Heritable risk factors for health complications in childhood cancer.

Childhood cancer patients and survivors suffer from serious health complications during and
after treatment. The effect of heritable risk factors on health complications is still insufficiently
understood. Heritable risk factors can be assessed at or early after a childhood cancer diagnosis as
they do not change over time. These markers can then guide treatment, prophylactic and supportive
measures, and follow-up care.

At the start of my PhD, I identified several gaps in the research on heritable conditions
affecting childhood cancer patients and survivors. First, population-based data on heritable cancer
predisposition syndromes were scarce. The Swiss Childhood Cancer Registry hosts a large, curated
dataset of more than 10,000 people affected with childhood cancers including demographic, cancer
diagnosis and treatment, follow-up, and survival data. Information on heritable conditions, i.e.
cancer predisposition syndromes, had been continuously collected but not analyzed. Second,
systematic nationwide collection of germline DNA in Swiss childhood cancer patients and
survivors had not been established and no centralized biobank to collect, process, store, and
analyze germline DNA was available. There was no structure in Switzerland to combine clinical
information with germline DNA and tumor data. Third, there was a lack of data on genetic
modifiers for rare outcomes. For several health complications, research on genetic modifiers had
been performed, but often this information was generated from small patient samples with
conflicting published results. This was the case in sinusoidal obstruction syndrome (SOS) which
occurs after hematopoietic stem cell transplantation (HSCT) and some chemotherapies without
HSCT. Most analyses had used a candidate-gene approach which is limited to prior knowledge for
selection of candidates. Finally, as many health complications associated with childhood cancers
are rare diseases and large cohorts difficult to assemble, there was a need for novel approaches to
the identification of meaningful candidate genes for further genotype-phenotype analysis without
relying on prior knowledge.

I addressed the first gap by analyzing data from the Swiss Childhood Cancer Registry of 8074
patients and survivors, and found that 94 were diagnosed with a second primary neoplasm (SPN)
by age 21 years. I identified 304 patients with cancer predisposition syndromes (CPSs) diagnosed
in regular clinical practice. I found that the incidence of SPNs was more than 10-fold higher in
childhood cancer patients than the incidence of neoplasms in the general population. The
cumulative incidence of SPNs 20 years after first primary neoplasm diagnosis was 23% in patients
with CPSs and 2.7% in those without. I found that CPSs were associated with an almost 8-fold
increase in risk for SPNs, while chemotherapy, radiotherapy, hematopoietic stem cell20 Abstract
transplantation, and older age at first primary neoplasm diagnosis each increased the risk 2-fold
(Publication I, chapter 4.1).

To address the second gap, I set up the protocol for nationwide germline DNA selfcollection
for childhood cancer survivors from their homes (entitled “Germline DNA Biobank
Switzerland for childhood cancer and blood disorders”, BISKIDS) as part of the existing Geneva
Biobank for Hematology and Oncology in Pediatrics (BaHOP). Of 928 invited eligible survivors,
463 (50%) participated. I identified foreign citizenship (odds ratio [OR] 0.5, 95% confidence
interval [CI] 0.4–0.7), older age at study invitation (OR 0.5, CI 0.4–0.8), and having a known cancer
predisposition syndrome (OR 0.5, CI 0.3–1.0) to be associated with non-participation
(Publication II, chapter 4.2). I then set up the genotype-phenotype research project entitled
“Genetic risks for childhood cancer complications in Switzerland” (GECCOS) using samples
collected through BISKIDS. We designed three sub-projects on pulmonary complications, hearing
loss, and second primary neoplasms. This project will serve as a backbone for further association
studies (Publication III, chapter 4.3).

To address the third gap, I performed a systematic review on sinusoidal obstruction
syndrome to better understand what was known on genetic predictors for this complication
(Publication IV, chapter 4.4). I found that, despite more than 20 years of research on genetic
markers for this health complication, only a few genetic predictive markers were identified in more
than one sample. Genetic markers that were tested in multiple studies showed often conflicting
results. Only glutathione S-transferase alpha 1 (GSTA1) variants and methylenetetrahydrofolate
reductase (MTHFR) variants in high-risk patients after busulfan-based conditioning regimens were
repeatedly identified. Only one study performed whole-exome sequencing with replication of
findings in an independent cohort. Finally, I was part of a team of researchers from the
CANSEARCH research platform in pediatric oncology and hematology of the University of
Geneva and the Swiss Institute of Bioinformatics of the University of Lausanne. We developed a
pipeline on how to prioritize genes and genetic variants for analysis on health complications
associated with specific treatment exposures (Publication V, chapter 4.5). We combined in vitro
differential gene expression of lymphoblastoid cell lines after busulfan exposure with clinical
whole-exome sequencing on SOS. We used a combined test statistic to prioritize genes and gene
variants for further analysis. For the pipeline we used SOS after busulfan exposure during HSCT,
but our model might serve other rare diseases as well.