This study in 2 patients with non-Herlitz JEB describes for the first time to our knowledge revertant mosaicism of LM-332 (Figure A). The identification of different second-site mutations again demonstrates that single individuals can undergo multiple reversion events (10
). All rescue mechanisms involved nucleotide changes in the DNA: c.565-3T→C, c.596G→C;p.G199A, c.619A→C;p.K207Q, c.628+42G→A, and c.629-1G→A (Figure B). The second-site mutations had effects at the RNA level and thereby at the protein level.
Schematic drawing showing the different second-site mutations that all corrected LAMB3:c.628G→A.
The inherited G→A transition at the last base of exon 7 (c.628G→A) converts a negatively charged glutamic acid residue into a positively charged lysine (p.E210K). This substitution takes place within the N-terminal globular domain of the short arm of the LM-332 β3 chain, which has been postulated to be critical in the association of LM-332 with other structural proteins of the BMZ, including laminin-311 (18
). Such polarity change may disturb protein-protein interactions of LM-332 and reduce epidermal-dermal adhesion. A more profound effect is observed on mRNA splicing, as the 628-nucleotide change takes place in the consensus sequence at the 5ι splice site (14
). The c.628G→A mutation generates 4 additional aberrant transcripts as described by Pulkkinen et al. (14
). RNA splicing is directed by the 5ι donor splice site, the 3ι acceptor splice site, and the branch point sequence that is located 18–40 nt upstream of the acceptor splice site (reviewed in ref. 21
). In the human genome, these splicing sequences are poorly conserved; only the GT at the 5ι end of the intron, AG at the 3ι end of the intron, and branch point adenosine at the 2ι position are almost invariable. Since natural splice sites can be different from the consensus sequence, both splice site strength and accessory regulatory sequences influence splice site selection. Various web-based resources for splice site prediction are available, such as Automated Splice Site Analyses, located at https://splice.cmh.edu. Computational analyses of our LAMB3
genomic sequences showed that the c.628G→A mutation weakens the individual information content (Ri
) of the natural donor site from 8.2 to 5.1 bits. Besides this 5ι splice site, 2 other donor sites with a high Ri
value were present in the analyzed sequence. These sites were TGGïΣΦGT, with an Ri
of 4.3 bits, and CAGïΣΦGT, with an Ri
of 3.2 bits, located 37 and 66 nt downstream of the natural exon/intron 7 border, respectively.
The second-site mutation c.628+42G→A lowered the Ri value of the alternative donor site TGGïΣΦGT drastically from 4.3 to 0.8 bits, whereas the CAGïΣΦGT site remained unchanged at 3.2 bits. The weakening of the TGGïΣΦGT donor site might favor use of the cryptic CAGïΣΦGT site by the spliceosome, thereby resulting in the observed larger mRNA transcript with retention of the first 66 nt of intron 7. The in-frame insertion led to incorporation of a stretch of 22 amino acids — SQCGYFSCPWNYGWKRKQSWSP — in the N-terminal domain of the β3 chain. Although this stretch contained 2 amino acids with basic side chains, 4 with acidic side chains, and 6 with bulky aromatic side chains, the resulting LM-332 protein was apparently functional, as it resulted in reversion of the skin phenotype.
The second-site c.596G→C changed a glycine to alanine at amino acid position 199, and second-site c.619A→C changed a lysine to glutamine at position 207. More importantly, both also affected mRNA splicing, as more normal-sized transcripts were present in the revertant cells. Computational analyses did not demonstrate effects on Ri
values of the splice sites used. We also excluded the possibility that the mutations altered exonic splicing enhancer (ESE) sequences using RESCUE-ESE software (http://genes.mit.edu/burgelab/rescue-ese). ESEs enhance splicing when present downstream of a 3ι splice site and/or upstream of the 5ι splice site (22
). A third factor that may influence splice site selection is a possible effect of these second-site mutations on the RNA secondary structure that contains the c.628G→A mutation. Single nucleotide alterations can affect the secondary structure of RNA, which in turn can influence RNA splicing (23
Two nucleotides downstream of the wild-type 3ι acceptor site of intron 7, CAGïΣΦAG, a cryptic splice site, GAGïΣΦGT, was present. Therefore c.628G→A generated not only a 64 bp–deleted transcript, but also the 66 bp–deleted transcript. The second-site mutation 629-1G→A reduced the strength of the natural acceptor site from 10.5 to 2.9 bits, whereas the cryptic splice site increased in strength from 6.0 to 7.8 bits. RT-PCR analysis indeed demonstrated preferential use of the cryptic splice site generating an in-frame rather than an out-of-frame transcript. Translation resulted in a smaller functional β3 polypeptide lacking 22 amino acids within the N-terminal domain of the short arm. The final second-site mutation, c.565-3T→C, was located in the 3ι acceptor site of intron 6 and gave a small increase, from 5.3 to 5.6 bits, in the Ri value. Although small, it might explain the higher production of wild-type mRNA transcript.
The other inherited mutation, p.R635X, which is the most common mutation in JEB patients of European origin, predicts a premature termination codon (PTC) within the coiled-coil rod of the LM-332 protein. The corresponding transcript level is expected to be markedly reduced due to nonsense-mediated RNA decay. RT-PCR around exon 14 indeed showed reduced expression of the 589 amplimers as described by Pulkkinen et al. (25
). Also, a smaller migrating 210-bp amplimer was detected for patient 078-01 belonging to an out-of-frame transcript lacking exon 14 (379 nt). Although this nonbeneficial transcript has not been reported before, it is not uncommon that PTCs induce exon skipping (26
Just over a year ago, several research groups separately identified the phenomenon of multiple in vivo reversions in single individuals, thereby challenging the concept that revertant mosaicism occurs through a single event. Multiple somatic reversions have now been observed for the FAH
gene in tyrosinemia type I (16
); the RAG1
gene in Omenn syndrome (17
); the CD3
ζ gene in severe combined immunodeficiency disease (27
); the COL17A1
gene in EB (10
); and here for the LAMB3
gene in EB. The number of recognized compensatory second-site mutations in inherited disorders is rapidly growing. Wada et al. (17
) described a patient with 6 different second-site mutations, all correcting the same delC mutation in the RAG1
gene, whereas Rieux-Laucat et al. (27
) identified 3 somatic second-site mutations that all reverted the nonsense codon p.Q70X in the CD3
ζ gene to a missense codon. The mechanism underlying the different second-site mutations in their patients remained unclear. An increased rate of reversion events can be evoked by mutational hot spots such as direct repeats or homonucleotide tracts (28
). Therefore, we analyzed the sequences around our second-site mutations. The known CAGG/CCTG mutation hot spot sequence that Levran et al. (29
) identified near some point mutations was not in close proximity — defined as more than 10 nt away — to any of the second-site mutations. Homonucleotide tracts were identified for 596G→C, 619A→C, and 628+42G→A, while a short direct repeat was identified for 629-1G→A, showing that the majority of the mutations was caused by misalignment-mediated errors.
Neither of the patients had exposure to chemotherapy or radiotherapy, excluding the possibility that they played a role in the development of these mutations. UV light could be also ruled out as a factor, as neither patient exposed his skin to sunlight, and the signature UV transition mutations, C→T and G→A, were only present in 2 of 5 second-site mutations (30
). The cause of the correcting DNA events in our patients thus remains uncertain. It must be random mutagenesis or a yet-unidentified remedial mechanism, and the repair data described here favor the latter.
Arguments in favor of random mutagenesis.
The average mutation rate in humans is estimated to be 175 mutations per diploid genome per generation by Nachman and Crowell (31
). It may well be that our patients have a higher mutation rate and that therefore different advantageous second-site mutations accumulated in keratinocytes. Other mutations in different genes or in different cell types that are not advantageous for the cell may just get lost. Accordingly, no second-site mutations were detected in peripheral blood or the fibroblast samples. Such an increased mutation rate can be the result of inactivation of a caretaker gene (32
). Inactivation results in genetic instabilities, causing an increased mutation rate affecting all genes. In light of this hypothesis, it is interesting to note that both patients developed cancer, which is known to result from an accumulation of somatic mutations (33
Arguments in favor of a directed mutagenesis.
In the heterozygous patient 078-01, all 3 second-site mutations correct the same 628G→A mutation. No second-site mutation was observed for the primary 1903C→T mutation on the other allele. The FAH
gene patient described by Bliksrud et al. (16
) was also heterozygous, and similar to our findings, both in vivo reversions were located on the same allele. No knowledge of the proper wild-type sequence is required to execute the reversion repair mechanism. While it can be argued that in a heterozygous patient, information on the correct sequence is available on the other allele, this can not be the case in a homozygous patient. In 029-01, homozygous for 628G→A, 4 different second-site mutations were identified. Also, the probands described by Wada et al. (17
) with 6 different reversions and by Rieux-Laucat et al. (27
) with 3 different reversions were both homozygous. More cases of in vivo reversion in homozygous and hemizygous patients have been described previously (2
Recently, we reported revertant mosaicism for type XVII collagen non-Herlitz JEB (10
). In these COL17A1
mosaic patients, the revertant patches remained stable during life and did not expand. Apparently, the revertant stem cells did not have a selection advantage compared with their deficient counterparts. Therefore, we concluded that the correcting mutations leading to the healthy skin patches of tens of square centimeters in size occurred during embryogenesis. In the LAMB3
mosaic patients described here, the situation was different. According to patient 078-01, his revertant skin patch increased in size, while patient 029-01 had not noted extension of the healthy skin area. This difference might be explained by the fact that patient 029-01 (c.628G→A/c.628G→A) likely has a higher level of LM-332 production in his deficient cells than patient 078-01 (c.628G→A/p.R635X), because the allele containing the nonsense mutation in the latter was not contributing to the LM-332 production. Therefore, expansion of reverted keratinocytes could indeed have been easier in proband 078-01 than in 029-01, as the deficient cells were less able to compete.
LM-332 is involved in cell locomotion and migration in wound healing (34
). Interactions with α2β1 and α3β1 are important for cell attachment, spreading, and migration, whereas binding to α6β4 results in stable anchorage without cell spreading (36
). Also, suppression of endogenous LM-332 in an oral squamous cell carcinoma cell line led to decreased cell attachment and increased migration (38
). The benefit of reversion is possibly related to the role of LM-332 in migration; however, this potential selection advantage of LM-332 revertant stem cells requires thorough investigation.
The treatment for genetic diseases seems to lie in gene therapy. Mavilio et al. (5
) recently showed in a phase I clinical trial the correction of LM-332–deficient JEB by transduction of retroviral vector expressing β3 cDNA. A 36-year-old man received a transplant of cultured epidermal sheets on both legs after removal of the outer skin layer. Transplantation was successful, as during the first year of follow-up, blistering, infections, inflammations, and an immune response were absent. Revertant mosaicism opens the fascinating possibility of “revertant cell therapy” for mosaic patients using patient’s own naturally corrected cells for transplantation. In LAMB3
revertant mosaic patients, one might take advantage of the patient’s own naturally corrected cells for skin transplantation. This autologous cell therapy bypasses the phase of molecular gene correction. Revertant mosaicism was thought to be a rare event, but our recent observations indicate that it might occur at a higher frequency than expected.