Clan Genomics: from OMIM phenotypic traits to genes and biology

Abstract

Clinical characterization of a patient phenotype has been the quintessential approach for elucidating a differential diagnosis and a hypothesis to explore a potential clinical diagnosis. This has resulted in a language of medicine and a semantic ontology, with both specialty- and subspecialty-specific lexicons, that can be challenging to translate and interpret. There is no ‘Rosetta Stone’ of clinical medicine such as the genetic code that can assist translation and interpretation of the language of genetics. Nevertheless, the information content embodied within a clinical diagnosis can guide management, therapeutic intervention, and potentially prognostic outlook of disease enabling anticipatory guidance for patients and families.

Clinical genomics is now established firmly in medical practice (Stankiewicz & Lupski, 2020). The granularity and informative content of a personal genome is immense. Yet, we are limited in our utility of much of that personal genome information by the lack of functional characterization of the overwhelming majority of computationally annotated genes in the haploid human reference genome sequence. Whereas DNA and the genetic code have provided a ‘Rosetta Stone’ to translate genetic variant information, clinical medicine and clinical genomics provide the context to understand human biology and disease. A path forward will integrate deep phenotyping, such as available in a clinical synopsis in the Online Mendelian Inheritance in Man (OMIM) entries, with personal genome analyses.

INTRODUCTION

Perhaps the greatest legacy of Victor A. McKusick M.D. (1921-2008), as we retrospectively celebrate his life in medicine and science to commemorate his 100^th birthday, was his recognition and cataloguing of diseases and birth defects as phenotypic traits (McKusick, 1975; McKusick & Amberger, 1993). The implicit understanding of gene action embodied in Online Mendelian Inheritance in Man (OMIM) (https://omim.org), and gene product ‘or mis-action’ underlying the genetic perturbations that distort homeostatic biological mechanisms resulting in disease, is forever with us to build on in our quest to practice better medicine for all stakeholders. Victor as a thought leader, echoing the themes developed in my 2018 McKusick Lecture (Lupski, 2019a), taught us to focus on the patient, the family, the extended family and clan, and study disease traits rather than ‘disease populations’ (McKusick, 1975). That is, rather than a group or ‘cohort of patients’ with a similar disease phenotype ascertained and collated under the category of a disease diagnosis, or entity thought to identify a potentially unified ‘disease process and population of patients’, Victor’s focus was a simple genetic question: what phenotype defines the Mendelian trait? From one perspective, Victor embraced the ‘N = 1 family’ experiment; this N = 1 approach was successful for the first next-generation whole genome sequencing (WGS) approach to molecular diagnosis (Lupski et al., 2010).

I will also now contend that Victor’s foresight may have ‘set the stage’ for the synthesis that facilitated the merger of genetics and genomics as separate scientific disciplines or ‘fields of thought’ which now illuminates human biology and gene and genome evolutionary biology as never before. From a gene and genetic or inheritance perspective, it is also of interest to contemplate what types of mutations, including structural variation, repeat expansions, and complex genomic DNA rearrangements, were: i) NOT delineated by model organism genetic and mutational studies, ii) not anticipated by the heuristic Watson-Crick DNA double helical model, and iii) not elucidated by the ‘Rosetta Stone’ that we all know as the genetic code (Table 1). Yes, human genetics is different than other forms of genetic study – and NOT just because it is done on human subjects by human investigators (Lupski, 2019b).

Table 1.

Human Genome Mutations

a) Structural Variants (SV)

b) Trinucleotide repeat, and other simple sequence repeat (SSR), expansions

c) Complex Genomic Rearrangements (CGR)

d) Multi-locus Pathogenic Variation (MPV)

^*Mutational types NOT delineated by model organism genetic studies

NOR anticipated by the heuristic W-C double helical model for DNA.

In this deferential work, I attempt to capture the incredible excitement of the present AND future in genetics and genomics. Moreover, how guided by Victor’s hand and perhaps by what may have been going on in his head, we all are gaining unprecedented insights into human biology and disease. The synthesis of genetics and genomics through the study of human biology allows us to expand our understanding of mutagenesis, evolutionary biology, and new mutation in sporadic disease traits and birth defects, as well as to foment new knowledge of somatic and germ cell mutagenesis in disorders such as cancer. We can better appreciate the role of mutagenesis at different stages of the organismal life cycle (Liu et al., 2014), investigate systematically Mendelian disease traits and birth defects in multicellular organismal biology; and perhaps even stimulate emerging fields like perizygotic/developmental mutagenesis (Liu et al., 2017), teratogenic mutagenesis (B. Franco et al., 1995; Link et al., 2019), and environmental mutagenesis (Yauk et al., 2015). In this context, the Clan Genomics Hypothesis (Lupski, Belmont, Boerwinkle, & Gibbs, 2011) provides the theoretical framework for a data - generating heuristic approach of family-based genomics and rare variant analyses to the study of disease. Moreover, in addition to research implications and an hypothesis-driven approach to the genetics of disease, Clan Genomics has immediate ramifications for clinical genomics practice (Lupski, Liu, Stankiewicz, Carvalho, & Posey, 2020).

Clan Genomics – a framework for rare variant family-based genomics

In 2011, the Clan Genomics Hypothesis was posited and the complex allelic architecture of human disease summarized formally (Lupski et al., 2011). The implication of Clan Genomics was that recent mutation may have a greater influence on susceptibility to, or protection from, disease than is conferred by variations that arose in distant ancestors. This was conceptually illustrated by a ‘heat map’ in the color shades of the rainbow with the ‘hotter colors’ (red/orange) overlying the siblings in a nuclear family, yellow the parents, and the ‘cooler colors’, e.g., green, showing more distant ancestors in the clan. The rare variants (copy number variant, CNV; single nucleotide variant, SNV; indels) with large effect have arisen recently in the family/clan/population history (Lupski, 2019a, 2019b; Lupski et al., 2011). Therefore, new mutations in you and your recent ancestors, and novel combinations aggregated in one’s personal genome from your parents, account for many medically actionable variant loci.

Clan Genomics provided a framework for a rare variant parsing of genome-wide variant allele data from the assayable portion of individual personal genomes and examining for Mendelian expectations. The hypotheses being tested, rare variant alleles and Mendelian expectations, explores pathogenic variation that might contribute to disease trait manifestation (Figure 1). During the last 10 years, the Clan Genomics hypothesis has been tested worldwide in hundreds of thousands of personal genomes – to date, no data have emerged that warrant rejection of the hypothesis.

FIGURE 1.

Clan genomics and family based rare variant genomic analyses. Different genetic mechanisms for homozygous biallelic variation aggregating at a locus. (a) Family based genomics. Illustrated is a typical pedigree; standard symbols are used with females as circles and males as squares, double horizontal line demarcating consanguinity and filled symbols designating those individuals manifesting trait. Proband only (orange shade background) and quad and trio approach illustrated with background yellow. In accordance with Mendelian expectations for an AR trait, carrier parents with variant alleles at a locus can independently segregate these and in a child at each genetic locus the alleles may become homozygous by aggregating in the personal genome of an affected child. Note if parental consanguinity then homozygosity more likely to occur. New mutations in distant ancestors can be brought to homozygosity through identity-by-descent, IBD (green shade). wt, wildtype or reference allele; DUP, duplication; − / −, mutation with two LoF alleles. (b) Uniparental disomy, UPD, can result in homozygosity for multiple AR trait loci of genes mapping on the same chromosome. It can result in homozygous marker genotypes at a locus when only one parent is a carrier for a variant allele; thus, distorting segregation and Mendelian expectations at loci mapping to the involved chromosome. UPD usually occurs by trisomy to disomy rescue and thus may increase in frequency with maternal age. (c) New mutation deletion CNV at a locus can uncover or ‘unmask’ a recessive trait variant allele: d*, a carrier variant allele inherited from the mother becomes ‘unmasked’ by a de novo deletion CNV mutation at the locus; lightning bolt, new mutation CNV

Gene dosage, alleles, and disease traits

The Charcot-Marie-Tooth type 1A (CMT1A) duplication (Lupski et al., 1991) explained the distal symmetric polyneuropathy (DSP) (England et al., 2009) trait of Charcot-Marie-Tooth (CMT) disease, but studies of its genesis also led to the delineation of: i) the first human genome structural variant (SV) mutagenesis mechanism of non-allelic homologous recombination (NAHR), ii) explained how SV could result in altered interpretation of marker genotypes and ‘mis-mapping’ of a disease locus (Lupski, 2003), iii) helped conceptualize genomic disorders (Lupski, 1998), and iv) formalized a triplosensitive versus gene haploinsufficiency locus for an autosomal dominant (AD) trait. But perhaps the most impactful insights from the locus to the disciplines of genetics and genomics were the documentation of trait manifestation due to gene dosage (Lupski et al., 1992) (Figure 2) - even before the disease driver gene, PMP22, was found. The invention of assaying gene dosage as a biomarker for disease (United States Patent and Trade Office, USPTO#s: 5,306,616; 5,780,223) recognized this unique genetic quality of the disease locus; the triallelic nature rather than a biallelic variant locus one observed with an autosomal recessive (AR) trait locus or a monoallelic variation that could be observed at an AD trait locus (Figures 2 and 3). Moreover, it set the stage for reciprocal recombination by NAHR resulting in the potential for mirror traits as well as delineating new genomic disorders based on this recurrent rearrangement mechanism, i.e. one that occurs repeatedly and a predicted genomic instability locus (Dittwald et al., 2013; Sharp et al., 2006).

FIGURE 2.

CMT, chromosomal syndromes, genomic disorders, Mendelizing disease traits: contributing genomic variation. A heuristic illustration of the multiple genetic and genomic variant ways a patient may have a CMT distal symmetric polyneuropathy (a) Chromosome 17 (Ch17) karyogram is shown with the dosage sensitive PMP22 gene designated by an asterisk. Specific chromosome abnormalities that have been reported with a CMT phenotype of distal symmetric polyneuropathy (DSP) include: direct duplications, inverted duplication, inherited translocation and de novo translocation. All result in three copies of wild type PMP22. (b) CMT is most often due to inherited or de novo submicroscopic genomic duplication (CMT1A; MIM# 118220), but has also been reported with exon deletion CNV, point mutation (Roa, Garcia, Suter, et al., 1993), biallelic PMP22 T118M allele (Roa, Garcia, Pentao, et al., 1993; Shy et al., 2006), Yuan-Harel-Lupski duplication (YUHAL; MIM# 616652) (Yuan et al., 2015), and Multiple de novo CNV (MdnCNV) (Liu et al., 2017); the latter two mutational mechanisms not illustrated here. (c) Studies in CMT genes including PMP22 (Roa, Garcia, Pentao, et al., 1993; Shy et al., 2006) and EGR2 (Warner et al., 1998) revealed both biallelic and monoallelic variants at the locus associated with AR and AD neuropathy traits, respectively. The PMP22 gene (horizontal rectangle) on two chromosome 17 homologues (horizontal lines); PMP22 T118M recessive DSP trait alleles and combinations of alleles at the CMT1A/HNPP locus.

FIGURE 3.

Variant alleles, biallelic AR disease trait genes, triallelic inheritance and mutational burden models. (a) Depicts alleles at a gene locus and a typical biallelic AR trait locus - Stargardt macular dystrophy (STGD1; MIM #248200). Note duplication CNV can result in a triallelic locus and if fully informative may show three distinguishable marker genotypes. Digenic inheritance can have one variant allele at each of two loci OR alternatively be triallelic as described for Bardet Biedl ciliopathy. Note this terminology of ‘triallelic inheritance’ does NOT distinguish which gene/locus has two alleles and which is monoallelic – perhaps a limitation of the nomenclature ‘digenic triallelic’. Moreover, the term is perhaps agnostic to an interpretation of a ‘biallelic AR trait locus and dominant modifier’ (b) Modeling triallelic inheritance in BBS. Note three variant (i.e. mutant) alleles at two BBS loci may be required for BBS manifestations, i.e. penetrance, in some families. (c) A family with BBS. Pedigree analysis and Sanger sequencing of BBS loci documented the mutational burden required in some families for trait manifestation. Evidence for such a potential model has been confirmed with the elucidation and study of additional ‘BBS associated genes’ and variant allele mutation types (Lupski, 2003).

Genomic disorders integrated the categories of Mendelian disease and chromosomal syndrome (McKusick, 2007), and even perhaps copy number variation (CNV) and SNV/SNP in common disease/complex traits, into a context of genome variation(s) contributing to disease (Figures 2 and 3). That is, disease from the perspective of biological perturbations from homeostasis or distorting ‘biology in balance’. Copy number, whether of chromosomes, genes, or regulatory regions is all about biology in balance.

The genetic description of triallelic inheritance in Bardet Biedl Syndrome (BBS; MIM# 209900) (Katsanis et al., 2001) and other ciliopathies, described purely by pedigree analysis and Sanger di-deoxy sequencing of genes, conceptually may perhaps be thought of as a form of digenic inheritance and perhaps some of the first genetic evidence providing insights into mutation burden models and the idea of oligogenic inheritance and/or multi-locus pathogenic variation (MPV) in patients or families versus polygenic inheritance in populations. When experimentally expanding BBS studies to capture exon-disruptive CNV, some 18.5% of 92 unrelated subjects with BBS show evidence for such a third potentially pathogenic ‘BBS gene allele’ (Lindstrand et al., 2016). Nowadays, leveraged by omim.org, and rare variant family-based genomics approaches (Posey et al., 2019), the functional annotation of genes and variant alleles contributing to perturbations from biological homeostasis resulting in disease would surely make Victor proud. Indeed, it is interesting to reflect on just how far human and medical genetics has come – Victor described only 738 conditions in the 1993 paper (McKusick & Amberger, 1993). In the ‘post-genomics era’, family-based rare variant genomics approaches are also providing experimental evidence for mutational burden models (Bis-Brewer et al., 2020; Gonzaga-Jauregui et al., 2015).

With data from more and more family-based genomics studies becoming available, it perhaps also is becoming more conceptually crystallized that there is really no such thing as a ‘recessive gene’ or ‘dominant gene’, but rather AR traits or AD traits. From this perspective it is ALL about variant alleles (Figure 3) and mutation STRENGTH, i.e. magnitude of effect; whether null, hypomorph, hypermorph, antimorph, or neomorph alleles (Muller, 1932). Whether loss-of-function (LoF) or gain-of-function (GoF) is important!

If biallelic variant alleles are required for trait manifestation, a Mendelian expectation of AR trait segregation will be fulfilled. Moreover, as documented with retinal disease and the ABCA4 gene locus, the AR Stargardt macular dystrophy (STGD; MIM #248200) can depend on the selected combination of alleles. This age dependent ‘trait penetrance’, as measured by decade of disease onset, is readily observed in a family (i.e. N = 1) given the high incidence of pseudodominance in the population (Lewis et al., 1999). Moreover, in an extended family or clan, whether the patient manifests STGD or retinitis pigmentosa (RP), can be influenced to a great extent by the selected combination of null alleles (Shroyer, Lewis, Yatsenko, & Lupski, 2001). Note that in the original 1997 description of the Stargardt macular dystrophy gene, ABCR renamed ABCA4, despite the study of many dozens to hundreds of families no ‘double null’, i.e. severe LoF combination of alleles, was ever observed; a ‘finding’ prompting speculation that such combinations might result in a distinct visual impairment recessive disease trait (Allikmets et al., 1997) subsequently shown to be RP.

New mutation and disease

From the perspectives of both genetics and genomics, the data from clinical genomics have provided tremendous insights about new mutation and disease – here I use the term ‘clinical genomics’ in its broadest sense of the meaning as defined in 2016 (Lupski, 2016) - personal genome variation contributing to distorting biology from balance. From the studies of the initial ~10,000 to 50,000 personal genomes of patients investigated by chromosomal microarray (CMA) (Harel & Lupski, 2018; Lupski, 2009), of those molecularly diagnosed by genomic disorder associated CNV, particularly in sporadic developmental delay/intellectual disability (DD/ID) traits (Lupski, 2007), it was essentially ALWAYS due to de novo mutation CNV. Moreover, it was NOT due to a common variant structural variation or inherited copy number neutral SV or CNV. Even in the large families with AD CMT eventually shown to be due to the CMT1A duplication, one could trace the new mutation in the pedigree and family or clan, to an individual in an antecedent generation (Patel et al., 1990).

The first ~ 2,000-3,000 clinical exomes (cES) (Y. Yang et al., 2014) showed remarkably that, of the 242 molecularly diagnosed AD disease trait cases, i.e. those conditions with monoallelic variants, 86% were de novo, 1% were inherited from a mosaic parent, and 1% were inherited, but the latter at an imprinted gene locus! As expected from this primarily outbred population patient cohort that was studied, the majority of the 182 AR disease trait biallelic variant gene/loci (104/182, 57%) were compound heterozygotes and 33% (N=59 families) were homozygous. Perhaps NOT expected was the observation of 2% compound heterozygotes due to SNV + CNV (N = 4)! Even more startling, the 33% homozygotes AND 3% of these being homozygotes due to uniparental disomy (N = 5) where only one parent was a carrier; trisomy to disomy rescue may be more frequent in clinical cases than initially thought (Spence et al., 1988).

From studies on the initial ~8,000 cES about 5% of molecularly diagnosed cases had multi-locus pathogenic variation (MPV) – notably, and certainly NOT anticipated by any mutational theory, BOTH variants were de novo in 44.7% of cases diagnosed by monoallelic variants in each of two loci (Posey et al., 2017). In relatively common genetically heterogeneous conditions such as DD/ID, it has been well established that in outbred populations new mutations play a major role (Lupski, 2010). But even in the founder population of Finland, one study in which 27 unrelated probands had variants in known ID genes, 75% were de novo alleles (Jarvela et al., 2021). Perhaps as we learn how to characterize better these cognitive phenotypes, and rare variant smaller sized CNV, e.g. < 100 Kb, additional potentially contributing variant alleles will be elucidated (Lupski, 2015a; Mannik et al., 2015). Identifying rare SVs has provided insights into neuropsychiatric disease (Lupski, 2008); finding a potentially medically actionable SV in a personal genome can prompt management and therapeutic decisions (Bodkin et al., 2019). Indeed SV, in particular reciprocal duplication and deletion genomic disorders, provided some of the first insights into the mirror traits of microcephaly and macrocephaly and phenotypic extremes from neurotypical behaviors (reviewed in (Lupski, 2015b)).

If a new mutation occurs during development, it can result in mosaicism (Campbell, Shaw, Stankiewicz, & Lupski, 2015; Lupski, 2013). Germ cell mosaicism can result in a recurrence greater than the ~1% often cited. Recurrence risk may be elevated higher if an apparent de novo pathogenic variant in a sporadic affected child is ‘phased’ to the maternally inherited allele (Campbell et al., 2015; Campbell, Stewart, et al., 2014; Campbell, Yuan, et al., 2014). Low-level parental somatic mosaicism may be more frequent than currently recognized (Gambin et al., 2020) and may have significant implications when genetic counseling is warranted.

Mutation rates, chromosomal syndromes, genomic disorders & Mendelian disease traits

It follows that if new mutations contribute to disease, the mutation rates can affect the frequency with which one observes particular rare variant alleles or combinations of alleles at a locus, contributing to conditions or syndromes (Table 2). Studies from genomic disorders found that intergenerational locus specific mutation rates for genomic disorder associated CNV were two to three orders of magnitude greater than point mutation (Lupski, 2007). Whereas point mutations occur at a frequency of ~1 – 2 X 10⁻⁸ per locus per generation (Kong et al., 2012; Rahbari et al., 2016; Sun et al., 2012), de novo CNV mutations associated with genomic disorders occur at a rate of ~ 10⁻⁴ to 10⁻⁶. Of note, by sperm PCR typing studies for the SV mutagenesis events that occur recurrently by NAHR at a locus, de novo mutations occur at a frequency of ~ 10⁻⁵ to 10⁻⁶; deletion CNV occur twice as frequently as duplication CNV (Turner et al., 2008). PRDM9 variation strongly influences homologous recombination/NAHR hotspot activity and meiotic instability in humans. While different de novo mutation rates at the hereditary neuropathy with liability to pressure palsies (HNPP) deletion/CMT1A duplication locus can be detected to occur by sperm PCR in males with different PRDM9 alleles, the de novo deletion/duplication ~ 2:1 ratio remains (Berg et al., 2010). Importantly, these observations suggest HR rates can differ in individuals from different populations.

Table 2.

Mutation rates and disease

Condition/syndrome/Dz	variant type	de novo per locus mutation rates
Mendelian disease trait	SNV	10−8
Genomic disorder	CNV	10−5 to 10−6
Chromosome syndrome	aneuploidy	10−3

The high mutation rates of SV mutagenesis at recurrent genomic deletion regions may make these loci a significant contributor to AR disease trait loci in populations due to a compound heterozygous CNV + SNV allelic combination (Yuan et al., 2021; Yuan et al., 2020). New mutation deletion CNV may allow an ‘entry point’ into many recessive disease trait genes/loci and even potentially enable a ‘haploid genetic/genomics approach’ to the functional biology of the human genome (Yuan et al., 2021). In some respects, the ‘autozygome’ reflects new mutation variation in antecedent generations of the family or clan (Khalak et al., 2012). Human gene ‘knockouts’ from autozygome studies have also enabled medical annotation of the human genome and informed the interpretation of GWAS of complex traits (Maddirevula et al., 2018).

Gene action: ascribing function to computationally annotated genome transcripts

The total number of genes annotated computationally in the human genome is about 20,000 – 25,000 according to (https://web.ornl.gov/sci/techresources/Human_Genome/project/). The vast majority remain to be ascribed a function; however, the rates of functionally characterizing human genes continue to soar, the slope of the rate is at a steady state, and there is NO EVIDENCE of a ‘flattening of the curve’ (Posey et al., 2019). Given the rapid accrual of knowledge in human gene function and genomic medicine, it is notable that reanalysis of extant cES data can assist molecular diagnosis and gene discovery (P. Liu et al., 2019). It is quite remarkable how, like omim.org, GeneMatcher (https://genematcher.org/) (Sobreira, Schiettecatte, Valle, & Hamosh, 2015) and the Matchmaker Exchange (Azzariti & Hamosh, 2020), further leveraged by DECIPHER (https://www.deciphergenomics.org/) (Firth et al., 2009), have connected virtually the entire world of rare disease physicians, researchers, families, and all stakeholders in the pursuit of understanding disease. This global community of humanity is perhaps a living and maturing ‘clinical reality’ catalyzed by the Human Genome Project; and in the interpretation of the heuristic data-driven models (Posey et al., 2019).

Having a DRAFT haploid reference genome enabled the characterizations of deviations from the human diploid state, i.e. CNV, observed in genomic disorders (Lupski, 2009). Some investigators have argued the delineation of genomic disorders and clinical application of chromosomal microarrays analysis (CMA) was the most unanticipated medical benefit of the human genome project! These CNV in personal genomes could affect gene action of one, two or many genes and elucidate mutational mechanisms affecting DNA/genome structure as opposed to the Watson-Crick base pair changes and gene mutations underlying many different Mendelian traits in model organisms and Homo sapiens.

SV mutagenesis mechanisms

Three major mutational mechanisms underlie DNA SV mutagenesis resulting in CNV. These include: i) non-homologous end joining (NHEJ), ii) NAHR, and iii) the replication-based mechanism which incorporated the idea of DNA replication and ‘template switching (TS)’ - microhomology mediated break induced recombination (MMBIR). All three, homologous (NAHR) and nonhomologous/microhomologous (NHEJ and MMBIR) are important to genomic DNA rearrangements and evolution of the human genome.

The NAHR mechanism underlies many recurrent genomic disorders and chromosomal rearrangements (Table 3). Interestingly, gene family loci and loci with genes/pseudogenes in ‘linked proximity’ seem particularly susceptible to genomic instability, where the different genes can act as NAHR substrates to generate deletion and duplication CNV ‘alleles’. This has been nicely elucidated at the ATAD3A locus with SV alleles causing a mitochondrial disease (Gunning et al., 2020; Harel et al., 2016) with cholesterol metabolism perturbations. The high frequency of such alleles also results in many combinations of CNV and SNV at that locus (Yap et al., 2021).

Table 3.

NAHR and chromosome abnormalities in constitutional & cancer genomes

Chromosome Abnormality#	Genome Architecture
(a) microdeletion	direct repeats
(b) microduplication	direct repeats
(c) inversions	inverted repeats
(d) recurrent translocations	olfactory receptor gene clusters
(e) isodicentric chromosomes*	non-sister chromatid exchange inverted repeat substrates

^#genomic NOT cellular (e.g. aneuploidy) mutational mechanism

^*iso17q (isodicentric17q) – most common chromosomal abnormalities in cancer

References (a) (Chen et al., 1997); (b) (Potocki et al., 2000); (c) (Giglio et al., 2001; Lakich, Kazazian, Antonarakis, & Gitschier, 1993); (d) (Ou et al., 2011); (e) (Barbouti et al., 2004)

The replication mechanism for SV mutagenesis (Lee, Carvalho, & Lupski, 2007), Fork Stalling Template Switching, or FoSTeS/MMBIR, is a means to repair one-ended, double-stranded, DNA, i.e. oeDNA, generated at collapsed forks or telomere ends (Lowden, Flibotte, Moerman, & Ahmed, 2011; Yatsenko et al., 2012). FoSTeS/MMBIR appears to be important in: i) nonrecurrent genomic rearrangements, ii) generating complex genomic rearrangements (CGR) due to two or more template switches (TS), iii) can be the mechanism for chromosomal insertional translocations, iv) underlie primate genome evolution, v) be responsible for hypermutation at a locus, vi) can result in exon shuffling, and vii) even be involved in evolving new genes (Beck et al., 2019; Carvalho et al., 2013; Carvalho, Zhang, & Lupski, 2010; Gu et al., 2016; F. Zhang et al., 2009; Zuccherato, Alleva, Whiters, Carvalho, & Lupski, 2016). Evolution as a ‘school of thought’ plays an important role in the teaching and practice of medicine (Nesse et al., 2010); a theme echoed throughout the 2009 Sackler colloquium marking the 200^th anniversary of Charles Darwin’s birth and 150^th anniversary of Origins of Species.

Perhaps a fourth emerging mutational mechanism, Alu-Alu mediated rearrangement (AAMR) (Lupski, 2019a; Song et al., 2018) is a special type of MMBIR mediated by a template switch (TS after oeDNA, break repair at collapsed DNA replication forks) between ~300 bp Alu repetitive elements that share substantive microhomology (Mayle et al., 2015). AAMR can cause exon shuffling and is important to gene evolution and the evolution of disease-associated alleles. From a rare disease perspective, the AAMR mutational mechanism is important because it can result in intragenic exonic deletion and duplication CNV; particularly CNV of size 100Kb > x > 100bp. From a mutational mechanism standpoint, it is also interesting that the microhomology/microhomeology required for TS and replicative recombination, such as MMBIR, of oeDNA is different from double strand break repair (DSBR) (Bahrambeigi et al., 2019).

Machine learning and supercomputer modeling of direct Alu-Alu repetitive sequence element studies of the human genome and gene structure show that genes have differing relative risk for susceptibilities to genomic instability mediated by AAMR (Song et al., 2018) as determined by AluAluCNVpredictor (http://alualucnvpredictor.research.bcm.edu:3838/). Exonic deletion CNV (Dharmadhikari et al., 2019; Duan et al., 2021; Hengel et al., 2021; Mitani et al., 2021) and exonic duplication CNV (Okamoto et al., 2014) can be important to biallelic AR disease traits. Thus, AAMR can drive new mutation generating AR trait ‘carrier state’ alleles. Furthermore, in this important N = 1 clinical case of an Iraqi family, such SLC13A5 deletion CNV alleles can bring clarity to eponymic clinical diagnoses and the genetic heterogeneity potentially underlying such disease traits as Kohlschutter-Tonz Syndrome (KTZS; MIM #226750) and Developmental and Epileptic Encephalopathy 25 (DEE25; MIM #615905) (Duan et al., 2021).

Interestingly, the finding of the NDRG1 exonic duplication CNV in a Turkish family (Okamoto et al., 2014) was perhaps the first evidence that this gene was something more than a ‘Romanian gypsy neuropathy’ gene (CMT4D; MIM: #601455) and could be an important contributor to CMT distal symmetric polyneuropathy (DSP) worldwide - and perhaps also influenced my thinking about the derivation of Founder alleles and Steel syndrome in the US Puerto Rican population versus Clan Genomics homozygosity and AR traits (Gonzaga-Jauregui et al., 2020). Could the condition described in the US Native American population referred to as ‘Navajo Neuropathy’ in some families be due to a CNV allele of yet another CMT gene (El-Hattab et al., 2018) – perhaps other than MPV17 (CMTEE; MIM# 618400)?

The FoSTeS/MMBIR mechanism has taught us that chromosomal, genomic, and genic rearrangements can be quite complex (Carvalho & Lupski, 2016; F. Zhang et al., 2009). A particular type of CGR SV mutagenesis DNA rearrangement end-product is worth noting; a ‘nested’ duplication-triplication-duplication, i.e. DUP-TRP-DUP, resulting from only two TS. This DUP-TRP-DUP structure (Figure 4) can triplicate dosage sensitive genes like PMP22 (F. Zhang et al., 2009), LIS1 (Bi et al., 2009), MECP2 (Carvalho et al., 2011), and PLP1 (Beck et al., 2015).

FIGURE 4.

Complex Genomic Rearrangements: DUP-TRP-DUP a mutational signature of constitutional and cancer genomes. (a) Patterns observed for duplication – triplication/inversion - duplication, DUP-TRP/INV-DUP on ChX. Array Comparative Genomic Hybridization (aCGH) pattern showing a structural variant copy number pattern DUP-TRP-DUP complex genomic rearrangement (CGR) with multiple copy number transition states, but only two recombinant junctions reflecting two template switches (TS). Red dots represent gain at that specific interrogating oligonucleotide with the extent of the deflection (log ratio of signal) reflecting the copy number gain at the locus: DUP, duplication; TRP, triplication. Black normal diploid copy number. Red horizontal line duplicated genomic interval; blue horizontal line triplicated genomic interval. Below, interpretation of array data and the idealized interpretation of the copy number changes followed by the actual arrangement of the DNA sequence structure (note a, a’ and c, c’ duplicated whilst segment b, b’ triplicated) in DUP-TRP-DUP with two recombinant junctions, jct1 and jct2, shown. (b) Example autosomal genes triplicated by DUP-TRP-DUP; three different chromosome loci and genes triplicated by DUP-TRP-DUP are shown. Right column shows disease trait observed. (c) Disease traits associated with gene triplication. Note the triplication of the alpha synuclein locus SCNA and that of the gene PMP22 in adult-onset Parkinson disease and Charcot-Marie-Tooth disease, respectively. LIS1 (PAFAH1B1) is the gene deleted in 17p13.3 in association with isolated lissencephaly (MIM #607432) (d) The genomic sequence (GS) interpretation of the DUP-TRP-DUP mutational signature in cancer genomes. DUP-TRP-DUP as a mutational signature of SV mutagenesis in the cancer genome was described in 2020 (Li et al., 2020). Interpretive key with relative dosage versus ‘diploid state’ of autosomes. Note duplication of one chromosome, but three copies of locus due to the presence of the other ‘non-rearranged’ autosomal chromosome (Riccardi & Lupski, 2013). Note general CGR structure of DUP-TRP-DUP and the multiple possible orientations of genomic segments that can be derived from just ‘two jumps’; i,e, two TS, multiple possibilities. Right column phenotype observed with somatic mutagenesis; DUP-TRP-DUP is a frequent mutational signature of SV mutagenesis in cancer genomes.

DUP-TRP/INV-DUP has been delineated recently as the rearrangement end product underlying the alpha synuclein gene, SCNA, triplication causing Parkinson disease (Robak et al., 2020; Zafar et al., 2018). This Parkinson associated triplication was described in 2003 (Singleton et al., 2003), However, the mechanism for its de novo mutational genesis has been described just recently almost 20 years later! Indeed, studies on Parkinson-associated SCNA multiplication mutations from Korea show these arise de novo from a microhomology-mediated replicative repair (MMBIR) SV mutagenesis mechanism with each representing a new mutation in ancestors of the clan (Seo et al., 2020). Remarkably, DUP-TRP-DUP was shown to be one of the 12 most common SV mutagenesis ‘signatures’ observed in cancer genomes (Figure 4) (Li et al., 2020)! Genomic intervals with enriched Alu repetitive elements may be more prone to genomic instability and DUP-TRP-DUP (Gu et al., 2015) (Figure 5). DUP-TRP/INV-DUP can be driven by TS at inverted Alu given 5’ to 3’ direction and ‘polarity of priming’ and ssDNA.

FIGURE 5.

Copy number and orientation in Complex Genomic Rearrangements. (a) the proposed model for derivation of DUP-NML-DUP; array CGH as shown in Figure 4. Duplicated genomic segments shown depicted by red horizontal bars. Alu element orientation and direction depicted by arrowheads; filled arrowhead color shows Alu family member according to KEY. Note ‘directionality’ of replicative repair due to TS. (b) Derivation of DUP-TRP/INV-DUP. Note only two TS, but multiple possibilities, Moreover, whether a resultant DUP-NML-DUP or DUP-TRP-DUP there are only two TS postulated and only two breakpoint junctions found. CGR structure also influenced by whether TS occurs between chromatids or chromosome homologues (Carvalho & Lupski, 2016; Carvalho et al., 2015). The orientation of the TS can result in genomic inversion. Whether the TS occurs before or after the migrating replication fork can affect ultimate rearrangement end product outcome and even potentially lead to a rolling circle mechanism and resultant genomic segment amplification (Hastings, Ira, & Lupski, 2009).

The same mutational mechanisms drive cancer genomes and constitutional genomic rearrangements associated with genomic disorders and Mendelian conditions (Table 3 and Figure 4). From this perspective, the most common cancer associated chromosomal abnormality, i17q, is actually an isodicentric 17q driven by NAHR (Barbouti et al., 2004) and non-sister chromatid exchange of an inverted low-copy repeat (LCR) in 17p11.2. The complexity of genomic changes observed in cancer and in constitutional genomes can be extensive as evidenced by the phenomenon, of chromothripsis and chromoanasynthesis (Liu et al., 2011; Maher & Wilson, 2012; Stephens et al., 2011) and what has been more recently defined as chromoanagenesis (Crasta et al., 2012). Intriguingly the organismal phenotype with such chromothripsis-like changes might be only a Mendelian condition such as Cornelia de Lange Syndrome (CDLS1; MIM #122470) or Coffin Siris Syndrome (CSS1; MIM #135900) and due to interruption of noncoding gene sequences, NIPL and ARID1B respectively, and thus NOT revealed by cES (Grochowski et al., 2021; Plesser Duvdevani et al., 2020). Also, From the mutational mechanism perspective it is quite intriguing to observe the types of complex genomic DNA rearrangements, including chromothripsis, that have been reported in association with CRISPR-Cas9 gene editing (Kosicki, Tomberg, & Bradley, 2018; Leibowitz et al., 2021; Simeonov et al., 2019)

Birth defects and the compound inheritance gene dosage model

Recent studies show that combinations of alleles can have profound consequences for developmental abnormalities and may be particularly impactful in the etiology of sporadic birth defects. The genetic model, formalized in the compound inheritance gene dosage model (Figure 6), pairs a null allele at a locus with a hypomorphic noncoding variant allele in trans. This can result in gene action and expression slightly below haploinsufficiency that may occur during a specific developmental period when cell morphogen gradients may play an important role. Interesting how gene dosage (gene copy number) and gene dosage (gene expression) can be challenging to experimentally differentiate from a molecular standpoint. Note a chromosome rearrangement ‘position effect’ is often molecularly related to a gene dosage (expression) alteration. The ‘compound inheritance gene dosage (CIGD)’ model takes gene dosage/expression windows to an entirely new level of illumination regarding subtle perturbations of gene expression during development and the potential molecular underpinnings of penetrance and position effects.

FIGURE 6.

Birth defects and the compound inheritance gene dosage model. (a) An AD trait may result from a monoallelic null variant allele at a locus resulting in haploinsufficiency. Biallelic null variants may result in lethality due to complete LoF at the locus. Some birth defects may result from gene action below haploinsufficiency, but not complete loss of gene function. (b) Illustrates function and thresholds for phenotypic expression. Note homozygosity for the hypomorphic allele results in no abnormal phenotype. (c) In the compound inheritance gene dosage (CIGD) model, a rare variant null allele at a locus is paired with a common variant noncoding allele; to result in gene action/expression below haploinsufficiency; TBX6 associated congenital scoliosis (TACS) results from hemivertebrae usually involving the lumbar spine (J. Liu et al., 2019). Interestingly, when a non-coding ‘up-expression’ allele, i.e. a hypermorph, is paired with a null allele the vertebral anomalies localize to the cervical spine (Ren et al., 2020). (d) Penetrance of CAKUT, congenital anomalies of the kidney and urinary tract and unilateral renal agenesis, can be related to gene action and gene dosage/expression at the TBX6 locus.

At the TBX6 locus (Figure 6), this compound inheritance gene dosage model accounts for 12-15% of all operable congenital scoliosis in China (Wu et al., 2015). Interestingly, the TBX6/Tbx6 associated congenital scoliosis (TACS) results in hemivertebrae and butterfly vertebrae in the lumbar spine (J. Liu et al., 2019; N. Yang et al., 2019). Duplication and overexpression of TBX6, resulting in a hypermorphic allele, causes spine vertebral defects localized to the cervical versus lumbar spine (Ren et al., 2020). Of note, mutations in another gene in this gradient dependent somitogenesis NOTCH/TBX6 gene cascade, RIPPLY2, cause cervical vertebral anomalies and a Klippel-Feil AR disease trait phenotype (Karaca et al., 2015).

Compound inheritance and gene dosage/expression underlie developmental disorders of the kidney, including congenital anomalies of the kidney and urinary tract (CAKUT) and renal agenesis (N. Yang et al., 2020), and of the lung (Karolak et al., 2019). Interestingly, this CIGD model may be particularly relevant to developmental defects in organs that lie on polar opposites of the body plane axis; perhaps stochastic effects of transcriptional regulatory developmental cascades play a role? For CAKUT this genetic model of compound inheritance connects GWAS signals to driver genes in recurrent deletions associated with genomics disorders (Verbitsky et al., 2019). The genetic model may enable clinical interpretation of medically actionable non-coding variants, which can be ‘common variant alleles’ that may have been recognized by GWAS studies.

A pragmatic, and potentially clinically useful molecular diagnostic approach might consider using ‘exome capture’ of long read rather than short read sequencing to build out gene structure and evaluate for potential pathogenic variation in noncoding regulatory sequences and introns. The medically actionable focus is thus on the gene – entire gene structure and regulation, NOT just coding sequence and variants that might affect coding information that are assayed by cES.

Multilocus pathogenic variation & aggregation of personal genome mutational burden

Multilocus pathogenic variation (MPV) is the aggregation of two or more molecular diagnoses, i.e. pathogenic variants at two or more gene loci, resulting in a blended phenotype due to a combination of disease traits associated with each individual gene/locus (Posey et al., 2017). Blended phenotypes may present as two clinically distinguishable traits, distinct blended trait phenotypes, or overlapping disease trait blended phenotypes; the latter often challenging to distinguish from phenotypic expansion of variation at a single locus (Posey et al., 2017). Whether or not certain combinations of MPV genes also result in epistatic interaction and synthetic trait/phenotypes remains to be explored.

When multilocus pathogenic variation results in a dual molecular diagnosis, it can be for an AD + AD trait, AD + AR trait, AD + XL trait, AR + AR, etc., trait combinations. If ONE or BOTH are monoallelic loci, e.g. AD + AR, AD + AD, then the aggregation in a patient genome might be from combinations of de novo mutation and inherited variant alleles (Figure 7). AR + AR combinations are rarely observed in outbred populations, but in admixed populations with an increased coefficient of consanguinity the AR + AR combinations can dramatically increase in frequency from Clan Genomics (Lupski, 2019a).This occurs because of long track de novo haplotypes resulting in either large-sized ROH (runs of homozygosity) brought to homozygosity by identity-by-decent (IBD) of linked MPV loci or distributive ROH for unlinked MPV. This has been demonstrated clearly for the arthrogryposis phenotype (Bayram et al., 2016; Pehlivan et al., 2019), and now for the neurodevelopmental disorders (NDD) phenotype (Mitani et al., 2021) in the admixed Turkish population.

FIGURE 7.

Aggregation of multilocus pathogenic variation in a personal genome. (a): interphase FISH molecular diagnostics of BAB1006 (Potocki et al., 1999) revealing a loss of the PMP22 gene (red) on chromosome 17 due to HNPP deletion with a gain of the RAI1 gene (two green dots) on the other chromosome 17 homologue due to PTLS duplication. (b) Horizontal karyogram of Chr17. Note cytogenetic map position of PMP22 (red) in 17p12 and RAI1 (green) in 17p11.2. (c) Southern analysis of three generations of family HOU365. Note the HNPP deletion, as evidenced by the Southern blot visualized HNPP rearrangement junction fragment (HNPP del jct; orange), that is inherited for at least three generations in individuals with operative AD carpal tunnel syndrome. (d) The PTLS duplication (PTLS dup) occurs de novo as evidenced by the appearance of the pulsed field gel electrophoresis (PFGE) junction fragment (jct; orange).

The study of disease in admixed populations reveals that the higher estimated coefficient of inbreeding (F) values manifest in an increased genome-wide burden of long-sized (>3 Mb) regions of homozygosity (ROHs), inferring ‘young haplotypes’, derived as de novo variant alleles in antecedent generations of the clan (Coban Akdemir et al., 2021). In some respects, such ‘young haplotypes’ represent copy number-neutral SV mutagenesis. These ROHs were shown to be enriched for ultra-rare, multi-locus, homozygous, deleterious variants and wherein their combinatorial effects can produce blended phenotypes accounting for the observed disease. These data support a Clan Genomics model for disease in a population. This unique combination of rare variants located in ROH regions in each individual genome characteristic of their recent family lineage, pedigree, or clan may explain their clinical phenotypic features and thereby their pathogenicity, providing further support for the Clan Genomics hypothesis and potentially illuminating La Reunion paradox (Beckmann, 1996; Zlotogora, Gieselmann, & Bach, 1996) in human genetics.

The concept of ‘Contiguous Gene Syndromes’, as formalized by Schmickel (Schmickel, 1986) in his Journal of Pediatrics paper, is potentially another example of MPV thinking. Perhaps ‘patient BB’ crystalized this thought for the ‘cytogenetics world’ when a syndromic patient appeared to have a phenotype that ‘blended’ multiple X-linked disease traits (Francke, 2013). Subsequently several of these mapped genes for the XL traits were found to be physically mapped together in Xp21.1 (Figure 8a). The concept was further generalized by the elucidation of the Yuan-Harel-Lupski (YUHAL; MIM #616652) syndrome and two dosage sensitive genes, for triplosensitive loci, ‘physically linked’ by submicroscopic nonrecurrent duplication CNV encompassing both (Figure 8b). Perhaps YUHAL is better thought of as a ‘multi-genic trait’? Finally, the 9q34.11 deletion CNV ‘map’ extends the thinking even more (Figure 8c). It is amazing to think that MUCH of this medical genetic thinking took place BEFORE there was a decent draft of the haploid reference genome.

FIGURE 8.

Contiguous gene syndromes, CNV, genes and multigenic disease traits. (a) Three physically linked genes: DMD, CGDX, and MCLD and a cytogenetically visible deletion of Xp21.1 in patient BB (Francke, 2013). Below and horizontally is chromosome X karyogram with an expansion of the Xp21.1 locus and physically linked genes for the condition mapped within the deletion interval. (b) The Yuan-Harel-Lupski syndrome (Yuan et al., 2015) (YUHAL; MIM #616652) is caused by, at least, a blended phenotype consisting of two triplosensitive genes, PMP22 and RAI1, linked on one nonrecurrent duplication rearrangement involving chromosome 17, dup17p11.2p12 (red horizontal line); thus, YUHAL could be considered as a multigenic trait. Above chromosome 17 karyogram with physical ‘genomic’ map of 17p12p11.2. Note the green (HNPP del), white, red (PTLS dup) for the personal genome (boxed purple) of BAB#1006; family HOU365 in figure 6. (c) Four physically linked genes, STXBP1, SPTAN1, ENG and TOR1A may be individually or combinatorically deleted in chromosome del9q34.11 syndrome (Campbell et al., 2012).

Quantitative phenotyping to investigate genes and biology

Moving forward, robust clinical phenotyping will continue to play an important central role in characterizing a patient’s disease process and in generating a differential diagnosis, as well as providing supportive evidence for exploring a potential molecular diagnosis. The information content inherent in clinical phenotyping may perhaps be used more and more in a quantitative rather than a qualitative way to investigate disease biology and for gene discovery research. Here, omim.org and the clinical synopsis of each entry translated to HPO terms, are very powerful as they provide information content to explore traits and the biology of disease (Amberger, Bocchini, Scott, & Hamosh, 2019). The Human Phenotype Ontology (HPO) enables semantic unification of common and rare disease (Groza et al., 2015). HPO is a structured hierarchical ontology that is amenable to quantitative computational analyses and phenotype comparisons via similarity matrices. Leveraging information in omim.org, particularly disease trait information embodied in the clinical synopsis of each entry, allows a quantitative phenotyping approach to disease biology and assessment of potential ‘contributing genes.’

Two semantic similarity approaches used with some success are the Resnik (Resnik, 1995) and the Lin (Lin, 1998) methods. Semantic similarity scores are to natural language processing, a promising application of machine learning (ML) and artificial intelligence (AI) in genomics, similar to what addition is to algebra. The Resnik method has been used to investigate the biological underpinnings of blended traits in MPV (Posey et al., 2017). It has also been used to leverage phenotypic information by conditioning a stringent variant calling pipeline of cES reanalysis and clinical molecular diagnosis of known, OMIM defined, genes (P. Liu et al., 2019). Such foundational phenotype analyses highlight how the highest quality base, phenotypic data in OMIM can be leveraged to understand the genetic contribution to all phenotypes as our computational approaches evolve to include ML and AI. Below, by way of illustration, four areas are highlighted for further exploration. These include: i) MPV and blended traits, ii) leveraging mouse KOMP (Knockout Mouse Project; https://commonfund.nih.gov/komp2) data (GOtoHPO) for gene/disease trait association, iii) hypermutation/gene dosage and mirror traits, and iv) genetic heterogeneity and genotype/phenotype correlation of an individual gene allelic series. To be clear, many challenges still remain.

The Lin method incorporates information content from the hierarchical representation of the common ancestor of shared terms and allows finer gradations of similarity score. Thus, where ‘Microcephaly’ (HP:0000252) would be considered distinctly different from ‘Macrocephaly’ (HP:0000256), they BOTH share the common ancestor, ‘Abnormality of skull size’ (HP:0000240) on the phenotypic hierarchical tree. The Lin method might be particularly adept when extensive clinical phenotyping information is available and for potential mirror trait analyses. We have used the Lin method and omim.org to study an MPV consisting of three known and one candidate gene in which pathogenic variation had not yet been associated with a disease. In this case, a knockout (KO) mouse had been previously described, and HPO2GO was used to quantitatively assess the contribution of that specific gene to the patient’s observed complex neurological disease (Herman et al., 2021).

To explore quantitatively the mutational contribution to phenotype in a patient with multiple de novo CNV (MdnCNV), we used Lin similarity to assess known OMIM genes mapping within duplication CNV intervals. Remarkably, the greatest contributor was NSD1 associated with Sotos syndrome (MIM #117550) due to haploinsufficiency by either SNV LoF or deletion CNV (Kurotaki et al., 2002). Indeed, the patient phenotype quantitatively matches the mirror trait described in patients with apparent reciprocal duplication of NSD1 deletion genomic disorder and (L. M. Franco et al., 2010) in about a dozen patients since the original report. Moreover, quantitative phenotyping provided evidence implicating some contribution from a second gene/locus SMARCA1 (Du et al., 2021).

To explore genotype/phenotype correlations in Robinow syndrome (RS), a genetically heterogeneous trait (DVL1, DVL3, FZD2, NXN1, ROR2, WNT5A) with gene products converging on noncanonical WNT signaling, quantitative phenotyping was applied. RS patients with DVL1 and DVL3 variant alleles clustered together, and all patients share a mutational mechanism of −1 frameshifting mutations in the penultimate exon. FZD2 variant allele traits clustered into a missense mutation group and a LoF group; interestingly one outlier had a variant allele that resided in a distinct domain. For NXN, what was perhaps most interesting was a patient outlier who combined an SNV with an ~ 1 Mb deletion CNV in 17p13.3. Also the most common non-RS phenotype in the cohort was four patients with FDG1 variants associated with Aarskog-Scott syndrome (AAS; MIM #305400) that clustered together within the differential diagnosis cluster quantitatively showing the remarkable clinical phenotype overlap (C. Zhang et al., 2021).

CONCLUSION

The ‘post-genomic’ era is upon us, but the work to do on human genes AND disease phenotypes remains - and it is immense! As clinical genomics continues to ALWAYS strive to improve, so will our understanding of human genes and the perturbations to ‘biology in balance’ that can result in disease. The partnership among clinicians, clinical geneticists, molecular geneticists, bioinformaticians, computer scientists, mathematicians, geneticists, and human disease scientists, and patients and families, will remain critical as the field of clinical genomics and molecular diagnostics continues to evolve. Clinical phenotyping, robustly performed and quantitatively assessed, will remain central to the practice of medicine. Free exchange of information, data, clinical ‘know-how’, professional expertise, intellectual curiosity, and the pursuit of knowledge for its own sake, for ALL countries of the world will benefit ALL humankind. Happy birthday Victor!