Reduced Cancer Incidence in Huntington’s Disease: Analysis in the Registry Study

INTRODUCTION

Huntington’s disease (HD) is an inherited neurodegeneration of mid-life onset. It is caused by an expanded CAG repeat at the 5’ end of the HTT gene, which is translated to give a polyglutamine tract at the N-terminus of the encoded protein, huntingtin (HTT) [1]. It is characterised by a movement disorder, cognitive decline and variable psychiatric symptoms with early and continuing cell death in the striatum [1, 2]. The disease is inexorably progressive and there is no treatment that can prevent the neurodegeneration [2].

Studies in the systematic population registries of the Scandinavian countries show a reduction in the expected age-matched incidence of cancer in subjects with a diagnosis of HD and spinal andbulbar muscular atrophy (SBMA, Kennedy’s disease) another disease caused by an expanded CAG tract [3, 4]. Scandinavian countries have country-wide registers that record whole population health so mining these data can be potentially very productive. Most countries do not have such comprehensive and well maintained registers and thus it is hard to replicate such studies. However, the Oxford Record Linkage Group examined rates of cancers in an all-England population study, the English linked hospital episode statistics (LHES), 1999–2010, and showed a similar effect [5]. They examined HD, SBMA, and a larger group of hereditary ataxias (HA), which include the CAG repeat expansion associated spino-cerebellar ataxias (SCAs). The rate ratio cancer diagnosis in 4865 people with HD was 0.53, with no effect seen in subjects with a diagnosis of SBMA or HA. More recently Coarelli et al. [6] questioned a cohort of HD and SCA patients directly about their experience of cancer and found standardised incidence ratios of 0.21 in HD and 0.23 in SCAs.

A reduction in cancer incidence is seen in other neurodegenerative conditions. A systematic examination of records from over 500,000 subjects in observational studies of CNS disorders showed a robust lower cancer co-occurrence in all neurodegenerations, with strikingly significantly lower rates in Alzheimer’s disease (ES 0.32) [7]. Their HD data was derived from Ji and Sorensen [3, 4]. By contrast Freedman et al. [8] observed only a modestly reduced risk of Alzheimer’s disease in cancer survivors in a prospective analysis of 1.16 million subjects in a Medicare population (HR = 0.87 (95% CI = 0.84–0.90). To follow up these studies we examined the incidence of cancers in the EHDN REGISTRY study directly. We obtained the data from REGISTRY and the population rates from the Globocan study [9]. We also examined the effect of the CAG length and the age of onset of HD on cancer incidence.

MATERIALS AND METHODS

RESULTS

There were 6540 patients identified from the EHDN REGISTRY study [11] with appropriate data: details are given in Table 1. Eight subjects were excluded as they were collected outside Europe (Singapore) and four had incomplete records. The 6528 subjects remaining were collected in European clinics, had a clinical diagnosis of HD and 173 (2.65%) of these patients also had information consistent with a cancer diagnosis. In the whole cohort 52.6% of subjects are female and 47.4% male. There is a higher proportion of females amongst those with a diagnosis of cancer than in those without (Table 1): this is attributable to the relative youth of our HD sample as the incidence of breast cancer is higher at relatively young ages compared with many other cancers. The sample size is too small to split by specific cancers to explore this further. The average age of the non-cancer HD subjects at the point of data collection (age at last visit) is over 10 years younger than in those with a cancer diagnosis and the mean CAG repeat length is longer in the non-cancer cohort than in the subjects with cancer (Table 1).

The distribution of the cancer diagnoses in the 173 subjects is given in Table 2. The standardised incidence rate (SIR) of cancers in the HD population was calculated (Table 2) and for all cancers was 0.26 (CI 0.22–0.30). All cancers were observed at significantly lower levels than in the European population though there were differences in the rates between the types of cancer. Uterine and skin cancers had age-standardised incidence rates closest to the European levels with colorectal, breast and prostate cancer all recorded at less than half the rates in the European population.

Those with a cancer diagnosis have, on average, over 10 years later age of onset of HD than those without. Testing whether this is significantly different, is, as noted above, complicated by the fact that cancer incidence rises as people age, and those with later ages of onset of HD will, on average, live longer and thus be more likely to develop a cancer [12]. Modelling this using a cancer diagnosis at any time in the past as a time-dependent covariate in a Cox proportional hazards model [10] of time to HD onset, shows that those with a cancer diagnosis at any age are very slightly more likely to have HD onset than those without cancer at the same age (HR 1.2773; 95% CI 1.06 to 1.61, p < 0.001). A cancer diagnosis during the past year is strongly associated with an increased rate of HD onset (HR 18.95, 95% CI 13.01 to 23.20, p < 0.001).

As we noted that the subjects with cancer diagnoses have shorter CAG repeat lengths than those without (Table 1), we investigated whether this is significantly associated with the likelihood of developing cancer. Given that the CAG repeat showed the expected significant inverse correlation with age of HD onset (p < 2×10^–16), the apparent association of cancer with CAG length could be a result of people with shorter CAG repeats on average living longer (as age at death in HD is inversely correlated with expanded HTT allele CAG repeat length), and thus having a greater risk of developing cancer [12]. To test whether this bias in lifetime length accounts for the apparent association between CAG length and cancer diagnosis we used a Cox proportional hazards model [10] with time from birth to cancer diagnosis as an outcome (death and HD diagnosis both as censoring) and CAG length as a covariate. We also adjusted for age and sex and used different baseline hazards for the different countries. There are no discernible effects of CAG repeat length on cancer incidence.

DISCUSSION

The standardised incidence ratio for all cancers detected in the REGISTRY subjects is substantially lower than that in the non-HD population, as reported previously [3–6]. Cancer might be underdiagnosed in the HD population in later stages of the illness as potentially relevant signs or symptoms may not be noticed or acted upon, or may be masked by HD symptoms – cachexia for instance, common in cancers, is also common in late stage HD. Our findings in the REGISTRY study show a lower rate of cancers than that reported in previous studies of cancer in HD subjects in population cohorts [3–5] which could indicate under-ascertainment of cancer in our sample. However, the recent study examining French HD and SCA populations and asking directly about cancer show a similar SIR to that which we show in this study [6]. However, it is also likely that having a diagnosis of HD, and therefore coming to clinical attention, makes diagnosis of any comorbidities more likely. Turner et al. [5], studying hospital admission records in England, found that there was an increased rate of cancer diagnoses in the first year after admission for HD. The overall decrease in the rate of cancers that they observed among HD patients (rate ratio = 0.71) was made more extreme if the first year was excluded (rate ratio = 0.53). This indicates that under-diagnosis is less likely in this population, rather than more likely: we showed a similar effect in the REGISTRY subjects. The only cancer they found to be as common as in the general population was lung cancer, which Turner et al. [5] attributed to the higher rate of smoking in the HD subjects [13]. We had only four cases of lung cancer in our study, too few to study separately.

The observation that two subjects were taking tamoxifen but had no recorded cancer diagnosis indicates that one of the reasons for the lower rate of all cancers observed here might well be poor recording of comorbidities in the REGISTRY database. This could result from poor recall of the participant, lack of knowledge of their partners and carers in clinic or from subjects dropping out of the study after a diagnosis of cancer due to treatment or other effects of the cancer itself and issues in the systematic collection of medical history in clinic.

In an attempt to overcome the limitations of these data we investigated the effect of CAG repeat length on the time to a cancer diagnosis with the view that any correlation might implicate the CAG length at the HTT locus in promoting or delaying cancer. No such effect was observed. To further investigate any link between the two diseases we also examined whether having a cancer diagnosis was associated with age at onset of HD. This is more difficult as there are competing risks: cancer risk is age-related, and a cancer diagnosis may influence time to death as well as time to HD onset, as may HTT CAG length. However, we observe that a diagnosis of cancer at any time slightly increases the likelihood of HD onset: this marginal effect requires replication in further studies. In addition, the study of Turner et al. [5] found that cancer incidences were higher in the year around HD onset and we see a similar effect in the REGISTRY data. This latter is likely to be an ascertainment bias: subjects receiving clinical attention are more likely to have any comorbid condition detected and therefore under-diagnosis of cancer is less likely in this population.

What might underlie the later onset of cancer in HD subjects? Defects in the DNA damage response (DDR) cause cancers [14] and the DDR has recently been implicated in altering the age at motor onset of HD and other repeat disorders [15, 16]. The direction of this effect is unclear and alterations in the operation of the DDR have been shown to be both protective and deleterious in HD and other neurodegenerations. Mismatch repair and base-excision repair have both been implicated as promoting degeneration in the repeat disorders, possibly through somatic expansion of repeats [17, 18]. Many other neurological diseases are caused by genetic defects in the DDR [19] but conversely, multiple protective effects of the DDR in neurodegenerations have also been observed [20–27].

One consequence of the involvement of the DDR in modifying HD onset might be that although age at onset is later in subjects who have had a cancer diagnosis, if they have cancer in the presence of lower DDR activity, then that cancer might be more aggressive with a faster course. HTT has been implicated in acceleration of breast cancer development and metastasis in mouse models of HD and to regulate cell division in mammary stem cells [28]. In subjects with breast cancer a reduction of ovarian cancer was shown in BRCA2 mutation carriers also carrying longer HTT CAGs [29], along with a paradoxical finding of increased metastasis and younger ages of cancer onset. Direct examination of metastatic breast cancer showed HTT mRNA downregulation in primary tumours and that the expression and localisation of the tight junction protein ZO1 was controlled by HTT [30]. Lower expression of HTT was correlated with less HTT and ZO1 proteins at tight junctions, poorer differentiation of tumour cells and was predictive of worse cancer prognosis [31].

It is not clear how the effect of increased CAG repeats, which translate to expanded polyglutamine tracts in the cognate proteins, could potentially mediate cancer risk but the underlying biology of HD offers clues. Huntingtin (HTT) is expressed in all cells: cell death is promoted by mutant HTT [32–34] and non-mutant HTT is anti-apoptotic [35]. HTT localises to spindle poles at mitosis and has a role in cell fate in neurons [36] that may extend to other cell types: therefore in neuronal cells it has been suggested that HTT regulates the balance between survival and death. Most of the experimental work examining these functions has only looked at long (>40) CAGs compared with a single normal range CAG length thus our knowledge of the downstream effects of small differences in the CAG repeat length below 40 CAGs on HTT biology are limited. If small modulations in CAG length in HTT and potentially other genes containing polymorphic CAG repeats impose a relatively small effect on cell fate decisions over a long period then they could well be one of the multiple factors that influence whether a cell divides or dies, contributing to the risk of uncontrolled cell division. The effect could in part be explained by the RNA generated from the expanded CAG repeat in HTT. sCAGs are small CAG repeat RNAs generated from the HTT gene [37] which are toxic to neurons [38] and may operate via an RNAi-based mechanism and downregulate trinucleotide repeat-containing survival genes, leading to tumour cell death [39]. Similar findings in other neurodegenerative diseases [7] might implicate a broader biological relationship between cell survival and cell death in the CNS, manifesting as neurodegeneration. Thus determining the relationship between CAG length in HTT and other polymorphic CAG repeat loci could well reveal fundamental biological mechanisms underlying both cancer risk and neurodegeneration.

ACKNOWLEDGMENTS

We wish to acknowledge the time and effort of all participants in this study. We thank Daniel Nelmes for helpful advice regarding cancer diagnoses. Funding: the European Huntington’s Disease Network is funded by the CHDI Foundation, Inc.: this project was EHDN project 478, and this work was supported by the MRC (MR/L010305/1).

AUTHOR CONTRIBUTIONS

The REGISTRY Investigators of the European Huntington’s Disease Network collected the data. PM, RP, RR, RN and DF analysed the data, AH assisted in interpreting the clinical data, PH and DF supervised the data analysis, LJ conceived the study and wrote the main manuscript text. All authors reviewed the manuscript.

CONFLICT OF INTEREST

The authors declare no competing financial interests.