Collection of the data from the known CMMRD cases

In 2008, all 78 CMMRD patients known at that time were compiled in a table.8 Since then, all subsequently published patients with a molecularly confirmed diagnosis of CMMRD were added to this list. Retrospectively, two patients were excluded from the original list because they carried MLH1 variants with arguable pathogenicity, that is, NM_0002439.3:p.Ala441Thr11 and p.Lys618Ala.12 As of June 2013, the table lists 127 patients from 79 families published in 63 papers (see online supplementary Table S1) as well as 19 so far unpublished patients (12 families). Of these 146 patients, 58% (60% of the families) carried biallelic PMS2 mutations while the remaining roughly 40% of patients/families are more or less equally distributed among MSH6 and MLH1/MSH2 biallelics (see table 1). This distribution is in contrast to LS patients of whom the large majority carry heterozygous MLH1 or MSH2 mutations and only a minority heterozygous PMS2 mutations. In part, this difference may reflect the reduced penetrance of heterozygous PMS2 mutations. Studies in unselected cohorts of colorectal cancer patient show that the prevalence of PMS2-associated LS is higher than previously thought.13 ,14

In the following, we will describe the clinical phenotype as it can be deduced from these 146 patients. Data of the so far unpublished patients are also included in the overall analysis presented here, but detailed information on these patients will be published elsewhere. Since nearly all of the patients are described as individual medical case reports, potential selection bias must be kept in mind. To assess the full cancer spectrum and true cancer risk associated with CMMRD natural history, further studies that screen for CMMRD in unselected patients’ cohorts are needed. These studies will also uncover the full spectrum and the frequency of non-neoplastic CMMRD features.

Malignancies in CMMRD patients

In almost all (145/146) of the patients reported so far, malignancies or at least premalignancies, that is, mainly bowel adenomas, were reported. Three biallelic mutation carriers who were reported in the original papers to have no tumours15–17 developed malignancies under surveillance18 (Gerdes and Illencikova 2013, unreported data). Although these observations suggest an extraordinary high tumour risk in CMMRD patients, data from a larger number of presymptomatically tested patients followed over a longer period of time are needed to assess the penetrance of CMMRD with respect to cancer development. Premalignancies and benign neoplasms not taken into account, 139 patients developed a total of 223 malignancies. In this calculation, multiple synchronous colon carcinomas were counted as only one malignancy. Table 2 summarises the spectrum and age of diagnosis of these malignancies. Online supplementary table S2 renders more details by splitting the data according to the mutated MMR gene.

In more detail, 48 malignancies of the haematopoietic system were diagnosed in 45 patients. Three patients had two different metachronous malignancies of the haematopoietic system. Within this group, non-Hodgkin's lymphomas (NHL) were the most prevalent tumour entity. Of the 31 NHL, 20 were derived from the T-cell linage and 4 from B-cell lineage while lineage determination was not available for the remaining. Among the nine acute lymphoblastic leukaemias, four were classified as T-cell acute lymphoblastic leukaemia (T-ALL) while the rest were unclassified ALL. Acute myeloid leukaemia (AML) (five patients) was less common. Classification was difficult in three cases with haematological malignancies reported as atypical chronic myeloid leukaemia, malignant lymphoma and acute leukaemia.

Taken together, a total of 81 brain/CNS tumours were identified in 78 CMMRD patients. Among this tumour entity, the largest group were the high-grade gliomas (58 tumours), including 34 glioblastomas (this also includes two gliomatoses cerebri), 8 anaplastic astrocytomas, 6 oligodendrogliomas and 2 gliosarcomas. Assuming that they were of or eventually developed into high grade, also four unspecified gliomas, three unspecified astrocytomas and one fibrillary astrocytoma WHO II were included in the group of high-grade gliomas. Less common were intracranial embryonal tumours. CNS primitive neuroectodermal tumours (CNS-PNET) developed in eight patients and medulloblastoma in seven patients. Further, there are individual cases with specific brain tumour entities that do not fall into the two main categories (see table 2 and see online supplementary Table S2) and five unspecified brain tumours reported.

A total of 88 LS-associated malignancies developed in 59 patients. The vast majority were colorectal cancers (CRC). Six patients developed metachronous CRC and at least 16 patients presented with synchronous CRC. Because the exact number of synchronous carcinomas was not always given, we counted multiple synchronous CRC as only one malignancy, but the metachronous carcinomas each individually. Calculating by this way, a total of 59 CRC were found in 53 patients (table 2 and online supplementary table S2). Other LS-associated tumours were less frequent. A total of 12 patients developed 18 independent metachronous cancers of the small bowel (duodenum, jejunum, ileum), 6 patients endometrium cancer, 4 cancer of the urethral tract and 1 an ovarian cancer.

As can be deduced from table 2, the distribution of age at diagnosis was broad in each of the three main tumour groups. Nevertheless, there is clearly a trend of haematological tumours developing at earlier age (mean age at diagnosis 6 years) than brain/CNS tumours (mean age at diagnosis 9 years) and LS-associated tumours (mean age at diagnosis 17 years). The youngest patient with colon cancer was 8 years at diagnosis, while the youngest patient with a NHL was only a few months old, and brain tumours were diagnosed in three patients already at the age of 2 years.

Figure 1 illustrates the differences in the prevalence of these three main tumour groups in CMMRD patients with biallelic MLH1/MHS2 mutations compared with those with MSH6 and PMS2 biallelic mutations, respectively. Brain tumours were significantly more prevalent in PMS2 mutation carriers than in MLH1/MSH2 biallelics (p=0.01), and there was a trend of them being more prevalent also in MSH6 than in MLH1/MSH2 biallelics. Haematological malignancies are significantly more prevalent in MLH1/MSH2 than PMS2 biallelics (p=0.04). Also, haematological malignancies were more often the primary tumour in MLH1/MSH2 mutation carriers comprising 38% of the cases (vs 25% and 16% of the MSH6 and PMS2 mutation carriers, respectively). In contrast, 50 and 60% of the PMS2 and MSH6 mutation carriers, respectively, but only 30% of the MLH1/MSH2 biallelics developed a brain tumour as first malignancy. No substantial differences for the prevalence of LS-associated tumours were observed between the groups.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Differences in the prevalence of the three main tumour types, that is, haematological, brain and LS associated, in CMMRD patients with the three different genotypes, that is, MLH1 or MSH2 (MLH1/MSH2), MSH6 and PMS2 biallelic mutations, are shown. The number of patients with the respective tumour type within the genotype group is given above the representative bar, and the mean age at tumour diagnosis is given in years (y) within each of the bars. Statistically significant differences in the prevalence of the tumour types between the genotypes are indicated. χ² testing was used for statistical analysis.

Looking at the mean age at diagnosis of the malignancy, MLH1/MSH2 biallelics tend to develop each tumour entity at lower age than MSH6 and PMS2 biallelics (figure 1 and see online supplementary table S2). This trend is also reflected in an overall lower mean age at diagnosis of the first malignancy in CMMRD patients with MLH1/MHS2 mutations than in MSH6 and, particularly, PMS2 biallelics (see table 1). However, the latter numbers are not very different (7.5 years in MLH1/MHS2 biallelic vs 8 and 10 years in MSH6 and PMS2 biallelics, respectively). This may be partly due to the fact that the group of MLH1/MSH2 biallelics contains two siblings compound heterozygous for the hypomorphic MSH2 mutation c.1A>G and a deletion encompassing exons 1–6 of this gene (c.1-?_1076+?del). These patients developed multiple LS-associated tumours in their fourth and fifth decades of life.19 Excluding these two siblings, the differences between the genetic groups are more obvious (5 vs 8 and 10 years). The percentage of biallelic mutation carriers with more than one malignancy was lowest in MLH1/MSH2 and highest in PMS2 biallelics (see table 1). This may indicate that the chance to survive the first tumour and develop a second metachronous malignancy is better in PMS2 biallelic mutation carriers than in MLH1/MSH2 biallelics. This observation adds to the notion that patients with biallelic MLH1/MSH2 mutations show overall a more severe phenotype than those with biallelic MSH6 and PMS2 mutations.

There were also a number of additional malignancies that do not fall into the three main tumour categories and that were diagnosed each in so far only one genetically proven CMMRD patient. These malignancies included embryonal tumours and sarcomas. Apart from the genetically proven CMMRD patient listed in table 2, rhabdomyosarcomas were also reported in two patients with an inferred diagnosis of CMMRD.6 ,20 Rhabdomyosarcoma, the most common soft tissue sarcoma in children, is discussed to be associated with NF1.21 Given the phenotypic overlap of CMMRD with NF1, CMMRD patients with rhabdomyosarcoma and signs of NF1 may have been misdiagnosed in the past. Hence, there is evidence that the frequency of CMMRD among patients with rhabdomyosarcoma may be higher. Also, there is so far unreported evidence that osteosarcoma may be a tumour entity that is more frequent in CMMRD than previously thought (Brugières for the French CMMRD consortium unreported).

Premalignancies and non-malignant tumours in CMMRD patients

The spectrum of tumours found in CMMRD patients includes also premalignancies and non-malignant tumours (see table 3). A high percentage of the published CMMRD patients developed adenomas of the gastrointestinal tract, which are considered the premalignant lesions from which gastrointestinal cancers develop. Adenomas of the colon and rectum were reported in 52 (36%) of the patients, and many of them developed multiple synchronous adenomas ranging from a few to up to 100 polyps22 reminiscent of (attenuated) FAP.23 ,24 In eight patients, adenomas of the small bowel were reported. At least one patient developed gastric polyps.

Hepatic adenomas were found in three patients with a genetically confirmed diagnosis of CMMRD,25 as well as in one of the siblings from the original Turcot family10 and in one so far unreported CMMRD patient. The latter two cases are not included in table 3 because the patient reported by Turcot was never molecularly confirmed and the unreported patient was found to be homozygous for a so far unknown PMS2 missense variant and, hence, his molecular diagnosis could not be confirmed unambiguously.

Benign cutaneous or plexiform neurofibromas were reported in eight patients and one patient had an optic glioma.26 The latter might be a sign of (mosaic) NF1 in this CMMRD patient who also had six CALMs and freckling and, therefore, met the diagnostic criteria of NF1. This WHO grade I tumour is not listed in table 1 as it is not considered a malignant brain tumour. But it is noteworthy that Ostergaard and colleagues27 report a patient with a brain tumour of the left parieto-occipital area, which initially showed the histology of a pilocytic astrocytoma. This tumour recurred within 6 months, then characterised as anaplastic astrocytoma.

Pilomatricomas, also termed calcifying epithelioma of Malherbe, are benign tumours of the hair follicles. In two patients with genetically confirmed CMMRD syndrome, multiple pilomatricomas were reported. The authors proposed that the presence of multiple pilomatricomas that is seen very rarely in otherwise healthy individuals should raise a suspicion for an underlying MMR repair defect when observed in a paediatric cancer patient.28 Of note, there may also be an association of multiple pilomatricomas with MUTYH-associated polyposis,29 a polyposis syndrome that has certain phenotypical and patho-mechanistic overlap with CMMRD since both recessively inherited tumour predisposition syndromes are caused by DNA repair defects leading to attenuated polyposis although usually in different age groups.

One patient had polyps of the vocal cord in infancy; the link of this latter sign with CMMRD is not proven yet.

Non-neoplastic features of CMMRD

The most prevalent non-neoplastic features seen in CMMRD patients are CALMs and other pigmentary alterations. More than 60% (91/146) of the CMMRD patients were reported to show at least one CALM or hyperpigmented skin area (table 4) and only four published19 ,30–32 and one so far unpublished patient have been explicitly reported to lack these pigmentary alterations. The vast majority of these patients had multiple (two or more) CALMs, but they did not always reach the critical number of ≥6, which is needed to be a diagnostic criteria of NF1. Several authors stress that the degree of pigmentation and the shape of these hyperpigmented skin macules often differed from the CALMs typically seen in NF1 patients. Nevertheless, several patients showed classical NF1-associated CALMs. Twenty-seven patients showed (often in addition to CALMs) other features diagnostic for NF1. These features included freckling in 14, neurofibromas in 8, Lisch nodules in 5 and tibia pseudarthrosis, sphenoid wing dysplasia and optic glioma each in 1 patient. Also, 3 of the 27 patients were reported to have a NF1 phenotype, but the signs were not specified.13 ,33 Because several patients showed a segmental/regional distribution of NF1 signs6 ,34 and a postzygotic NF1 mutation was found in one patient,35 it is thought that classical CALMs and other typical NF1 signs in CMMRD patients result from postzygotic NF1 mutations present in a segmental or mosaic status. Another pigmentary abnormality reported in at least nine CMMRD patients are areas of skin hypopigmentation, also reported as ‘ash leaf spots’ or ‘vitiligo’ (table 4).

An analysis of three patients with biallelic PMS2 and eight with biallelic MSH6 mutations has shown that constitutional deficiency of these MMR genes leads to impaired immunoglobulin (Ig) class switch recombination characterised by a decrease or absence of IgG2, IgG4 and IgA concomitant (particularly in young patients) with increased IgM levels, that is, hyper-IgM syndrome.36 ,37 IgA deficiency indicative of this humoral defect was also observed in a patient with a homozygous MSH2 mutation.38 Hence, in total 12 CMMRD patients were reported to show defects in Ig class switch. In most of the other CMMRD patients, this parameter was not tested since an increased susceptibility to severe and persistent infection that would be indicative of an immune defect was noticed only in very few CMMRD patients.

Eight CMMRD patients had congenital malformations. Agenesis of the corpus callosum (ACC) is reported so far in four patients, and in three of these patients grey matter heterotopia was also detected.30 ,39 Baas and colleagues30 render evidence that these features are seen more frequently in CMMRD than in the general population. Therefore, it was suggested that ACC with or without grey matter heterotopia is associated with CMMRD and should raise the suspicion of this disorder in a paediatric cancer patient. It was speculated that these isolated and largely asymptomatic brain malformations in CMMRD patients result from early embryonic somatic mutations in one or more of the genes implicated in the callosal development. Equally, it may be speculated that somatic mutations in developmental genes (presumed to occur more frequently in CMMRD) could be responsible also for other malformations that were observed in CMMRD patients. Non-therapy-induced brain cavernomas were seen in two published patients25 ,40 and one so far unreported patient. The combination of congenital asplenia, left isomerism and a ventricular septum defect was observed in one patient.23 It will be worthwhile to systematically analyse whether these and other malformations are also found at a higher frequency in CMMRD patients than in the general population.

Furthermore, two patients developed lupus erythematosus31 ,41 and two haemangioma of the skin.30 ,42

Diagnostic criteria that should raise the suspicion of CMMRD syndrome in a cancer patient

An early and definite diagnosis of CMMRD in a paediatric and young adult cancer patient is desired for several reasons. (1) Although the true cancer risk of CMMRD is currently not known (and may be overestimated since we have to rely on a potentially biased cohort of patients), the available data from the published patients with this rare disorder suggest that it is extraordinarily high. CMMRD patients who survive their first malignancy or in whom a premalignancy has been removed have a high risk of developing a second often different (pre-) malignancy. To improve the prognosis, surveillance should be offered at least for bowel cancer and brain tumours as recently proposed by the European consortium.43 (2) Early and definite diagnosis may also be a prerequisite in the future to adjust treatment to the underlying MMR defect and the high risk of a secondary malignancy. Currently, no information is available regarding the optimal treatment for CMMRD patients. But careful attention should be given to a possibly reduced efficacy and increased cytotoxicity of certain chemotherapeutic agents due to constitutionally impaired mutation repair.7 ,44 Since MMR-deficient cells are profoundly resistant to O⁶ methylators such as temozolomide,45 ,46 the risk of therapy failure may be increased in CMMRD patients treated with this drug. Furthermore, these drugs may increase the risk of second primary tumours by accelerating the rate of unrepaired mutations. (3) Siblings of a CMMRD patient have a 25% risk of having inherited the same genotype and, hence, equally a high risk for childhood cancer. (4) The parents of CMMRD patients and 50% of their siblings as well as other more distantly related family members are heterozygous for the MMR mutation(s) and, therefore, have an increased risk for LS-associated tumours in adulthood. A definite molecular diagnosis is needed to offer the families of CMMRD patients appropriate counselling and discuss with them the options of predictive testing as well as prenatal/preimplantation diagnostics if this is desired.

It might be considered desirable to establish a diagnosis before the development of the first malignancy in a biallelic MMR gene mutation carrier. Currently, however, this seems to be only achievable for siblings of a tumour patient with a molecularly confirmed diagnosis. The non-neoplastic features of CMMRD may be too subtle and unspecific to raise the suspicion in a child without a malignancy if there is no family history of CMMRD or diagnosis of LS in the family. Therefore, it was decided at the first meeting of the C4CMMRD consortium that, at the time being, the group will only propose clinical criteria that should raise the suspicion of CMMRD when present in a cancer patient. Questions that were discussed among the group were

1. Which tumour entities should raise a suspicion?

2. What is the age limit for a patient with a specific tumour entity to be suspected?

3. Which non-neoplastic feature should be considered a diagnostic criterion?

Ad A): The spectrum of malignancies in CMMRD is very broad (see table 2). Therefore, any malignancy in a paediatric (young adult) patient could be a CMMRD-associated one. Nevertheless, the index of suspicion should be higher in tumour entities that are overrepresented in CMMRD patients compared with their proportion in all tumours of the general population. Taking this into account, we developed a scoring system for the suspected diagnosis of CMMRD. Tumours assigned three points are highly specific for CMMRD syndrome. A diagnosis of CMMRD should be suspected in patients with these tumour entities no matter whether they show additional (non-neoplastic) features of CMMRD or not. Malignancies assigned two points are overrepresented in CMMRD but less specific. Additional features or tumours that add up to three points need to be present in patients with these malignancies to suspect CMMRD. All other malignancies are assigned one point , and additional tumours or features strongly pointing into the direction of CMMRD need to be present to raise a suspicion that should entail further diagnostic steps.

Colorectal or other cancers of the LS spectrum are extremely rare below the age of 25 years even in LS and FAP. Because they are highly suggestive for CMMRD, which phenotypically presents in several cases as a very early onset form of LS, three points are assigned to these tumours.

Multiple bowel adenomas are a frequent finding in CMMRD, which hence shows clinical overlap also with (attenuated) FAP. Herkert et al23 proposed that in the absence of proven APC or MUTYH germline mutations, patients with childhood-onset adenomatous polyposis should be considered for MMR gene mutation testing, especially when they have features of NF1. In CMMRD, both the adenoma formation and the adenoma carcinoma transition may be accelerated due to the greatly enhanced mutation rate in neoplastic and non-neoplastic tissue. In agreement with this notion, most (35/52) of the CMMRD patients with adenomas showed high-grade dysplasia in at least one of them or had synchronous bowel cancer. Therefore, CMMRD syndrome should be considered as a differential diagnosis also in a patient under the age of 25 years (for considerations concerning the age limit of 25 years, see ad (B)) with (i) multiple adenomas if a heterozygous APC mutation or biallelic MUTYH mutations are absent (exclusion of POLD1 and POLE hotspot mutations may also be considered47) or (ii) in a patient with a single adenoma with high-grade dysplasia. Consequently, multiple bowel adenomas and absence of an APC/MUTYH mutation and/or one adenoma with high-grade dysplasia under the age of 25 have a score of three points.

High-grade gliomas, including glioblastoma, gliosarcoma, anaplastic astrocytoma, oligodendroglioma and others, are rare tumours in childhood and adolescence. According to the data of the French Brain Tumour Data Bank48 and the German childhood cancer register (http://www.kinderkrebsregister.de), they represent 15% of all brain/CNS tumours in childhood and adolescence. Hence, assuming that a quarter of all tumours in childhood are brain/CNS tumours, high-grade gliomas represent less than 5% of all tumours in childhood/adolescence. In contrast, 26% of all malignancies so far seen in CMMRD patients are high-grade gliomas. Equally, CNS-PNET constituting 3.5% of all so far reported malignancies in CMMRD (vs 0.5% of all childhood tumours in the general population) are overrepresented in CMMRD. Among the haematological tumours, NHL are overrepresented in CMMRD. NHL constitute ∼5–7% of all childhood tumours but 14% of all CMMRD-associated malignancies. Of note, at least 65% (20/31) of the CMMRD-associated NHL were from the T-cell lineage and at least 42% (13/31) were T-cell lymphoblastic lymphomas. This is in contrast to the distribution in the general population where ∼65% are from the B-cell lineage and more than 50% are mature B-NHL.49 Hence, NHL of the T-cell lineage mainly account for the overrepresentation of NHL in CMMRD. Taken together, T-NHL, high-grade gliomas and CNS-PNET are considered malignancies typical for CMMRD and are assigned two points in the scoring system.

Ad B): The age at diagnosis of the first malignancy ranges from 0.4 to 39 years in CMMRD patients (table 1). However, the vast majority, that is, 120 of the 146 CMMRD patients, were younger than 18 years of age when their first tumour was diagnosed. Only 17 patients were between 18 and 25 years when they developed their first tumour, which was a CRC in 11 patients, a glioblastoma in 1 patient and an oligodendroglioma in 1 patients. The remaining four patients had colorectal adenomas when they were young adults. Only four patients were older than 25 years when they were diagnosed with their first tumour; in all four cases a CRC.13 ,19 ,41 These four patients carried at least one allele a likely hypomorphic MSH2, MSH6 and PMS2 mutation, respectively, and, therefore, may genetically as well as clinically represent an intermediate phenotype between CMMRD and LS. Based on the observations in the patients reported so far, we set the age limit at diagnosis of the first tumour at 25 years for patients with CRC, colorectal adenomas and malignant gliomas. For all other patients with malignancies, the age limit is <18 years (see table 5).

Ad C): Non-neoplastic features are weighted according to their specificity for CMMRD and their frequency in the general population with one point or two points.

Presence of (segmental) NF1 signs, primarily multiple (≥6) CALMs and freckling, is so far the most commonly reported non-neoplastic feature associated with CMMRD, and it has been stressed already in several reports that presence of one or more of these signs should raise the suspicion of CMMRD in any paediatric cancer patient with the exception of children/young adults with clearly NF1-associated malignancies, such as a peripheral nerve sheath tumour (MPNST) or a juvenile myelomonocytic leukaemia (JMML) or with a parent who is also diagnosed with NF1. Hence, this highly specific feature is weighted two points.

Although many CMMRD patients came to attention through the presence of NF1 signs, our analysis of the reported cases clearly indicates that limiting suspicion to patients showing classical NF1 signs will miss a number of patients. In several CMMRD patients, the number of CALMs has been reported to be below six, which is the necessary number to be diagnostic for NF1. Furthermore, several CMMRD patients do not show classical NF1-associated CALMs but have different kinds of skin hyperpigmentation and hypopigmentation. As an isolated finding these alterations are frequent in the general population. A study assessing the frequency of CALMs among school children aged 4–11 years showed that ∼20% (146/732) had one CALM. However, only 4.1% (30/732) and 1.2% (9/732) had two and three CALMs, respectively.50 Merks and colleagues51 report a slightly lower frequency of solitary and multiple CALM in 13% and 3.3% of school children, respectively. Hence, the likelihood of having a childhood malignancy and two or more CALMs (hyperpigmented macules) by pure chance is very low. The same is likely the case for hypopigmented macules. Therefore, the presence of two or more skin areas of hyperpigmentation or hypopigmentation with a minimum diameter of 1 cm should be another feature that should raise the suspicion of CMMRD. This feature is also weighted two points.

Brain malformations such as agenesis of corpus callosum and non-therapy-induced cavernoma as well as pilomatricomas/epitheliomas of the Malherbe are rare in the general population. But they seem to have a higher incidence in CMMRD patients (see sections ‘Premalignancies and non-malignant tumours in CMMRD patients’ and ‘Non-neoplastic features of CMMRD’). Therefore, these features should also raise the suspicion of CMMRD when present in a childhood cancer patient and are, thus, weighted depending on their frequency in the general population 1–2 points. The frequency of brain malformations in CMMRD patients still needs to be evaluated prospectively. As long as they have not been shown to be more frequent, we do not advocate performing cranial MRI unless required for other clinical reasons to test for these features only in order to confirm the diagnosis CMMRD in a cancer patient. The brain malformations add one point to the score of a cancer patient. Multiple pilomatricomas, which are extremely rare and always indicate an underlying genetic defect, add two points. A single pilomatricoma adds only one point.

Consanguinity of the parents and/or homozygosity for one (founder) mutation is observed in 46 of the 91 so far reported families with CMMRD patients. However, consanguineous marriages are common in several ethnic and religious groups of the European population. Therefore, we decided that consanguinity should add one point and raise the suspicion of CMMRD in a patient with a paediatric cancer that is overrepresented in CMMRD syndrome and in any other paediatric cancer patient who has one of the other features.

Due to the impaired class switch recombination, reduced IgG2/4 and/or IgA levels could be observed in several CMMRD patients. But this feature is neither very specific for CMMRD nor particularly rare in the general population. Hence, it is also assigned one point.

Other criteria for suspecting CMMRD in a child with a malignancy are (i) a sibling with a malignancy and (ii) the molecular diagnosis of LS or an LS-associated tumour before the age of 60 years in a first-degree , second-degree or third-degree relative of the patient. When assigning points to these criteria, it was taken into consideration that typical CMMRD tumours in a sibling should raise a higher index of suspicion (two points) than any other tumour in a sibling (one point). Equally, a definite diagnosis of LS in the family is assigned two points, whereas an LS-associated tumour before the age of 60 years in a first-degree, second-degree or third-degree relative of the patient is scored only one point. A LS-associated tumour in a relative that has been shown to be microsatellite stable, and hence, is very likely a sporadic tumour should not be taken into account.

According to the three-point scoring system summarised in table 5, CMMRD syndrome should be suspected in an individual who reaches a score of minimum three points. The points should be counted by adding the points assigned to the malignancy/malignancies and those assigned to additional features. When using this system, it has to be kept in mind that several of these features listed in table 5 may be present also in a patient with a different (childhood) cancer predisposition syndrome, like Fanconi anaemia, ataxia telangiectasia, Bloom syndrome, Peutz–Jeghers syndrome, tuberous sclerosis and others. These syndromes are usually associated with characteristic anomalies that are not listed in table 5. Therefore, patients with signs specific for another cancer predisposition syndrome except for NF1 should first be tested for the other cancer predisposition syndrome. Patients who have clearly NF1-associated malignancies such as JMML or MPNST and signs reminiscent of NF1 should be first tested for a NF1 mutation. There is also an overlap of CMMRD with Li Fraumeni syndrome (LFS). The tumour spectrum of the latter encompasses brain tumours and, to a lesser extent, haematological neoplasms. In families fulfilling Chompret's criteria,52 a TP53 germline mutation should be ruled out first.

Diagnostic steps to substantiate the diagnosis CMMRD

Patients reaching a score of three points should be further analysed to confirm or refute the suspected diagnosis of CMMRD syndrome. The diagnostic steps in these patients largely follow the protocols developed for LS, which involves analysis of microsatellite instability (MSI) and/or immunohistochemical (IHC) staining of the four MMR proteins.

IHC analysis to assess the expression loss of the affected MMR protein can be effectively employed in all solid tumours of CMMRD patients and has the advantage to guide subsequent mutation analysis in the four MMR genes. In general, biallelic truncating mutations in PMS2 or MSH6 will result in isolated loss of these proteins, whereas mutations in MLH1 or MSH2 will lead to concurrent loss of MLH1/PMS2 or MSH2/MSH6, respectively, since MLH1 and MSH2 are the obligatory partners in the formation of MLH1/PMS2 and MSH2/MSH6 heterodimers. Of note, in the case of an underlying missense mutation, IHC may show normal expression of the affected MMR gene, which may be a possible pitfall when using IHC analysis to confirm suspected CMMRD. In contrast to LS where expression loss is observed only in neoplastic cells, in CMMRD patients IHC detects expression loss of one (or two) of the MMR proteins in both neoplastic and non-neoplastic tissues. Hence, negative IHC staining in neoplastic and surrounding normal cells should not be interpreted as a failure of proper staining and care should be taken to use a (on slide) staining control from a different individual. In principle, expression loss of one of the MMR genes can also be demonstrated in blood lymphocytes of CMMRD patients, as shown by western blotting37 but possibly also by IHC staining on cytospin preparations.

MSI tests for small misalignments, that is, insertion–deletion loops (IDL), which arise during DNA replication and remain uncorrected in the absence of a functional MMR system. It takes advantage of the fact that repeated-sequence motifs consisting of units of one, two or a few more nucleotides, that is, microsatellites, are frequent targets of IDL errors that result in shortening or lengthening of these sequence motifs, a phenomenon termed MSI. In tissues derived from clonally proliferating MMR-deficient cells, that is, neoplastic cells, MSI is easily detected by PCR amplification and fragment analysis of a set of microsatellite markers.53 MSI analysis following the current protocols for LS uses a panel of 5–6 dinucleotide54 and/or mononucleotide55 repeat markers. This approach is a reliable tool to diagnose MMR deficiency in gastrointestinal and other LS-associated tumours of CMMRD patients. However, standard MSI analysis as applied in LS frequently fails to show MSI in brain tumours42 ,56 ,57 and other malignancies.30 Currently, the reasons for this observation are unknown, but it has been shown that more subtle shifts (shortening or lengthening) of microsatellite alleles can be observed also in brain tumours and may be indicative of CMMRD.58 Furthermore, it might be worth testing whether a different panel of microsatellite markers would be more sensitive in brain and other tumours of CMMRD patients.

In principle, MSI can be observed also in DNA from normal, that is, non-neoplastic, cells of CMMRD patients. However, because altered microsatellite alleles are present only in a small proportion of the cells from normal tissue, so far most approaches used technically demanding single-molecule analyses (eg, small-pool PCR) to show MSI in normal tissue, that is, germline MSI (gMSI).59 Recently, a much more simple method to detect gMSI was presented.60 This assay relies on the analysis of ‘stutter’ peaks typically associated with microsatellite PCR products. When quantified by a novel publicly available software application, the relative peak height of the ‘stutter’ peaks of selected dinucleotide microsatellites significantly increases in DNA of CMMRD patients compared with normal controls as has been confirmed also in a larger cohort of samples (Bodo et al., in preparation). The main limitation of this assay is that the relative peak height is not altered in patients with CMMRD due to biallelic MSH6 mutations. Nevertheless, if this assay shows in a larger cohort of samples that it is a reliable, simple and rapid tool to detect CMMRD at least in patients with PMS2, MLH1 or MSH2 mutations, it would be a good screening tool and an alternative to IHC in cases where appropriate tissue is not available.

Taken together, both IHC staining of the MMR genes and MSI analysis are diagnostic methods to substantiate the suspected diagnosis. Since IHC will also guide target-gene mutation analysis and has been shown to render reliable results in most solid tumours, it is considered the preferred method. But as outlined, both methods have potential pitfalls and may fail to confirm the suspected diagnosis. Therefore, we recommend combining both assays if needed. The final confirmation of the diagnosis CMMRD should come from the determination of the causative biallelic mutations in the patient.

Counselling and genetic testing

According to the recommendations of national and international human genetic societies and the legislation of most European countries, genetic counselling must be offered to the patients and/or their parents prior to performing mutation analysis in the affected child. Patients and/or their parents should be informed by a team of paediatric oncologists and medical geneticists about the suspected diagnosis if this is substantiated by MSI and/or IHC analysis. Considering the burden of this syndrome, psychological support should systematically be proposed to families. The family has to be informed of potential therapeutic implications of the test result and also of the high risk for a second malignancy in a patient with a positive test result. Genetic counselling must also include information on the potential 25% recurrence risk in a sibling and on the risks for LS-associated cancer in possible heterozygous mutation carriers, particularly both parents.

With the informed consent of the patient and/or the parents, mutation analysis will be initiated. Reliable and robust comprehensive analysis exists for all four MMR genes now including the historically difficult PMS2 gene. However, when analysing PMS2, which is the affected gene in more than 50% of CMMRD patients, special care has to be taken to avoid pitfalls that arise from high prevalence of hybrid alleles that result from sequence exchange of the functional gene with its pseudogene PMS2CL.33 ,61–66

Preferentially, targeted gene mutation analysis is performed. However, in cases where tumour tissue is not available for IHC analysis or the results are inconclusive, mutation analysis of all four MMR genes can be considered. It is expected that with the implementation of next-generation sequencing techniques this may be possible at reasonable costs in many laboratories in the near future.