Two varieties of an inversion breaking the factor VIII gene (F8) and causing half of all severe hemophilia A have previously been described. These were thought to involve the intron 22 sequence: int22h-1  interacting with either the proximal (int22h-2), or the distal (int22h-3) of two extragenic and telomeric copies. However, new sequence data show that only int22h-3 should be involved in these inversions . We explain this discrepancy and note that the new sequence data predict a confounding complication for mutation detection in some families with hemophilia A.
In 1993, it was shown that int22h-3 is in inverse orientation to int22h-1 and the same was thought to be true for int22h-2 because BclI blots indicated that both int22h-2 and int22h-3 recombined with int22h-1 to cause inversions . By contrast, the new sequence of the human X chromosome  shows that int22h-1 is in the same orientation as int22h-2 so that their interaction should lead to deletion and duplication rather than inversion. These data show that int22h-2 and int22h-3 form the arms of an imperfect palindrome separated by a 65 kb loop containing the Bcl1 sites that characterize int22h-2 and int22h-3. A further 40 kb of duplicated sequence extends the palindrome (Fig. 1).
The involvement of int22h-2 in recombinations causing inversions that disrupt the F8 gene may be explained by postulating that recombination between the arms of the palindrome has led to a polymorphism entailing the inversion of part of the palindrome including its loop. This would create alleles where the int22h-2 repeat is the most telomeric and the one in inverse orientation to int22h-1 (Fig. 1). Some findings support this hypothesis, thus, the F8 mRNA of a patient with an inversion involving int22h-2 indicated that transcription of the inverted part of the F8 gene had extended from exon 22 into cryptic exons in the distal arm of the palindrome and to an exon external and telomeric to this arm [4,5]. Furthermore, long PCR in one of our inversion patients has indicated that the centromeric portion of int22h-1 is joined to an int22h copy flanked by sequences that are at the centromeric end of the loop of the palindrome described by the new sequence data . This int22h-2 sequence would recombine to produce an inversion only if it had been the most telomeric and hence, the one in inverse orientation to int22h-1 as postulated above. In fact, the frequency of this polymorphic allele could explain the rarity of the inversion involving int22h-2 (i.e.: 1/6).
The palindromic arrangement of int22h-2 and int22h-3 may favor their intra-chromosomal gene conversion and thus help explain why these sequences are fivefold more identical when they are on the same chromosome .
Recombination between similarly oriented int22h-1 and int22h-2 may cause highly deleterious 0.5 Mb deletions [2,7] or, alternatively, duplications. The latter could cause problems in the characterization of hemophilia A mutations as a normal but non-functional copy of a F8 gene region may obscure relevant mutations. Furthermore, the long PCR test for the detection of int22h-related inversions  would not distinguish between an inversion carrier and the carrier of a duplication.