Discussion

Dissection of Biallelic Effects from other Determinants of Phenotype
Although phenotypic consequences, referred to here as biallelic effects, resulting from two different mutant alleles in compound heterozygote patients have been postulated, such effects have historically been difficult to distinguish from the influence of environment and genetic background. We used a genetically defined mammalian model system under controlled environmental conditions to reveal phenotypic effects attributable specifically to combinations of differentially mutated Xpd alleles.
The observed biallelic effects were of three general types. In the first, the allele associated in a homozygous state with a phenotype closer to wt singularly determined the phenotypic outcome, a phenomenon widely known in human recessive disease. Because these Xpd alleles functioned at or near wt levels with respect to a particular function, we call these effects “dominant”. Such alleles can also be referred to as “separation of function” alleles, because they allow dissection of the roles of multifunctional proteins in specific phenotypes.
Secondly, highlighting the potential relevance of current findings to all diploid organisms including humans was the observation that in one compound heterozygous animal, the Xpd allelic relationship could shift from A dominant |a recessive to A recessive |a dominant with respect to different phenotypes in a time-dependent and tissue-specific manner (see below and Table 2).
In the third type of biallelic effect, known as interallelic complementation, two mutant alleles produced a phenotype closer to wt than either could alone in a homo- or hemizygous state. As summarised in Table 2, examples of all types of biallelic effects were observed in a variety of Xpd-associated phenotypes, ranging from brittle hair to segmental progeria.

TFIIH in Transcription and Repair: Mechanisms of XPD Disease Pleiotropy
We observed differences in the ability of XpdTTD versus homozygous lethal Xpd†XPCS and Xpd†XP alleles to function in two transcription-related phenotypes separated in the organism by both time and space: embryonic lethality and terminal differentiation of enucleating skin and blood cells. The preblastocyst-stage homozygous lethality shared by the XpdKO, Xpd†XPCS, and Xpd†XP alleles most likely reflects a defect in basal transcription that is incompatible with life. In XpdTTD/ †XPCS and XpdTTD/ †XP compound heterozygous mice, embryonic lethality was fully rescued by the XpdTTD allele. Because embryonic lethality was also fully rescued in XpdTTD/KO hemizygous mice, the XpdTTD allele can be considered as wt and thus dominant to each of the homozygous lethal alleles (XpdKO, Xpd†XPCS, and Xpd†XP) with respect to this particular phenotype (Table 2).
TTD-specific cutaneous and anaemic features, on the other hand, are thought to result from a specific kind of transcriptional insufficiency caused by depletion of unstable TFIIH during the terminal differentiation of skin, hair-shaft, and blood cells [16,24]. In compound heterozygous mice, both homozygous lethal Xpd†XPCS and Xpd†XP alleles were able to alleviate XpdTTD-specific cutaneous and anaemic features and can thus be defined as dominant over the XpdTTD allele with respect to these phenotypes. We conclude that the defects leading to embryonic lethality and aberrant terminal differentiation of the skin, hair, and blood represent two qualitatively and/or quantitatively different transcriptional deficiencies. During early embryonic development, XpdTTD is dominant over the Xpd†XPCS and Xpd†XP alleles, whereas later in the ontogenesis of skin, hair-shaft, and blood cells, the situation is reversed.
In its role in the repair of UV photolesions, the Xpd†XPCS allele imparted a clear UV survival benefit over a single XpdTTD allele or two XpdXPCS alleles independent of expression levels, which is consistent with interallelic complementation. However, the observation that no other cellular or biochemical UV-related parameters were improved in XpdTTD/ †XPCS argues against complementation of this repair activity in the rescue of TTD progeroid symptoms in vivo.

Interallelic Complementation and XPD Function
What does interallelic complementation tell us about the mechanism of XPD function? Interallelic complementation is most often observed in multimeric proteins with multiple functional domains. Unfortunately, the structure–function relationship between disease-causing mutations and XPD functional domains, including detailed structural information on XPD or even its stoichiometry within TFIIH, remains unknown. However, based on the ability of cell extracts that are defective in two different TFIIH components (XPD and XPB) to complement NER activity in vitro [26], it is likely that TFIIH (or its components) can either multimerise or exchange at least during the NER reaction. Furthermore, XPD is known to be a “loosely bound” subunit of TFIIH [27]. We thus envisage the molecular mechanism of interallelic complementation to involve the exchange of XPD molecules within the TFIIH complex or turnover of TFIIH complexes containing different XPD molecules at the site of DNA damage during the course of the global genome as well as transcription-coupled repair of either UV-induced or endogenous DNA damage.

A Biallelic Paradigm for XPD Disorders
Recently, proteins originating from presumed null alleles were biochemically characterised as inactive in basal transcription [27], providing an explanation as to why these alleles failed to rescue lethality in haploid S. pombe with a null mutation in the XPD homologue rad15 [19]. Our data suggest that certain presumed null alleles, although unable on their own to support basal transcription, may in fact have a substantial impact on disease outcome in compound heterozygous humans, as they do in mouse models.
Clinical evidence in support of this hypothesis comes from a number of XP complementation group D patients that do not fit within the framework of the current monoallelic paradigm of XPD disorders (Figure 5). In contrast to two hemizygous XPDXPCS patients carrying the XPDG47R- or XPDR666W-encoding alleles who died of the disease before 2 y of age, two compound heterozygous XPDXPCS patients carrying the same XPDG47R- or XPDR666W-encoding alleles in addition to the presumed null XPDL461V+del716−730 both had considerably milder disease symptoms and survived more than ten times longer (A. Lehmann, personal communication) (Figure 5). Compound heterozygosity is also associated with the recently reported combination XP and TTD (XPTTD) syndrome [8]. Similar to the XpdTTD/†XPCS and XpdTTD/†XP mice described here, both patients with XPTTD described so far had intermediate hair cysteine values. Furthermore, XPTTD patient XP38BR carried a “causative” TTD mutation in one allele and a novel point mutation encoding XPDL485P in the other. Although the XPDL485P-encoding allele fails to complement viability in the haploid S. pombe rad15 deletion strain and is thus interpretable as a null allele [8], we nonetheless suggest that the combined XPTTD phenotype in this patient involves phenotypic contributions from both alleles. Taken together, these data suggest a shift to a biallelic paradigm for compound heterozygous patients in XP complementation group D.
Figure 5  Genotype–Phenotype Relationships in XPD Disorders
According to the current monoallelic hypothesis, phenotype is determined solely by the causative allele product. If a second, different allele is present, it is considered a functional null. There is a lack of any correlation between the site of the XPD mutation and the resulting disorder. We propose a biallelic hypothesis for compound heterozygotes in which both alleles can contribute to the phenotype. Examples of compound heterozygous patients in which a second, presumed null allele is likely to contribute to disease outcome are provided above in comparison to corresponding homo- or hemizygous patients with the same causative allele. Numbers in the schematic of the protein indicate the helicase domains.

Potential of Combined Recessive Alleles to Affect Phenotypic Diversity in Mammals
In humans, the clinical relevance of biallelic effects such as interallelic complementation remains unknown. Although interallelic complementation between two endogenous mutant alleles has been described in cells from a compound heterozygous patient with methylmalonic acidaemia, no observable effects on disease outcome were noted in the patient [28]. Thus, to the best of our knowledge, the amelioration of progeroid features observed here is the first in vivo demonstration in compound heterozygous animals of interallelic complementation relevant to a human disease. Keeping in mind that the ~1,200 alleles known to exist for the CTRF gene implicated in the common autosomal recessive disorder cystic fibrosis alone [29] can theoretically result in ~700,000 different allelic combinations, the potential number of allelic combinations of different recessive mutations and single nucleotide polymorphisms genome-wide is currently incalculable. We suggest biallelic effects as a previously underestimated yet important variable in considering genotype–phenotype relationships from autosomal recessive disease to normal phenotypic diversity in mammals. Extension of the above concept implies that recessive mutations can enter evolutionary selection in F1 provided that the second allele carries a different recessive alteration. Finally, our data highlight the potential of clinically relevant alleles previously designated as null, with little or no detectable expression or activity, to nonetheless contribute to phenotype.