Temporal order of RNase IIIb and loss-of-function mutations during development determines phenotype in DICER1 syndrome: a unique variant of the two-hit tumor suppression model

Patients and specimens

PPB patients (n = 124) and family members were ascertained through the International PPB Registry (IPPBR). Inclusion into this study required a pathologic diagnosis of PPB verified by central review (LPD, DAH). All subjects gave written consent for molecular and family history studies, as approved by the Human Research Protection Offices at Washington University in St. Louis (HSC#04-1154), Children's Hospitals and Clinics of Minnesota (IRB#98107), and Children's National Medical Center (IRB#4603; Pro0315). For families with more than one affected member, only data from the initial proband is included. Medical history and biological samples were collected and prepared for analysis as previously described^4,30. Tumor tissue was available for sequencing from a subset of patients. For two of these cases, DNA was isolated from unstained tissue on glass slides using the Pinpoint Slide DNA Isolation System (Zymo, Irvine, CA).

Definition of "disease foci"

Clinical data was abstracted from medical records and imaging studies. All children had pathologic confirmation of PPB. In addition, lung cysts, kidney cysts, CN, NCMH, SLCT, ERMS, thyroid cancer or nodules, pineoblastoma and/or pituitary blastoma were defined as evidence of syndromic disease. Lung cysts that were distinctly separate (anatomically separate in same lobe or in different lobes) were counted individually.

Mutation testing

Initial sequencing of blood and saliva DNA samples was by standard Sanger methods described previously⁴ or by a commercial laboratory (Ambry Genetics, Aliso Viejo, CA). Low-frequency variants were detected and quantified by targeted next-generation sequencing (NGS) using a custom multiplex PCR panel for DICER1 coding regions (Ion Torrent Ampliseq, Life Technologies, Grand Island, NY, USA) (Table S1)³⁰. NGS was performed on an Ion Torrent 318 v2 chip (ION PGM Sequencing 200 kit v2, Life Technologies) with an average of 6 samples per chip, to achieve an average depth of coverage of 3000 filtered reads. Signal processing, mapping and quality control were performed with Torrent Suite software v.4.0.2 (Life Technologies). Variant calls were made using the Torrent Variant Caller Plugin v.4.0, with somatic low stringency mutation workflow and default settings. BAM files of raw reads were reviewed using Integrative Genomics Viewer v2.3^37,38.

Annotation of sequence variants and the spectrum of possible mutations

DICER1 sequence variants were annotated with Alamut Batch software (Interactive Biosoftware, Rouen, France), with reference to DICER1 transcript record NM_177438.2. Nonsense, frameshift and canonical splice-site mutations were considered loss of function (LOF). Missense variants affecting codons 1705, 1709, 1809, 1810 and 1813 in the RNase IIIb domain were classified as “hotspot” mutations. For variants assayed by NGS, allele frequencies were calculated from filtered read counts. The SIFT algorithm was used to assess potential significance of novel missense mutations^39–41. All variants identified were deposited into ClinVar (accession numbers SCV000195560-SCV000195643). The numbers of possible single-nucleotide changes that can produce amino acid substitutions at the five hotspot codons or nonsense mutations anywhere in the DICER1 open reading frame, or disrupt canonical splice sites, were compiled from DICER1 transcript record NM_177438.2 and genomic record NG_016311.1.

NanoString genomic copy number assay

Molecular probes for NanoString Copy Number Assay at the DICER1 locus were developed in collaboration with NanoString Technologies, Inc., Seattle, WA (Table S2). Genomic DNA was fragmented and hybridized using the nCounter Prep Station, and hybridization signals quantified using the nCounter Digital Analyzer, according to NanoString’s recommendations. Preliminary analysis and quality control of the data were performed using nSolver Analysis Software version 1.1 (NanoString) with default copy number variation (CNV) analysis settings. CNVs were confirmed with high-density CNV array hybridization in a commercial laboratory (Prevention Genetics, Marshfield, WI).

Statistical analyses

The number of disease foci per patient and the age at DICER1 syndrome diagnosis were compared between mutation categories using nonparametric tests, due to the skewness of both clinical features and to the unbalanced sample sizes. Kruskal-Wallis tests were used to compare medians among the four mutation categories. Where a significant overall association was found, pair-wise post-hoc Wilcoxon rank sum tests were used to compare medians, and resulting p-values adjusted for multiple comparisons using the Sidak method. A p-value of 0.05 was considered statically significant and all analyses were performed using Stata V13 (College Station, TX).

Most predisposing DICER1 mutations are inherited loss of function (LOF) mutations

Our overall approach to detecting and categorizing predisposing DICER1 mutations in PPB children is shown schematically in Figure 1. We identified germline, heterozygous DICER1 mutations in 90 of the 124 probands in our cohort (72.6%; Table 1, Table S3). Nearly all (89) were detected by Sanger sequencing of exonic PCR amplicons. For one child in whom no mutation was detected by Sanger sequencing, blood DNA was probed by NanoString hybridization, which indicated deletion of one copy of exon 24. High-density CNV array hybridization was used to confirm a heterozygous deletion of ~ 1.1 kb, comprising all of exon 24 and parts of the flanking introns (c.5096-498_5364+356del). Paternal DNA was positive for the deletion, which was anticipated as this child has an uncle with CN. Only one previous instance of a large, intragenic deletion as a germline DICER1 mutation has been reported, which suggests such mutations are very rare⁴². The actual prevalence of large deletions is difficult to estimate because they are not readily detected by the targeted sequencing strategies applied for mutation screening in this study and most others.

Figure 1. Study design – Detection and categorization of DICER1 mutations in PPB probands.

A cohort of 124 children diagnosed with pleuropulmonary blastoma (PPB) was screened for predisposing DICER1 mutations by targeted Sanger sequencing and/or low-depth, next-generation sequencing (NGS) of DNA amplified from peripheral blood cells, saliva (buccal cells) or non-neoplastic surgical specimens. Sequenced PCR amplicons covered the 26 coding exons of the DICER1 open reading frame and flanking splice signals. DICER1 coding sequence or splice site mutations detected at approximately heterozygous frequency in blood or normal tissue cells were categorized as germline mutations. For patients in whom screening revealed no germline mutation, blood and/or normal tissues were analyzed for the presence of intragenic deletions or larger genomic alterations using NanoString copy number assay and CNV array, and for coding or splice site mutations present at low allele frequencies using high-depth NGS on the Ion Torrent platform. Wherever possible, matched tumor specimens were also sequenced on the Ion Torrent platform. Low-frequency DICER1 mutations detected in multiple normal and/or tumor samples, or in primary tumors of multiple organs, were categorized as mosaic mutations. Mosaic mutations were further categorized as loss-of-function (LOF) or RNase IIIb domain hotspot missense mutations. Patients for whom both LOF and hotspot mutations were identified in a single tumor, but not found in blood or normal tissue samples, were categorized as having tumor-specific, biallelic DICER1 mutations.

The spectrum of germline mutations is dominated by truncating, LOF mutations (Figure 2). These are mainly single-nucleotide substitutions that produce new stop codons (33 cases, 37%) and small insertions or deletions (indels) within exons that shift reading frame (44 cases, 49%). Seven mutations of consensus splice sites occur in our cohort; of which six are predicted to cause exon skipping during transcript splicing with resulting frameshift. The remaining splice site mutation, c.1752+1delG, is at the 5’ end of intron 10. Skipping of exon 10 would cause in-frame deletion of 81 amino acids near the end of the helicase domain. In all, 84 of 90 germline DICER1 mutations discovered in patients (93%) truncate the open reading frame before the end of the critical RNase IIIb domain, and are thus predicted to result in complete loss of DICER1 protein function even if the message escapes nonsense-mediated decay. Six non-truncating germline mutations were identified, including the intron 10 splice site mutation described above and five non-hotspot missense changes: I582T, L1583R and G1708E (each seen once) and D1822V (identified in two patients) (Table S4). The I582T substitution is at the distal end of the helicase domain (Figure 2), the role of which is unclear. L1583R is within the RNase IIIa domain and segregates with disease in a family⁴. The G1708E and D1822V mutations both fall within the RNase IIIb domain, near the metal-binding catalytic site. These two missense mutations are predicted to compromise protein function by the SIFT algorithm but their precise functional significance in DICER1 is unknown^39–41.

Figure 2. The spectrum of predisposing loss-of-function mutations in PPB/DICER1 syndrome.

A linear schematic of the DICER1 open reading frame is shown with annotated functional domains represented to scale. Sequence changes identified as inherited or de novo germline mutations in 90 PPB/DICER1 syndrome patients are indicated by position along the coding sequence. Mutations linked to the schematic by two, three or four fine lines are those discovered in a corresponding number of individuals from unique families.

DNA was available from both parents for 77 children with germline mutations, and Sanger sequencing of parental DNA was sufficient to confirm 67 of the mutations (87%) as inherited. Mutations in the ten patients whose parents had no DICER1 mutation detected by Sanger sequencing were provisionally considered de novo. To confirm this, targeted next generation sequencing (NGS) was performed in eight of the ten trios, yielding mutant allele frequencies between 42.0% and 57.1% in the probands but no conclusive evidence of the variants in parental blood. There were no statistically significant differences between de-novo and inherited germline LOF patients with respect to age at onset, numbers of disease foci or survival.

Penetrance of familial DICER1 LOF mutations was far from complete. Of the 67 families in this cohort with segregating LOF mutations, 29 include parents or siblings who are confirmed as mutation carriers but have no history of syndromic disease (Table S5). True penetrance is difficult to estimate because we have limited knowledge of how many germline DICER1 mutation carriers are phenotypically normal, as only a subset with overtly affected family members have been ascertained. Moreover, subclinical disease is common. Preliminary data from an ongoing NCI-sponsored DICER1 family history study indicate that ~ 87% of otherwise asymptomatic individuals with confirmed DICER1 mutations have thyroid nodules detectable by ultrasound and ~ 43% have lung cysts detectable by CT scan (D.R. Stewart and L. Doros, unpublished).

Among children with germline LOF mutations, age at first diagnosis of PPB or other syndromic disease was typically one to five years (70 of 90 patients), but this ranged from diagnosis within days of birth to as late as eighteen years. The most frequent syndromic condition after PPB was cystic nephroma, followed by thyroid disease (nodular hyperplasia or carcinoma), nasal chondromesenchymal hamartomas and embryonal rhabdomyosarcomas (Table 1, Table S5). The number of discrete disease foci per patient ranged as high as five or six (in two patients), but the majority of children in this group had experienced no more than two at the time of their most recent exam, and nearly half had only a single PPB tumor. None of the six patients with non-truncating germline mutations had unusual clinical features and as a group they were not distinguishable from patients with truncating mutations. Table S6 provides data on somatic hotspot mutations identified in all available tumors of PPB children.

Approximately 10% of predisposing DICER1 mutations are mosaic rather than germline

We and others have previously described biallelic DICER1 mutations in tumors of children who apparently have no germline mutation, inherited or de novo^6,14,35. Because PPB children are typically so young when affected, we hypothesized that at least some cases of this kind reflect mosaicism, i.e., a mutation present in some but not all cells of the body, because it occurred during post-zygotic embryonic development rather than being present in the zygote (as a germline mutation would be). To explore this possibility, we performed targeted, high-depth NGS of DICER1 coding exons in DNA from blood and/or other normal tissues of children who had tested negative for germline mutation by Sanger sequencing, and in matched samples of tumor tissue where available. We categorized a DICER1 mutation detected by NGS as mosaic when the following criteria were met: i. The mutation was evidently not a constitutional, germline allele because it was present at sub-heterozygous frequency (arbitrarily taken as below 35% of reads) in peripheral blood and/or other normal tissue samples. ii. The mutation was evidently not specific to a tumor, because the same mutant allele was detected in one or more normal, non-neoplastic tissue samples, OR, the same mutant allele was detected in multiple primary tumors arising in different organs (Figure 1). We identified ten children with predisposing mosaicism for either LOF or RNase IIIb hotspot mutations (Table 1).

Mosaic LOF mutations were detected in five children, at frequencies that ranged from 1.1% to 17.2% of allelic reads in DNA from blood, saliva or normal fibroblasts (Table S7). For three of these children, archival PPB tumor tissue was available, and in each the LOF mutation was present, as was an RNase IIIb domain hotspot mutation. Two of the five children with mosaic LOF mutations had a single focus of disease in a lung. The other three children each had two foci of disease, also restricted to the lungs. It might be anticipated that children bearing mosaic LOF mutations tend to have fewer disease foci than those with germline LOF mutations because the number of cells at risk for second hits is generally lower. No statistically significant difference of this kind can be discerned from the five mosaic LOF children in our cohort, but notably, none have developed syndromic tumors other than PPB. As this was not a population study, we cannot estimate how many persons with mosaic LOF mutations are asymptomatic but, by analogy to the low penetrance of familial LOF mutations, it could be a large proportion.

Five children harbored mosaic RNase IIIb domain hotspot missense mutations, detected in multiple primary neoplasms and/or non-neoplastic tissues (Table 2). None had family members with features of DICER1 syndrome, and the RNase IIIb hotspot mutations found in probands were not detected in parental blood, consistent with a postzygotic origin. NGS of tumor tissues from these children identified somatic LOF mutations or allele loss in some but not all specimens, with the caveat that allele loss can be difficult to detect in specimens with low tumor purity (i.e., PPB Type Ir, CN and NCMH). For one mosaic hotspot patient, specimens of a thyroid carcinoma and two separate ovarian Sertoli-Leydig cell tumors (SLCT) were available for NGS. The thyroid carcinoma and one SLCT had lost the second DICER1 allele, but the other SLCT had instead sustained a frameshift mutation (Table 2). This is consistent with underlying mosaicism for the RNase IIIb hotspot mutation and subsequent acquisition of independent LOF mutations in each tumor site.

Mosaic RNase IIIb hotspot mutations are associated with early-onset, multifocal disease

The five children with mosaic RNase IIIb domain hotspot mutations shared unusual clinical features. All were diagnosed with DICER1 syndrome early; within 15 months of birth. All presented with multiple cysts of the lungs and/or kidneys, which were accompanied or followed in all cases by multiple DICER1 syndromic tumors (Figure 3). Four of the five had CN as well as PPB. Other tumors included SLCT, thyroid nodular hyperplasia or carcinoma, NCMH, ciliary body medulloepithelioma and pineoblastoma. In addition, four of the five children experienced episodes of intestinal intussusception, which in three cases were associated with polyps of the small intestine discovered upon surgical intervention. Total numbers of discrete disease foci per patient were extraordinarily high, ranging from a minimum of 10 to as many as 24. Despite the small number of patients in this group, statistical analysis confirms clinical impressions that they are distinct from those with predisposing LOF mutations. Mean age at first DICER1 syndrome diagnosis was significantly earlier, and both mean and median numbers of disease foci are significantly greater in children with mosaic RNase IIIb mutations (Table 1). The association with juvenile-type intestinal polyps and intussusception is a novel feature of children with mosaic RNase IIIb hotspot mutations, not previously seen in children with other categories of DICER1 mutation.

Figure 3. Numbers and types of disease foci in DICER1 syndrome patients with mosaic RNase IIIb domain hotspot mutations.

For each of the five mosaic hotspot children identified in this study, an individual timeline indicates numbers of discrete foci of neoplastic disease and their histopathological types, graphed with respect to patient age at diagnosis. Across the lower portion of the chart, a single aggregate timeline (dark violet) represents the mean number of disease foci for all PPB/DICER1 syndrome patients with predisposing loss of function (LOF) mutations identified in this study, graphed with respect to patient age at diagnosis. The shaded areas (in lighter violet) surrounding the timeline for LOF mutation patients indicates one and two standard deviations above and below the mean. The range of foci number among all LOF mutation patients was 0 to 6 in all years of age represented (not shown). Abbreviations: CN cystic nephroma; CBME ciliary body medulloepithelioma (eye); NCMH nasal chondromesenchymal hamartoma; PPB pleuropulmonary blastoma, PinB pineoblastoma; SIP small intestinal polyp(s); SLCT Sertoli-Leydig cell tumor (ovary); TCa thyroid carcinoma; TN thyroid nodule(s).

Four of the five children with mosaic DICER1 hotspot mutations presented with cystic PPB (type I/IR) rather than sarcomatous disease (type II or type III) and all five have survived to date. However, this does not indicate a benign clinical course. Though all five hotspot mosaic children are alive, their clinical experiences have been complicated and arduous (Figure 3). Each has undergone multiple major surgeries and chemotherapies with concomitant morbidities. One child is alive with recurrent disease at last follow-up.

Tumor-specific, biallelic DICER1 mutations account for about 10% of PPB cases

In twelve children, we identified biallelic DICER1 mutations present at high allele frequencies in a PPB tumor, but not detectable in blood even with the benefit of high-depth NGS (Table S8). Tumors from these children had an RNase IIIb hotspot missense mutation and either a nonsense LOF mutation (n = 5) or allele loss (n = 7). All twelve children presented with a single PPB tumor and none developed additional foci of disease in the lungs or other organs over the course of subsequent follow-up. This is consistent with occurrence of both an RNase IIIb hotspot mutation and a LOF mutation or allele loss within a single, highly localized clone of somatic cells which then gave rise to the tumor. Absence of additional disease foci is a predictable outcome if both DICER1 mutations are restricted to the initial site of tumorigenesis. However, the absence of additional disease foci among children in this category did not indicate less dangerous disease. Of the 12 patients, 11 had advanced PPB (type II or III), and two succumbed (Table 1).

Currently unresolved cases

Twelve PPB probands in our cohort are negative for predisposing DICER1 mutations detectable in blood DNA by Sanger sequencing or NGS of coding exons. Ten of these children had a single focus of disease, and thus may be sporadic cases involving tumor-specific, biallelic DICER1 mutations, but tumor tissue is either not available or not of sufficient quality for confirmation. Two had multifocal disease, possibly reflecting DICER1 mosaicism that is not well represented in the blood lineage (below our limits of detection). Clinical features of the twelve unresolved cases and the status of further analyses pending or completed, including tumor sequencing, NanoString copy number assay and germline sequencing for additional candidate loci, are summarized in Table S9.

Genotype-phenotype correlation of predisposing mutations in PPB-DICER1 syndrome

All germline DICER1 truncating mutations are predicted to be essentially equivalent in their effect: complete loss of function in miRNA processing. This prediction is based partly on nonsense-mediated decay, but also reflects the functional domain structure of the DICER1 protein. All truncating mutations so far identified in PPB/DICER1 syndrome patients interrupt the open reading frame before the end of the critical RNase IIIb domain (Figure 1, Table S3). Neomorphic RNase IIIb domain function (skewed 5p/3p miRNA production) is a recurring feature of DICER1 tumors, and it is plausible that loss of all wildtype RNase IIIb function is required for it to become tumorigenic. Presumed equivalence of all truncating mutations is consistent with clinical findings: no correlations are apparent between locations of germline truncating mutations within the DICER1 gene and clinical features such as age of onset, number of disease foci, specific tissue sites involved or survival. Non-truncating germline mutations are too rare for correlations with clinical presentations or outcomes to be ascertained.

The natural history of PPB indicates a multistep genetic pathogenesis, and so it is not surprising that in some cases where no germline DICER1 mutation can be detected, one of the two “hits” required for tumorigenesis was acquired during embryogenesis in the form of somatic mosaicism. Mosaic mutations may ultimately prove important in the pathogenesis of many other sporadic childhood neoplasias, as demonstrated recently for retinoblastoma (RB1)⁴³.

Mosaicism for RNase IIIb domain hotspot missense mutations defines a special category of DICER1 syndrome patients that are phenotypically distinct from those who bear germline or mosaic LOF mutations. RNAse IIIb hotspot mutations have not been encountered as inherited alleles in this study or others, which suggests they are inviable^{4,8,10,11,14–16,21,26–30,34,44}. In addition to the five mosaic RNase IIIb hotspot patients in our cohort, three apparently similar cases have been reported (Table 2). Klein et al. described two infants with bilateral Wilms tumor and multiple cysts of the kidneys and lungs⁴⁴. Each child was found to be mosaic for a DICER1 RNase IIIb domain missense mutation, although in one case the mutation was at D1713; also an acidic residue within the RNase IIIb catalytic cleft, but not an established hotspot. De Kock et al. described an infant with pituitary blastoma and bilateral cysts of the kidneys and lungs in whom a de-novo hotspot mutation was detected at high allele frequency in blood as well as tumor¹¹.

Clinically, mosaic hotspot patients are distinguished by two features: i.) consistently early presentations of neoplastic disease, often by one year of age, and ii.) numerous discrete foci of disease developed concurrently or successively, usually involving more than one syndromic tissue/organ site (Figure 3, Table 2). The two features are related and can be interpreted within the conceptual framework of the emerging model for DICER1 syndrome pathogenesis, which provides important insight as to how tumor suppression by DICER1 fails^{6,26,33,35,36}. DICER1 is not a classical tumor suppressor gene for which “two hits” – loss of function in both alleles – are required to allow tumorigenesis. Neither is it haploinsufficient in the usual sense, i.e., that cells with only one expressed allele make wild-type protein, but not in sufficient quantity to fulfill its function. Rather, it is neomorphic function by mutant DICER1 protein, with substitutions of key amino acids in the RNase IIIb domain that causes tumor suppression to falter when it is not masked by expression of wild-type DICER1 protein. Unmasking of an RNase IIIb hotspot mutation may arise through any form of LOF mutation in the wild type allele, including allele loss. The two mutational events, RNase IIIb missense and LOF, may occur in either order and both are generally required to foment the initiation of tumorigenesis. However, as outlined below, RNase IIIb hotspot mutation is a low-probability event and LOF mutation is, relatively, a very high-probability event. The projected consequence of these lopsided probabilities is that occurrence of an RNase IIIb hotspot mutation becomes the rate-limiting step in onset of pathogenesis.

Rationale for the distinctive phenotype of mosaicism for RNase IIIb hotspot mutations

The RNase IIIb domain hotspots in DICER1 are a diminutive mutational target; five codons within an open reading frame of 1922 codons (0.26%). Moreover, molecular mechanisms by which RNase IIIb hotspot missense mutations can arise are restricted to errors of DNA replication and/or DNA repair that produce nucleotide substitution without disturbing the open reading frame. There are 36 possible single-nucleotide changes that can produce amino acid substitutions at these five codons, and only a subset of them has ever been identified in DICER1 syndrome tumors. The spectrum of pathogenic RNase IIIb hotspot mutations is thus very narrow. In contrast, the spectrum of possible LOF mutations is broad and mechanistically diverse. Of the 1922 codons in the DICER1 open reading frame, 675 can be converted to a stop codon by a single nucleotide change. A subset can be converted in more than one way, giving a total of 736 possible single nucleotide changes that result in a nonsense mutation. Among the other 16,562 possible single nucleotide changes in the DICER1 open reading frame, presumably some would be missense mutations that disrupt DICER1 protein function. The five non-hotspot missense mutations we detected as predisposing alleles in PPB probands are likely examples (Figure 2). The individual nucleotides of the DICER1 open reading frame present 5766 point locations at which insertion or deletion of one or a few nucleotides can shift reading frame. An additional 104 bases comprise canonical splice sites of the 26 DICER1 introns, where small sequence changes may result in exon skipping, with or without frameshift. The possibilities for LOF mutations also include larger intra-locus deletions and allele loss through copy-neutral loss of heterozygosity, segmental deletions or complete loss of chromosome 14. Absolute frequencies of these diverse DICER1 mutational mechanisms in a particular cell lineage cannot be modeled precisely, but it becomes clear that the aggregate likelihood of all possible LOF mutations is vastly greater than the likelihood of a neomorphic mutation in one of the five hotspot codons.

It follows that in a developing embryo or child with a germline (or mosaic) DICER1 LOF mutation, “second hits” occurring in a somatic cell will almost always be another LOF mutation, usually resulting in cell death or limited proliferation at most. Rarely, a second hit will be an RNase IIIb hotspot missense mutation, which allows for continuing cell viability and growth, though at the cost of skewed miRNA processing that may ultimately promote tumorigenesis in the surviving clones of cells. However, the low likelihood of incurring an RNase IIIb hotspot missense mutation in somatic cells means that months, years or a lifetime may elapse before one occurs. Further, the developmental context in which a second, hotspot mutation occurs may be important. There are apparently windows of risk for transformation, perhaps coinciding with certain periods of organ/tissue development when an “onco-fetal” gene program is normally active and subject to miRNA modulation, i.e., lung, kidney and brain in the embryo; uterine cervix and ovaries in pubertal girls^1–3,8,9,45. A low probability of RNase IIIb hotspot mutations as second hits during windows of risk may underlie the low penetrance and variable expression of familial LOF mutations in DICER1 syndrome.

For a developing child with a mosaic RNase IIIb hotspot mutation, the prospects are radically different. Somatic cells that bear the RNase IIIb hotspot mutation, masked by a wild type allele, will be viable and non-tumorigenic unless and until they sustain a second hit. However, cells with a preexisting RNase IIIb hotspot mutation are at high aggregate risk of acquiring a subsequent LOF mutation, because it can take any of the myriad forms outlined above. The probability of a secondary LOF mutation occurring during expansion of any given cell lineage over the course of prenatal and postnatal development is relatively high, and independent LOF mutations in multiple lineages may occur. If sufficient fractions of cells in critical lineages are affected, disturbed regulation of developmental gene expression programs arising from defective miRNA processing may be lethal in utero. For surviving embryos, onsets of tumorigenesis will tend to be early and, depending on embryonic distribution of the RNase IIIb hotspot mutation, foci of tumorigenesis may arise in one or more organ sites characteristic of DICER1 syndrome. Additionally, we hypothesize that in mosaic hotspot children, wider tissue/organ distribution of aberrant miRNA processing during development may produce syndromic features not seen in children with predisposing LOF mutations, such as juvenile-type small intestinal polyps, or the generalized somatic overgrowth noted by Klein et al⁴⁴.

Implications for mutation testing, clinical evaluation, and genetic counseling

Recent publications have outlined general recommendations for mutation detection and clinical evaluation for syndromic disease in patients with suspected DICER1 syndrome and family members^9,46–48. Here we add considerations of risk for multifocal disease and reproductive transmission of DICER1 mutations based on mutation category.

Most predisposing DICER1 mutations are germline and detectable by targeted Sanger sequencing from blood. Initial testing should include parents, to distinguish inherited from de novo mutations. Sanger sequencing will usually suffice to detect a parental mutation that is also constitutional, but may fail to detect mosaicism. There is growing appreciation that apparently de novo mutations in children with genetic disease sometimes stem from mosaicism in a parent, which can often be detected by more sensitive methods⁴⁹. For eight patients with apparently de novo mutations in this cohort, we found no evidence of mosaicism in parents by resequencing with high-depth NGS, but this limited finding does not exclude the possibility of parental mosaicism for families evaluated in the future.

For those patients who have a tumor with confirmed DICER1 mutation(s), but test negative for germline mutation by Sanger sequencing from blood, it will be important to distinguish as rigorously as possible between tumor-specific, biallelic mutations and the presence of underlying mosaicism. Mutations confined to the tumor will confer no risk for new foci of primary disease in the proband, and family members including potential offspring will be unaffected. Mosaicism, whether for an LOF mutation or an RNase IIIb hotspot mutation, will confer some degree of risk for additional syndromic neoplasias. It may be impossible to unequivocally rule out mosaicism, but techniques such as targeted resequencing by high depth NGS in multiple tissues can greatly improve diagnostic confidence, particularly with respect to RNase IIIb hotspot mutations. For patients who have more than one focus of disease but no germline or mosaic LOF mutation identifiable by targeted NGS of exons, testing for intragenic deletions or larger genomic alterations is recommended.

Patients carrying mosaic RNase IIIb hotspot mutations are predicted, on the basis of both clinical observations and mechanistic rationale, to have extraordinarily high risk as a group for developing multiple disease foci; approaching 100%. It will not be possible to predict individual risk for multifocal disease by allele frequency in blood, as this will not reveal the extent to which other somatic lineages harbor the mutation. Mosaic RNase IIIb hotspot patients will benefit from the most proactive program of family education and surveillance. The International PPB Registry recommends that potential benefits of renal ultrasound and surveillance chest CT be discussed with the family⁴⁸. The frequency of follow-up chest CTs and chest radiographs should be determined individually, based on patient age, medical history and previous imaging results. Continuing evaluations should include a yearly complete review of systems by a clinician familiar with DICER1 syndrome; yearly screening for ovarian SLCT with review of systems for endocrine dysfunction and pelvic ultrasound for females from early childhood through adulthood; yearly ophthalmologic examination and yearly thyroid examination by palpation and/or ultrasound. Pituitary blastoma and pineoblastoma are rare even in DICER1 syndrome and typically limited to the infant and young child. There is no consensus at this time on screening for intracranial neoplasms.

As prospective parents, patients who are mosaic for a DICER1 mutation face a theoretical risk for transmitting the mutation of up to 50%, depending upon whether and at what frequency it is present in germ cells. For carriers of a mosaic LOF mutation, the consequences of transmission will be similar to those of a germline LOF mutation carrier. For carriers of a mosaic RNase IIIb mutation, it is uncertain whether transmission could result in a live birth. The absence of RNase IIIb hotspot mutations as inherited alleles in all published studies implies they preclude development to term, but this remains speculative. The mosaic hotspot mutation identified in patient 101 of this cohort was discernable in blood by Sanger sequencing and present at 15% of NGS read counts in normal lymph node tissue (Table 2). Similarly in the two Wilms tumor patients reported by Klein et al. and one pituitary blastoma patient described by De Kock et al., de-novo hotspot mutations were readily detected in blood by Sanger sequencing^44,11. Whether the latter case is truly germline or mosaic with high representation in the blood lineage was unclear. Nonetheless, it is clear from these examples that human embryogenesis can tolerate a DICER1 hotspot mutation at high allele frequency in at least some cell lineages. It thus seems possible, though unlikely, that an inherited RNAse IIIb hotspot mutation could be viable.