differentially methylated regions.
revealed that subsequent to radiation therapy there was HOMER annotation of top 100 DMRs and then analysis

Chromosome
Non-responders
responders
Complete

Key:

Fig. 3  Copy number calls across cohort – ID on Y-axis

Stockton et al. Radiat Oncol (2021) 16:129 Page 8 of 12

(IQR 78–154) and the non-responder group a median 74

neoantigens (IQR 28–89, Wilcoxon rank sum p = 0.034).

DSB pathway and mismatch repair deficiency

Semi-quantitative analysis using QuPath was carried
out on white light DAB stained images of the expression
of yH2AX, Ku70/80, ATM, MLH1, MSH2 and MSH6
respectively. For the pathCR group the pre-treatment
biopsies only were studied and for the non-responder
group the post-treatment specimens.
For DNA repair proteins, pathCR cases showed signifi-
cant differences compared to the non-responder group
in the DNA repair associated proteins γH2AX (89% cells
expressing vs. 78% cells expressing, p = 0.007), ATM (80%
vs. 69%, p = 0.04) and Ku70/80 (88% vs. 29%, p < 0.001),
suggesting that a deficiency of normal DNA double
strand break proteins was associated with resistance to
treatment.
In the mismatch repair associated proteins, pathCR
group showed significant differences compared to the
non-responder group in MLH1 (70% cells expressing
vs. 84% cells expressing, p = 0.001) and MSH6 (76% vs.
88%, p < 0.001). However MSH2 expression (33% vs. 41%,
p = 0.10) was not significant. Especially for the mismatch
repair proteins, despite there being statistically signifi-
cant differences, these were only a few percentage points
different and little complete loss of expression was seen,
suggesting that the role of these proteins is uncertain in
this context.

Imaging mass cytometry analysis

A 40 marker image mass cytometry analysis was per-
formed, and cell counting using CellProfiler determined
that with 44 regions of interest across the responders
vs. non-responders there were significantly more (pro-
portions as part of total immune cells compartment
7.98% vs. 3.06%, p = 0.0021) CD8 + T-cells infiltrating in
Fig. 4  Example river plot of poorly responding cancer; Pre-treatment tumours with a complete response to radiotherapy than
biopsy demonstrates one clone in sample (clonality 0.57); post
those with a poor response.
treatment specimen shows multiple (five more) clones evolving as a
result of radiotherapy selection pressure
Discussion
In this study, we have demonstrated a number of interest-
by g:Profiler revealed only enrichment for methylation in ing findings comparing pre-treatment biopsies and post
the Human Phenotype pathway HP0000104 Renal Agen- treatment specimens in patients undergoing (nCRT) for
esis (p = 2.82 × ­10−2), which contains Wnt signalling and rectal cancer.
DNA repair genes (ATRX) and the GO:MF term Class II The most important finding is that in patients achiev-
MHC binding (p = 1.77 × ­10−2). ing a complete or near complete response to nCRT
is that the pre-treatment tumour has a high tumour
mutational burden [46]. This is because hypermuta-
Neoantigen prediction tion within the tumour due to an intrinsic DNA repair
Neoantigen prediction using pVacTools revealed a defect and leads to increased presentation of neoanti-
median of 78 neoantigens per sample (range 9–383) of gens because of indel and frameshift mutations [47].
which the pathCR group had a median 91 neoantigens This increases immunovisibility and we hypothesise
 Stockton et al. Radiat Oncol (2021) 16:129 Page 9 of 12

Fig. 5  Multi-marker image mass cytometry figures of non-responders (top two images) vs. responders (bottom two images). In left most images,
blue staining represents nuclear beta catenin (representing cellular nuclei) and red staining represents CD8 + T-cells. In right most images, blue
demonstrates tumour, red demonstrates CD8 + T-cells nad grey demonstrates stoma, showing that the T-cells infiltrate predominantly the
stromal compartment. Violin plot to the right shows significant differences between responders and non-responders in terms of percentage of
CD8 + T-cells (percentage shown is relative to total number of cells)

that the increased immunovisibility, coupled with acti- We have also demonstrated that there is enrichment for
vation of the immune system by irradiation, causes mutations in the mTOR/AKT signalling pathways, specif-
cGAS/STING activation which has been shown to lead ically PIK3CA but also as a general trend towards muta-
to a Type I interferon response [48] leading to migra- tions and epigenetic changes in this pathway. PIK3CA is
tion of immune cells and enhanced regression. Also, it a member of the mTOR signalling pathway and makes up
suggests that these patients may benefit from neo-adju- the alpha subunit of the PI3K protein. mTOR signalling
vant anti-PD1/PD-L1 or anti-CTLA4 immunotherapy has previously been highlighted as being of possible rele-
as high TMB (defined as > 10 muts/mb) has been shown vance in radiosensitivity [50] as a cellular marker of stress
to be correlated with responsiveness to these agents [1]. in the tumours of patients undergoing chemoradiother-
Another intriguing possibility is that genotoxic ther- apy. PIK3CA signals through the mTORC1/mTORC2
apy, coupled with radiation therapy could be delivered and exerts its downstream effect on AKT [51]. Targeted
as part of neoadjuvant treatment in the tumour, pos- agents for PIK3CA (apitosilib [52]), mTORC1/2 (visu-
sibly increasing neoantigen burden and making more sertib [53]) and AKT (MK2206 [54]) exist and are at vari-
patients suitable for immunotherapy [49]. Our sample ous stages of clinical development. We suggest that these
size of pathological complete responders is relatively agents may be utilised as part of a neoadjuvant therapy
small, however we deliberately chose tumour regres- strategy in order to increase response rates.
sion where absolutely no cells remained, which is a FBXW7, a gene previously implicated in cell cycle
very rare phenomenon, compared to the phenomenon control by ubiquitination? of cyclin-E1 [55] was also
of “minimal residual disease”, but we believed that this significantly enriched for mutations within it in this
would give a stronger biological signal. We found that study. Zhang et  al. demonstrated that FBXW7 also
this was replicated in a larger cohort (SCORT WS3) had a role in non-homologous end joining [56, 57],
although this cohort only underwent limited ampli- being a binding partner in the complex that repairs
con sequencing which can serve at best as a proxy for damage caused by double strand breaks. Mutations in
tumour mutational burden. FBXW7 may affect its ability to participate in NHEJ
Stockton et al. Radiat Oncol (2021) 16:129 Page 10 of 12

and therefore increase radiosensitivity. Mutational Conclusions

signature analysis of the exome sequencing dataset Therefore, we suggest that based on these findings, a
also showed multiple mutational signatures consist- number of factors contribute to the response to neo-
ent with enrichment in DNA repair, specifically faulty adjuvant chemoradiotherapy: hypermutation leading
homologous recombination (HR) and nucleotide exci- to increased neoantigen presentation; enrichment in
sion repair. The finding that both PIK3CA and FBXW7 defects in the mTOR signalling pathway; hypoxia regu-
mutations are both enriched in pre-treatment biopsies lated by miR-21-5p and an increase in clonal diversity.
and found in post-treatment specimens from cancers Our findings agree with those of Akiyoshi et  al. [1] in
that do not respond to neoadjuvant therapy means that increase in neoantigen diversity correlated with
that they may act as biomarkers for lack of response. response. Kamran [16] however, found no clear molec-
In oesophageal cancer, induction of novel mutations ular defects that predisposed to radioresistance. This
in cancer driver genes in post treatment specimens has could be due to the fact that their analysis was geared
been observed after chemoradiotherapy treatment and towards pathways of treatment resistance rather than
our findings may reflect this. those that lead to radiosensitivity.
Tumour heterogeneity is a significant problem across Our findings suggest a number of new therapeutic
all tumours due to drivers that may cause a differen- avenues for increasing responsiveness to chemoradio-
tial response to therapy because of mutational clonal- therapy in rectal cancer. We plan to further study these
ity. Our results show, unsurprisingly, that treatment of in complex 3D models such as organoids in order to
samples with neoadjuvant chemoradiotherapy causes an understand whether they will increase response rates
increase in clonal diversity, which may be a driver in the when appropriately targeted.
phenomenon of radiation resistance. However, we did
not demonstrate any difference between responders and
Abbreviations
non-responders in terms of MATH score as a measure PathCR: Pathological complete response; DNA: Deoxyribonucleic acid; PI3K:
of tumour heterogeneity, unlike other papers [58]. As we Phosphatidylinositol-3-kinase; AKT: AKR thymoma pathway; mTOR: Mam-
only sampled a single region of tumour, it is also possible malian target of rapamycin; nCRT​: Neoadjuvant chemo-radiotherapy; CCR​:
Complete clinical response; TEMS: Transanal endoscopic microsurgery; TAMIS:
the observed lack of difference in heterogeneity is purely Transanal minimally invasive surgery; DSB: Double strand break; HR: Homolo-
due to chance. gous recombination; NHEJ: Non homologous end joining; TRG​: Tumour
The immunohistochemical analysis of tumour samples regression grade; IHC: Immunohistochemistry; FFPE: Formalin fixed, paraffin
embedded; HLA: Human leucocyte antigen; MHC: Major histocompatibility
showed increased yH2AX and ATM expression was asso- complex; CNV: Copy number variant; SV: Structural variant; MATH: Mutant
ciated with response, as well as increased expression of allele tumour heterogeneity.
the mismatch repair proteins MLH1 & MSH6. We found
this puzzling, as we would have expected the opposite to Supplementary Information
be true, i.e. loss of expression was needed for response, The online version contains supplementary material available at https://​doi.​
especially in ATM and yH2AX. However, we hypothesise org/​10.​1186/​s13014-​021-​01853-y.
that these samples may have had regions of loss of nor-
mal response and these disappeared as a consequence Additional file 1. Supplementary data from mutational analysis (Mut-
SigCV, Intogen, Intogen Pathways and COSMIC mutational signatures).
of response to radiotherapy, or that, is as typical tumour
biopsies usually have low tumour content.
Clearly, the best way to understand these defects Acknowledgements
We acknowledge the support of the Genomics Birmingham Sequencing Facil-
and investigate them further would be to build a cel- ity (of which ADB is academic lead).
lular model of rectal cancer in order to modulate these
pathways in order to measure responsiveness [59]. Cur- Authors’ contributions
Manuscript: ADB, JS, IG, DM; Molecular analysis: JS, LT, CW, JJ, MD, TN, ADB;
rent cell lines have a bias towards their micro environ- Sample acquisition and patient cohort: IG, DM, ADB, S-CORT consortium,
ment and although provide reasonable models of single Statistical analysis: ADB. All authors read and approved the final manuscript.
pathway alterations lack the fidelity to measure therapy
Funding
response when modulated. The ideal model would be This study was supported by grants from the Academy of Medical Sciences
an organoid based rectal cancer therapy model [60] as and the Wellcome Trust (Ref 102732/Z/13/Z). ADB is currently supported by a
this provides both the 3D structure (enabling cell/cell Cancer Research UK Advanced Clinician Scientist award (Ref C31641/A23923).
communication a more representative element of intra- Availability data and materials
tumoural hypoxia) and more accurate response to ther- All data will be uploaded to the SRA upon acceptance of this manuscript.
apy, as well as the ability to evolve and resist therapy and
the ability to co-culture with other cells in the microenvi-
ronment such as T-cells and fibroblasts.
 Stockton et al. Radiat Oncol (2021) 16:129 Page 11 of 12

Declarations and then surgery, in the management of locally advanced rectal cancer. J
Natl Compr Canc Netw. 2014;12(4):513–9.
Ethical approval and consent to participate 13. van der Valk MJM, Hilling DE, Bastiaannet E, Meershoek-Klein Kranen-
NW Research Ethics committee (15/NW/0079). All patients gave informed barg E, Beets GL, Figueiredo NL, et al. Long-term outcomes of clinical
consent to participate. complete responders after neoadjuvant treatment for rectal cancer in the
International Watch & Wait Database (IWWD): an international multicen-
Consent for publication tre registry study. Lancet. 2018;391(10139):2537–45.
Not applicable. 14. Chadi SA, Malcomson L, Ensor J, Riley RD, Vaccaro CA, Rossi GL, et al.
Factors affecting local regrowth after watch and wait for patients with a
Competing interests clinical complete response following chemoradiotherapy in rectal cancer
The authors declare no competing interests. (InterCoRe consortium): an individual participant data meta-analysis.
Lancet Gastroenterol Hepatol. 2018;3(12):825–36.
Author details 15. Rombouts AJM, Al-Najami I, Abbott NL, Appelt A, Baatrup G, Bach S, et al.
1
 Surgical Research Laboratory, Institute of Cancer and Genomic Science, Can we Save the rectum by watchful waiting or TransAnal microsurgery
University of Birmingham, Vincent Drive, Birmingham B15 2TT, UK. 2 University following (chemo) Radiotherapy versus Total mesorectal excision for early
Hospitals Birmingham NHS Foundation Trust, Birmingham, UK. REctal Cancer (STAR-TREC study)? Protocol for a multicentre, randomised
feasibility study. BMJ Open. 2017;7(12):e019474.
Received: 13 March 2021 Accepted: 4 July 2021 16. Kamran SC, Lennerz JK, Margolis CA, Liu D, Reardon B, Wankowicz SA,
et al. Integrative molecular characterization of resistance to neoadjuvant
chemoradiation in rectal cancer. Clin Cancer Res. 2019;25(18):5561–71.
17. Agostini M, Janssen KP, Kim IJ, D’Angelo E, Pizzini S, Zangrando A, et al. An
integrative approach for the identification of prognostic and predictive
References biomarkers in rectal cancer. Oncotarget. 2015;6(32):32561–74.
1. Akiyoshi T, Tanaka N, Kiyotani K, Gotoh O, Yamamoto N, Oba K, et al. 18. Rasheed S, Harris AL, Tekkis PP, Turley H, Silver A, McDonald PJ, et al.
Immunogenomic profiles associated with response to neoadju- Hypoxia-inducible factor-1alpha and -2alpha are expressed in most rectal
vant chemoradiotherapy in patients with rectal cancer. Br J Surg. cancers but only hypoxia-inducible factor-1alpha is associated with
2019;106(10):1381–92. prognosis. Br J Cancer. 2009;100(10):1666–73.
2. Glynne-Jones R, Wyrwicz L, Tiret E, Brown G, Rodel C, Cervantes A, et al. 19. Sak A, Stuschke M. Use of gammaH2AX and other biomarkers of
Rectal cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment double-strand breaks during radiotherapy. Semin Radiat Oncol.
and follow-up. Ann Oncol. 2017;28(suppl_4):iv22–iv40. 2010;20(4):223–31.
3. Prete FP, Pezzolla A, Prete F, Testini M, Marzaioli R, Patriti A, et al. Robotic 20. Beggs AD, Domingo E, McGregor M, Presz M, Johnstone E, Midgley R,
versus laparoscopic minimally invasive surgery for rectal cancer: a et al. Loss of expression of the double strand break repair protein ATM
systematic review and meta-analysis of randomized controlled trials. Ann is associated with worse prognosis in colorectal cancer and loss of Ku70
Surg. 2018;267(6):1034–46. expression is associated with CIN. Oncotarget. 2012;3(11):1348–55.
4. Ngan SY, Burmeister B, Fisher RJ, Solomon M, Goldstein D, Joseph D, 21. Trenner A, Sartori AA. Harnessing DNA double-strand break repair for
et al. Randomized trial of short-course radiotherapy versus long-course cancer treatment. Front Oncol. 2019;9:1388.
chemoradiation comparing rates of local recurrence in patients with T3 22. Lal N, White BS, Goussous G, Pickles O, Mason MJ, Beggs AD, et al. KRAS
rectal cancer: Trans-Tasman Radiation Oncology Group trial 01.04. J Clin mutation and consensus molecular subtypes 2 and 3 are independently
Oncol. 2012;30(31):3827–33. associated with reduced immune infiltration and reactivity in colorectal
5. Bujko K, Wyrwicz L, Rutkowski A, Malinowska M, Pietrzak L, Krynski J, et al. cancer. Clin Cancer Res. 2018;24(1):224–33.
Long-course oxaliplatin-based preoperative chemoradiation versus 5 × 23. Blank A, Roberts DE, 2nd, Dawson H, Zlobec I, Lugli A. Tumor heterogene-
5 Gy and consolidation chemotherapy for cT4 or fixed cT3 rectal cancer: ity in primary colorectal cancer and corresponding metastases. Does the
results of a randomized phase III study. Ann Oncol. 2016;27(5):834–42. apple fall far from the tree? Front Med (Lausanne). 2018;5:234.
6. Dossa F, Chesney TR, Acuna SA, Baxter NN. A watch-and-wait approach 24. Bankhead P, Loughrey MB, Fernandez JA, Dombrowski Y, McArt DG,
for locally advanced rectal cancer after a clinical complete response Dunne PD, et al. QuPath: open source software for digital pathology
following neoadjuvant chemoradiation: a systematic review and meta- image analysis. Sci Rep. 2017;7(1):16878.
analysis. Lancet Gastroenterol Hepatol. 2017;2(7):501–13. 25. Koressaar T, Lepamets M, Kaplinski L, Raime K, Andreson R, Remm M.
7. Habr-Gama A, Sabbaga J, Gama-Rodrigues J, Sao Juliao GP, Proscurshim I, Primer3_masker: integrating masking of template sequence with primer
Bailao Aguilar P, et al. Watch and wait approach following extended neo- design software. Bioinformatics. 2018;34(11):1937–8.
adjuvant chemoradiation for distal rectal cancer: are we getting closer to 26. Ijsselsteijn ME, van der Breggen R, Farina Sarasqueta A, Koning F, de
anal cancer management? Dis Colon Rectum. 2013;56(10):1109–17. Miranda N. A 40-marker panel for high dimensional characterization of
8. Sanghera P, Wong DW, McConkey CC, Geh JI, Hartley A. Chemoradiother- cancer immune microenvironments by imaging mass cytometry. Front
apy for rectal cancer: an updated analysis of factors affecting pathological Immunol. 2019;10:2534.
response. Clin Oncol (R Coll Radiol). 2008;20(2):176–83. 27. Eling N, Damond N, Hoch T, Bodenmiller B. Cytomapper: an R/biocon-
9. Fokas E, Allgauer M, Polat B, Klautke G, Grabenbauer GG, Fietkau R, ductor package for visualisation of highly multiplexed imaging data.
et al. Randomized phase II trial of chemoradiotherapy plus induc- Bioinformatics. 2020.
tion or consolidation chemotherapy as total neoadjuvant therapy 28. Kamentsky L, Jones TR, Fraser A, Bray MA, Logan DJ, Madden KL,
for locally advanced rectal cancer: CAO/ARO/AIO-12. J Clin Oncol. et al. Improved structure, function and compatibility for Cell Profiler:
2019;37(34):3212–22. modular high-throughput image analysis software. Bioinformatics.
10. Pettersson D, Holm T, Iversen H, Blomqvist L, Glimelius B, Martling A. 2011;27(8):1179–80.
Preoperative short-course radiotherapy with delayed surgery in primary 29. Li H, Durbin R. Fast and accurate short read alignment with Burrows-
rectal cancer. Br J Surg. 2012;99(4):577–83. Wheeler transform. Bioinformatics. 2009;25(14):1754–60.
11. Bahadoer RR, Dijkstra EA, van Etten B, Marijnen CAM, Putter H, Kranen- 30. Garrison E, Marth G. Haplotype-based variant detection from short-read
barg EM, et al. Short-course radiotherapy followed by chemotherapy sequencing. arXiv preprint arXiv:​1207.​3907a​rXiv:​1207.​3907. 2012.
before total mesorectal excision (TME) versus preoperative chemoradio- 31. Jones D, Raine KM, Davies H, Tarpey PS, Butler AP, Teague JW, et al.
therapy, TME, and optional adjuvant chemotherapy in locally advanced cgpCaVEManWrapper: simple execution of CaVEMan in order to detect
rectal cancer (RAPIDO): a randomised, open-label, phase 3 trial. Lancet somatic single nucleotide variants in NGS data. Curr Protoc Bioinform.
Oncol. 2021;22(1):29–42. 2016;56:15 0 1- 0 8.
12. Cercek A, Goodman KA, Hajj C, Weisberger E, Segal NH, Reidy-Lagunes 32. Ye K, Schulz MH, Long Q, Apweiler R, Ning Z. Pindel: a pattern
DL, et al. Neoadjuvant chemotherapy first, followed by chemoradiation growth approach to detect break points of large deletions and
Stockton et al. Radiat Oncol (2021) 16:129 Page 12 of 12

medium sized insertions from paired-end short reads. Bioinformatics. interferon-dependent antitumor immunity in immunogenic tumors.
2009;25(21):2865–71. Immunity. 2014;41(5):843–52.
33. Raczy C, Petrovski R, Saunders CT, Chorny I, Kruglyak S, Margulies EH, et al. 49. Swift LH, Golsteyn RM. Genotoxic anti-cancer agents and their relation-
Isaac: ultra-fast whole-genome secondary analysis on Illumina sequenc- ship to DNA damage, mitosis, and checkpoint adaptation in proliferating
ing platforms. Bioinformatics. 2013;29(16):2041–3. cancer cells. Int J Mol Sci. 2014;15(3):3403–31.
34. Chen X, Schulz-Trieglaff O, Shaw R, Barnes B, Schlesinger F, Kallberg M, 50. Chang L, Graham PH, Hao J, Ni J, Bucci J, Cozzi PJ, et al. PI3K/Akt/mTOR
et al. Manta: rapid detection of structural variants and indels for germline pathway inhibitors enhance radiosensitivity in radioresistant prostate
and cancer sequencing applications. Bioinformatics. 2016;32(8):1220–2. cancer cells through inducing apoptosis, reducing autophagy, suppress-
35. Roller E, Ivakhno S, Lee S, Royce T, Tanner S. Canvas: versatile and scalable ing NHEJ and HR repair pathways. Cell Death Dis. 2014;5:e1437.
detection of copy number variants. Bioinformatics. 2016;32(15):2375–7. 51. Kennedy AL, Morton JP, Manoharan I, Nelson DM, Jamieson NB,
36. Saunders CT, Wong WS, Swamy S, Becq J, Murray LJ, Cheetham RK. Pawlikowski JS, et al. Activation of the PIK3CA/AKT pathway suppresses
Strelka: accurate somatic small-variant calling from sequenced tumor- senescence induced by an activated RAS oncogene to promote tumori-
normal sample pairs. Bioinformatics. 2012;28(14):1811–7. genesis. Mol Cell. 2011;42(1):36–49.
37. Gonzalez-Perez A, Perez-Llamas C, Deu-Pons J, Tamborero D, Schroeder 52. Dolly SO, Wagner AJ, Bendell JC, Kindler HL, Krug LM, Seiwert TY, et al.
MP, Jene-Sanz A, et al. IntOGen-mutations identifies cancer drivers across Phase I study of apitolisib (GDC-0980), dual phosphatidylinositol-3-kinase
tumor types. Nat Methods. 2013;10(11):1081–2. and mammalian target of rapamycin kinase inhibitor, in patients with
38. Lawrence MS, Stojanov P, Polak P, Kryukov GV, Cibulskis K, Sivachenko A, advanced solid tumors. Clin Cancer Res. 2016;22(12):2874–84.
et al. Mutational heterogeneity in cancer and the search for new cancer- 53. Basu B, Krebs MG, Sundar R, Wilson RH, Spicer J, Jones R, et al. Vistusertib
associated genes. Nature. 2013;499(7457):214–8. (dual m-TORC1/2 inhibitor) in combination with paclitaxel in patients
39. Martincorena I, Raine KM, Gerstung M, Dawson KJ, Haase K, Van Loo P, with high-grade serous ovarian and squamous non-small-cell lung
et al. Universal patterns of selection in cancer and somatic tissues. Cell. cancer. Ann Oncol. 2018;29(9):1918–25.
2018;173(7):1823. 54. Narayan RS, Fedrigo CA, Brands E, Dik R, Stalpers LJ, Baumert BG, et al.
40. Blokzijl F, Janssen R, van Boxtel R, Cuppen E. MutationalPatterns: compre- The allosteric AKT inhibitor MK2206 shows a synergistic interaction with
hensive genome-wide analysis of mutational processes. Genome Med. chemotherapy and radiotherapy in glioblastoma spheroid cultures. BMC
2018;10(1):33. Cancer. 2017;17(1):204.
41. Flensburg C, Sargeant T, Oshlack A, Majewski I. SuperFreq: integrated 55. Yeh CH, Bellon M, Nicot C. FBXW7: a critical tumor suppressor of human
mutation detection and clonal tracking in cancer. BioRxiv. 2018:380097. cancers. Mol Cancer. 2018;17(1):115.
42. Hundal J, Carreno BM, Petti AA, Linette GP, Griffith OL, Mardis ER, et al. 56. Zhang Q, Karnak D, Tan M, Lawrence TS, Morgan MA, Sun Y. FBXW7
pVAC-Seq: a genome-guided in silico approach to identifying tumor facilitates nonhomologous end-joining via K63-linked polyubiquitylation
neoantigens. Genome Med. 2016;8(1):11. of XRCC4. Mol Cell. 2016;61(3):419–33.
43. Dilthey AT, Mentzer AJ, Carapito R, Cutland C, Cereb N, Madhi SA, et al. 57. Zhang Q, Mady ASA, Ma Y, Ryan C, Lawrence TS, Nikolovska-Coleska Z,
HLA*LA - HLA typing from linearly projected graph alignments. Bioinfor- et al. The WD40 domain of FBXW7 is a poly(ADP-ribose)-binding domain
matics. 2019. that mediates the early DNA damage response. Nucleic Acids Res.
44. Tian Y, Morris TJ, Webster AP, Yang Z, Beck S, Feber A, et al. ChAMP: 2019;47(8):4039–53.
updated methylation analysis pipeline for Illumina BeadChips. Bioinfor- 58. Noorbakhsh J, Kim H, Namburi S, Chuang JH. Distribution-based meas-
matics. 2017;33(24):3982–4. ures of tumor heterogeneity are sensitive to mutation calling and lack
45. Raudvere U, Kolberg L, Kuzmin I, Arak T, Adler P, Peterson H, et al. strong clinical predictive power. Sci Rep. 2018;8(1):11445.
g:Profiler: a web server for functional enrichment analysis and conver- 59. Ganesh K, Wu C, O’Rourke KP, Szeglin BC, Zheng Y, Sauve CG, et al. A rectal
sions of gene lists (2019 update). Nucleic Acids Res. 2019;47(W1):W191–8. cancer organoid platform to study individual responses to chemoradia-
46. Rospo G, Lorenzato A, Amirouchene-Angelozzi N, Magri A, Cancelliere C, tion. Nat Med. 2019;25(10):1607–14.
Corti G, et al. Evolving neoantigen profiles in colorectal cancers with DNA 60. Nagle PW, Coppes RP. Current and Future perspectives of the use of
repair defects. Genome Med. 2019;11(1):42. organoids in radiobiology. Cells. 2020;9(12).
47. Goodman AM, Kato S, Bazhenova L, Patel SP, Frampton GM, Miller
V, et al. Tumor mutational burden as an independent predictor of
response to immunotherapy in diverse cancers. Mol Cancer Ther. Publisher’s Note
2017;16(11):2598–608. Springer Nature remains neutral with regard to jurisdictional claims in pub-
48. Deng L, Liang H, Xu M, Yang X, Burnette B, Arina A, et al. STING- lished maps and institutional affiliations.
dependent cytosolic DNA sensing promotes radiation-induced type I

Ready to submit your research ? Choose BMC and benefit from:

• fast, convenient online submission

• thorough peer review by experienced researchers in your field
• rapid publication on acceptance
• support for research data, including large and complex data types
• gold Open Access which fosters wider collaboration and increased citations
• maximum visibility for your research: over 100M website views per year

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions