The Role of Nuclear APOBEC Enzymes in Neoplastic Progression of Ulcerative Colitis

Introduction

The activity of nuclear APOBEC family deaminase enzymes, eg AID and APOBEC3B, is normally associated with immune response processes. However, in recent years, aberrant expression of APOBEC enzymes and consequently off-target deamination of cytidines to uracil in DNA have been suggested to contribute to genome instability, and even drive development of cancer.^1–3 In particular, aberrant expression of AID has been proposed to be a link between inflammation in ulcerative colitis (UC) and progression to colitis-associated colorectal cancer (CA-CRC).⁴ It is known that the NF-κB pathway, which is active in inflamed mucosa of UC patients, may induce aberrant expression of APOBEC enzymes.^5,6 Consequently, these enzymes may be a potential source of DNA damage and genome instability in the colonic enterocytes in addition to oxidative damage induced by reactive oxygen and nitrogen species (RONS) in UC genomes.^7,8 Here, we evaluated the expression of AID and other nuclear APOBEC family members and the potential genomic footprint of their activity in colorectal mucosa in a UC patient, who had progressed to CA-CRC (UC progressor). Moreover, we addressed the activity of the classical DNA damage response (DDR) pathway and the presence of genome instability in this patient (CA-CRC case). Then, we validated our findings in a cohort of seven UC progressors and 30 UC non-progressors.

Materials and Methods

Patient Material

We explored a CA-CRC case; a female patient diagnosed with UC pancolitis without concomitant primary sclerosing cholangitis (PSC) at the age of 49 years and treated with Salazopyrin. She has had numerous episodes of self-limiting flares without consulting specialist healthcare and did not have other diseases. Twenty years after diagnosis, at 69 years of age, the patient progressed to adenocarcinoma pT3N2b DUKES C in the sigmoid colon and concomitant low- and high-grade dysplasia in the colorectum. The patient underwent a proctocolectomy from which tissue specimens covering the entire colon and rectum were formalin-fixed and paraffin-embedded (FFPE) or embedded in Optimal Cutting Temperature (OCT) compound (64 FFPE specimens and 14 OCT compound embedded specimens each paired with an adjacent FFPE specimen). The specimens represented different stages of histological pathology; non-inflamed (ascending colon), chronic-active inflammation (hepatic flexure), low-grade dysplasia (LGD) (cecum), high-grade dysplasia (HGD) (cecum and rectum) and colorectal cancer (sigmoid colon). Whole blood was also sampled from the patient. In addition to material from the CA-CRC case, we used FFPE mucosa biopsies from a cohort of non-PSC UC progressors (7 patients) and non-progressors (30 patients) described elsewhere.⁹ The biopsies covered non-inflamed (73 biopsies from non-progressors, 21 biopsies from progressors), inflamed (22 biopsies from non-progressors, 4 biopsies from progressors), LGD (8 biopsies from progressors), HGD (1 biopsy from progressor) and cancer (1 biopsy from progressor) histology. The study was conducted following the Declaration of Helsinki (as revised in 2013).

The material used in this study has been approved by the Regional Committees for Medical and Health Research Ethics in Norway (2010/1093, 2014/722 and 2015/664). Written informed consent was obtained from the study participants, including the use and publication of detailed clinical information such as disease activity, course of disease and treatment.

DNA and RNA Isolation

DNA and RNA were isolated from four sets of adjacent FFPE and frozen sections, respectively, of colorectal specimens covering non-inflamed, inflamed, HGD and cancer from the CA-CRC case. For every 50 µm sectioned of each FFPE and OCT block, a control HE slide was evaluated for confirmation of the histology and ensuring that the inflamed, HGD and cancer tissue constituted more than 50% of the respective samples used for isolation. DNA was extracted using the Blood & Culture DNA mini kit (Qiagen), and RNA was extracted using the RNeasy micro kit (Qiagen) following the protocol with the DNase treatment included (DNase kit from Qiagen). DNA from whole blood was extracted using the Blood & Culture DNA midi kit (Qiagen). DNA and RNA concentrations were measured on a Nanodrop (Thermo Scientific). The DNA and RNA integrity were assessed using the DNA 12000 Reagent kit (Agilent) and the RNA 6000 Nano kit (Agilent), respectively, and run on a Bioanalyzer 2100 (Agilent).

qRT-PCR and Telomere Length

Reverse transcription of total RNA and quantitative RT-PCR analysis were run as described previously¹⁰ using primers specific for NFKBIA,¹¹ AICDA,¹² APOBEC1,¹² APOBEC3A,¹² APOBEC3B,¹² APOBEC3G¹² (Invitrogen) and TBP (Qiagen). For AICDA, ten cycles of pre-amplification were performed, using Taqman probes and TaqMan PreAmp Master Mix (Life Technology) and then run as above using the Taqman primers and TaqMan^® Gene Expression Master Mix (Life Technologies). Average telomere length was measured from genomic DNA as described previously.¹⁰ All assays were run in triplicates.

Whole Genome Sequencing and Mutation Signatures

DNA from whole blood and colorectal specimens covering histological inflamed, HGD and cancer was subjected to whole genome sequencing (WGS) using an Illumina HiSeq X-Ten platform with a depth of 30x for the blood-derived DNA and 80x for the colorectal specimen-derived DNA (BGI Genomics). Downstream analysis involved GATK best practices for read alignment and processing to generate BAM files.¹³ The analysis was performed using two variant callers; Strelka2 and Mutect2, which produced VCF files listing the detected variants.^14,15 For each sample, variants from Strelka and Mutect were merged. The blood-derived sample served as the germline reference, allowing subtraction of inherited mutations from the variant sets of the inflammation-, HGD-, and cancer-derived samples.

Mutation signature decomposition was performed using the SigProfilerAssignment refitting algorithm with the COSMIC mutational signatures database (v3.3) against the GRCh38 human reference genome assembly.¹⁶

Immunohistochemistry

Immunohistochemistry (IHC) of FFPE colorectal sections covering non-inflamed, inflamed, dysplastic and cancer histology from the CA-CRC case and from the progressor and non-progressor cohort was performed as described previously.⁹ To confirm the histology throughout a sample, a new HE section was evaluated for every 20 µm. FFPE colorectal sections of 4 μm were pretreated and stained with primary antibody according to the details given in Table 1. After incubation with the primary antibody, slides were incubated with Ready-to-use secondary antibody labelled with HRP (Agilent) for 20 minutes (min) at room temperature (RT). The reaction product was developed applying the Flex DAB+ Sub-Chromo (Agilent) incubating at RT for 10 min. Subsequently, the slides were counterstained in FLEX hematoxylin (Agilent) for 5 min. Between incubations the slides were rinsed/washed using 1x FLEX WASH BUFFER (Agilent). The percentage of positive enterocytes of the total number of enterocytes was scored in areas of relevant pathology for each slide by two pathologists independently.

Table 1 Antibodies and Staining Details Used for Immunohistochemistry

Statistical Methods

Statistical analyses were performed using GraphPad Prism 10 for Windows. One-way ANOVA followed by Dunnett’s multiple comparison test was performed for comparison of samples. An adjusted (adj) p-value < 0.05 was considered statistically significant.

Results and Discussion

Characterisation of a CA-CRC Case

The NF-κB pathway is central to the inflammation process in UC and may induce expression of APOBEC family proteins.^5–7 To examine the activity of the NF-κB pathway in the mucosa in the CA-CRC case, we performed qRT-PCR on RNA targeting NFKBIA mRNA, which encodes the NF-κB inhibitory subunit IκB-α and may serve as a surrogate marker for NF-kB activity.¹¹ Indeed, the NFKBIA mRNA was expressed at a higher level in the inflamed compared to the non-inflamed mucosa (adj p < 0.001), and at lower level in HGD and cancer mucosa (adj p < 0.0001) (Figure 1a). As NF-κB might induce expression of nuclear APOBEC proteins in the inflamed mucosa, we assessed the expression of nuclear APOBEC genes, AICDA (encoding AID), APOBEC1, APOBEC3A, APOBEC3B, and APOBEC3G by qRT-PCR (Figure 1b). AICDA and APOBEC1 mRNAs were expressed at significantly higher levels in the inflamed and HGD mucosa than in the non-inflamed mucosa (AICDA: adj p < 0.0001, APOBEC1: adj p < 0.0001). The proctocolectomy specimens used for RNA isolation comprise not only enterocytes but also other cell types such as immune cells infiltrating the mucosa. Consequently, the observed gene expression of AICDA and APOBEC1 mRNAs may have originated from other cell types. Therefore, we assessed the protein expression in the mucosa by IHC, which allows distinction between enterocytes and immune cells in the mucosa. Staining of FFPE sections covering the entire colorectum of the CA-CRC case with antibodies specific for AID, APOBEC1 and APOBEC3A/B/G showed that a low percentage of enterocytes expressed AID, and this was only seen in a few of the specimens (Figure 1c). Considering the qRT-PCR results, this shows that AID primarily is expressed in other mucosal cells than the enterocytes at the time of sampling. Indeed, the IHC staining of AID confirmed that mucosa infiltrating immune cells were positive for AID expression (data not shown). Nuclear APOBEC3s were absent in the enterocytes in all specimens which corresponds with the identified APOBEC3A/B/G mRNA levels (Figure 1c). Most enterocytes expressed APOBEC1 in all specimens examined (Figure 1c). This was an unexpected finding as APOBEC1 expression in humans seems to have a confined expression to the small intestine of the gastrointestinal tract.¹⁷ We therefore examined histologically healthy colon mucosa by IHC and found expression of APOBEC1 in the enterocytes throughout the entire colon (data not shown). Thus, APOBEC1 expression in enterocytes may be ascribed to normal physiology. Moreover, the higher mRNA levels of APOBEC1 in the inflamed and HGD samples may be ascribed to immune cells in the mucosa of the specimens.

Figure 1 Characterisation of colorectal specimens from a colitis-associated colorectal cancer (CA-CRC) case to assess nuclear APOBEC gene and protein expression and status of the classical DNA damage response (DDR). (a) Bar chart showing the gene expression of NFKBIA determined by qRT-PCR on RNA from single colorectal specimens from the CA-CRC case. (b) qRT-PCR on RNA from single colorectal specimens from the CA-CRC case to determine the gene expression of nuclear APOBEC genes. (c) Heat map showing percentage enterocytes in FFPE colorectal specimens positive for the given marker found by immunohistochemistry. The histology of the specimens is given on the left side of the heatmap and a tick mark corresponds to one specimen. The target marker is given below the heatmap. The colour scale is given on the right side of the heatmap. Specimens appearing with a grey colour indication could not be evaluated. d) Bar chart showing the telomere length of DNA from single colorectal specimens from the CA-CRC case determined by qPCR. In (a, b and d) the samples are normalised to the non-inflamed sample, and the average of three technical replicates are shown. One-way ANOVA followed by Dunnett’s multiple comparison test was performed for comparison of samples in (a, b and d). An adjusted p-value < 0.05 was considered statistically significant. Abbreviations: FC, fold change; infl, inflamed; LGD, low-grade dysplasia; HGD, high-grade dysplasia.

As APOBEC enzyme activity may lead to genome instability, we examined the status of the classical DDR pathway in the patient’s colorectum by IHC (Figure 1c). Independent of histology, most of the specimens had varying degrees of enterocytes with expression of the DNA damage marker 53BP1. However, specimens with dysplasia or cancer had a high percentage of enterocytes with 53BP1 expression. The expression of the DNA damage checkpoint proteins, pCHK1 and pCHK2, were absent in all specimens, whereas the tumour suppressor protein, p53, was primarily expressed in the LGD specimens. However, we cannot conclude whether p53 was activated in these specimens as we were not successful performing IHC targeting the phosphorylated p53. Enterocytes positive for the proliferation marker, Ki67, were found to varying degrees in specimens independent of histology. Overall, this suggests that despite the presence of 53BP1 signal in more than 50% of cells in some specimens, the classical DDR to double strand breaks was not activated in the mucosa in this CA-CRC patient.

Production of RONS as a consequence of the inflammation process in UC involves formation of a variety of DNA lesions and may cause genome instability.^18,19 Telomeres are particularly sensitive to RONS-induced DNA damage, thus we assessed telomere length as a surrogate marker for RONS-induced DNA damage.¹⁹ Significantly shorter telomeres were found in HGD (adj p = 0.0018), and cancer (adj p < 0.0001) mucosa compared to non-inflamed mucosa in the CA-CRC case (Figure 1d).

Finally, it has been suggested that transient bursts of APOBEC activity might drive carcinogenic progression.²⁰ To assess whether the CA-CRC case had any genomic footprint of APOBEC activity, we performed WGS of DNA isolated from an inflamed, a HGD and a cancer specimen from the CA-CRC case. Interestingly, the APOBEC3-associated COSMIC SBS2 mutation signature was found in the genome of all three specimens (data not shown). Moreover, the AID-associated mutation signature SBS85 was found in the three genomes (data not shown). However, as AID and nuclear APOBEC3 proteins were not expressed in the enterocytes, we cannot determine whether these signatures originate from former transient bursts of aberrant APOBEC activity in enterocytes or if they originate from APOBEC activity in eg immune cells present in the specimens. Consequently, we cannot exclude the involvement of APOBECs in the progression of UC to CA-CRC in this patient. In agreement with proposed RONS-driven genome instability from the telomere length analysis in the inflamed, HGD and cancer specimens of this patient, we identified the oxidation-associated mutation signatures SBS18 (in HGD and cancer genomes) and SBS29 (in all three genomes) (data not shown).

AID and Nuclear APOBECs and Genome Instability in a Cohort of UC Progressors and Non-Progressors

Next, we examined if the expression pattern of AID and other nuclear APOBEC proteins and the lack of classical DDR activation found in the CA-CRC case were specific for UC progression. We therefore performed IHC on FFPE biopsies representing non-inflamed, inflamed, dysplastic and cancer colorectal mucosa from a cohort of UC progressors and non-progressors (Figure 2). Like the CA-CRC case, expression of the AID protein was almost absent in the enterocytes independent of histology and progression status. The same applied for the nuclear APOBEC3 proteins except for the cancer biopsy, where 30% of enterocytes expressed the proteins. APOBEC1 expression was seen in a high proportion of enterocytes in most biopsies independent of histology and whether the biopsy originated from a progressor or a non-progressor.

Figure 2 Status of nuclear APOBEC protein expression and classical DNA damage response (DDR) in a cohort of UC progressors and non-progressors. Heat map showing percentage enterocytes in FFPE colorectal biopsies positive for the given marker found by immunohistochemistry in a cohort of UC progressors and non-progressors. The histology of the biopsies is given on the left side of the heatmap and a tick mark corresponds to one biopsy. The target marker is given below the heatmap. The colour scale is given on the right side of the heatmap. Biopsies appearing with a grey colour indication could not be evaluated. Abbreviations: P, progressor; NP, non-progressor; infl, inflamed; LGD, low-grade dysplasia; HGD, high-grade dysplasia.

In contrast to the CA-CRC case, a few of the progressors and non-progressors had enterocytes with activated CHK2 protein (pCHK2). However, it was not necessarily associated with expression of 53BP1, nor was it associated with a specific histology or dependent on progression status. Independent of progression status, p53 expression was seen in many of the non-inflamed and inflamed biopsies in varying percentages of enterocytes expressing p53 and this did not correlate with 53BP1 or pCHK2 staining. Most biopsies with dysplasia had p53 expression, whereas the cancer biopsy was negative for p53 expression. Again, there was no association with 53BP1 or pCHK2 staining or progression status. Ki67 positive enterocytes were found in varying percentages in biopsies independent of histology and progression status. Overall, this showed that presence of genome instability in the enterocytes did not necessarily induce the classical DDR pathway neither in progressors, nor in non-progressors, suggesting that the findings from the CA-CRC case were not specific to UC-progression.

Collectively, in contrast with what was reported by Endo et al⁴ previously, our findings from the CA-CRC case and the progressor and non-progressor cohort do not support a role for AID in driving progression of UC to CA-CRC. Nor did we find support for a role for other nuclear APOBECs. This is supported by a gene expression analysis performed in our lab of the inflamed colorectal mucosa from the progressor and non-progressor cohort.⁹ However, the patients included in the present study had a minimum duration of disease of eight years (median 16.1 years (8–49) and 24 years (8–53) for non-progressors and progressors, respectively).⁹ If transient bursts of AID/APOBEC expression are occurring earlier than eight years after diagnosis, we would not have identified it by IHC in the present study. In contrast to our study, a study by Gushima et al found AID expression by IHC in inflamed as well as neoplastic mucosa of UC patients.²¹ Interestingly, the patients included in that study had a shorter disease duration for both non-progressors and progressors (median 8.5 years (0.5–32) and 15 years (0.5–47), respectively, only half a year in some patients). This might explain the apparent discrepancy between the findings of Gushima et al and our study. Duration of disease was not reported by Endo et al for their patient cohort.⁴ On the other hand, a gene expression study on colonic biopsies from newly diagnosed and treatment naïve ulcerative colitis patients did not suggest AID or APOBEC enzymes to be important as early as diagnosis of disease.²² Thus, we cannot exclude the possibility that AID and nuclear APOBEC3 enzymes may be expressed in enterocytes of the colon. It is possible that bursts of DNA cytidine deamination contribute to genome instability, in addition to RONS-induces DNA damage, early in the disease course of UC. Notably, independent of progression status, DNA damage in enterocytes of colorectal UC mucosa may not necessarily initiate the classical DDR pathway.

The cross-sectional design of the study and the limited number of patients and neoplastic samples used to validate the results of the initial in-depth analysis reduce the generalisability of our findings. However, the stringent selection of patients with homogenous phenotypes (non-PSC UC) strengthens the validity of our findings.

Conclusion

In conclusion, our findings suggest that AID or other nuclear APOBECs are unlikely to drive the progression of UC to CA-CRC late during course of UC. However, we see footprints of AID and APOBEC3 activity in the genome of the inflamed and neoplastic colorectal mucosa in the CA-CRC case and cannot exclude the possibility that AID and nuclear APOBEC3 enzymes may contribute to genome instability early during course of disease.