Congenics in the pathway from quantitative trait loci detection to gene identification

is that the way to go?

The proclamation that the sequencing of the human genome has been completed, and that science is moving onto proteomics and beyond, has led many to assume that the basic search for genes that underlie the pathogenesis of human disease is done and over with. This is not so, in particular for complex diseases, and the field of genomics continues to be live and kicking. It is true that much is already known about the genomic location of many of the genes responsible for disease. A large number of quantitative trait loci (QTLs), defined as genomic segments which incorporate genes that account for quantitative traits, have already been detected and reported in a wide variety of models simulating complex diseases in humans. In the field of hypertension, linkage analysis and interval mapping using experimental crosses designed to unveil the responsible genes have in fact yielded QTLs on nearly all chromosomes. Unfortunately, merely knowing the chromosomal location of genes is not sufficient, because this knowledge has not led to the actual identification of the culprit genes in the vast majority of cases. The major stumbling block in the process appears to reside in the step from QTL detection by linkage analysis to the identification of the gene(s) within the QTLs that are causative of complex diseases.

The ability to dissect out and positively identify candidate genes within QTLs has been hampered primarily by the large QTL size, with their span ranging in most cases from approximately 5 cM to > 40 cM. This is translated in some species from approximately 10 to over 80 million base pairs. Obviously, such genomic segments are by far too large to sequence. In addition, such large genomic segments most likely incorporate several dozens to hundreds of genes, all of which are potential candidate genes. The inability to easily pinpoint within the QTL the responsible gene(s) has led researchers to seek methods that allow them to narrow down the QTL span. Once the QTL is narrow enough, it can be sequenced and genes and mutations within these genes can be sought relatively effectively. What is considered narrow enough? A QTL span approaching 1 cM is currently thought to be ‘reasonable’ or ‘workable’ in terms of the cost of sequencing and the number of genes that it incorporates.

How can one narrow the span of a QTL? The strategies that are currently available to the researcher aiming for a ‘workable range’ are very limited. One way to achieve this is through fine mapping. This strategy calls for increasing the number and genomic density of the genomic markers, usually microsatellites, that are used in interval mapping to delineate the QTL. The success of the method depends heavily on the ability of the researcher to detect microsatellites that are informative and that are located within or in very close vicinity of the QTL segment of interest. Another approach uses congenic strains. The current paper by Joe et al. [¹] is representative of such an effort to go a step beyond the initial detection of a QTL by linkage analysis in a rat model of hypertension and to narrow down the QTL to a ‘workable’ range. By constructing new congenic strains and substrains and through the use of substitution mapping, Joe et al. [¹] were able to reduce the span of their previously detected QTL on rat chromosome 1 from 17 cM to 1.4 cM which is more manageable, in terms of the ability to detect within it likely candidate genes.

The concept of congenics is not novel and is not unique to the rat. It has been termed as a standard procedure of experimental mammalian genetics. However, its major application has been in rodents, mostly in the mouse and the rat, with most of the ensuing publications surfacing after 1995 relating to the search for the genetic basis of complex diseases. The use of congenic strains has in fact evolved during the past decade as the major working algorithm used by researchers in their attempts to narrow down the span of QTLs.

Congenic strains, of which both the concept and applications are elegantly reviewed by Rapp [²], are animal strains in which the native genomic background is maintained unchanged except for a specific genomic segment of interest which incorporates the QTL that is transferred from a contrasting strain. When an entire chromosome is thus transferred, a consomic strain has been constructed. When a segment incorporating an entire QTL has been moved, a congenic strain has been created. When only part of a QTL has been transferred, a congenic substrain has been generated. The transfer of genomic segments from one strain to the other, whether incorporating an entire chromosome or only part of it, is by random recombinations that occur spontaneously during meiosis. The construction of these strains is tedious. It involves initial crossbreeding of animals from contrasting strains, subsequent repeated backcrossing onto one of the parent strains depending on the genomic background that is to be maintained and, finally, brother-to-sister mating to fix the transferred allele on the appropriate genomic background. It is the task of the researcher to identify those animals amongst the offsprings of the backcrosses in which the desired recombinations have occurred, favouring the transfer of select genomic segments of interest from one strain to the other. The process is very lengthy, requiring backcrossing over at least eight generations, unless the more laborious ‘speedy congenics’ alternative strategy is applied, which allows the construction of congenics within four to five generations of backcrossing.

What is the expected yield from congenic strains? Primarily, congenic strains allow the researcher to confirm that the QTL previously detected by linkage analysis indeed carries a gene or a set of genes that are causally associated with the phenotype of interest, hypertension in the case of Joe et al. [¹]. Thus, construction of congenics falls within the science of functional genomics. When a genomic segment is transferred from a normotensive strain onto a hypertensive background, a reduction in blood pressure in the hypertensive strain is anticipated. Indeed, this is what Joe et al. [¹] reported in their study, thus proving unequivocally that the QTL they worked on was functionally related to hypertension. What happens, however, when a culprit QTL is transferred from the hypertensive strain onto the normotensive background? Does this procedure lead to a rise in blood pressure in the otherwise normotensive strain? Such results would have been an even more convincing and conclusive proof of the hypertensive nature of the QTL. And yet, Joe et al. [¹] do not provide the results of such a reciprocal congenic strain. This is true not only for this research group, but also for other researchers who have failed to achieve the anticipated increase in blood pressure. Some have reported only a partial increase in blood pressure, others no increase at all. It appears that it is much more difficult to increase blood pressure on a normotensive background than to decrease blood pressure on a hypertensive background. Why this is so is not entirely clear. It has been speculated that a yet undefined permissive background is required to allow blood pressure to increase, and that mere transfer of a QTL containing a hypertensive gene or genes onto a normotensive background is not sufficient to generate hypertension. It is possible that susceptibility genes, in this case resistance genes, are involved. Thus, hypertension may develop in the absence of resistance genes but fails to develop in their presence. When the hypertensive QTL is transferred from the hypertensive animal onto the genomic background of the normotensive animal without removing or neutralizing the genes that confer resistance to hypertension, blood pressure fails to go up. As a result of the difficulties encountered in attempting to raise blood pressure on the normotensive background, many researchers have simply discontinued their attempts to construct and study the reciprocal congenic strains.

The second major gain from transferring QTL segments across strains and constructing consomic, congenic strains and substrains, beyond the functional proof validating the causative association between the QTL and the phenotype, is the ability of the researchers to narrow down the span of the QTL to a desired range. Joe et al. [¹] are currently reporting to have successfully narrowed down the QTL to approximately 1.4 cM, which is a truly remarkable achievement in itself.

At this point, it may be worthwhile to reflect on the theory that lies behind the usefulness of congenic strains. The assumption that the concept underlying the congenics approach is valid has led researchers such as Joe et al. [¹] and many others to invest much of their time, effort and resources in the construction of these strains. Is the concept truly sound and foolproof enough to justify such a major commitment? For monogenic diseases, where one gene accounts for one phenotype, the congenic strain could have been an ideal tool in achieving the desired goal. However, practice has shown that the causative genes in monogenic diseases are relatively easy to detect and that construction of congenic strains is not necessary to achieve this goal. For polygenic disease, if only two genes were to account for one phenotype and if each gene were to act alone and independently, then the congenic strains would also have been useful. However, if gene-to-gene interaction were to occur, then the researchers would have had to construct double congenics (i.e. congenic strains that simultaneously incorporate the two QTLs). On the other hand, if each of the interacting genes had a major effect on its own, single congenics would have still been helpful in detecting them. In the case of hypertension, which is thought to be polygenic and multifactorial, it is likely that many more than one or two genes are causally involved. In the Dahl model that Joe et al. [¹] used in the current study, Rapp and colleagues previously detected and reported QTLs on nearly all the rat autosomes and chromosome X [^2–4], indicating the involvement of a large number of genes affecting blood pressure, certainly more than one or two. Is it reasonable in such a case where multiple genes are involved, and gene-to-gene interaction is so much likely to occur, to utilize single congenics, as opposed to double, triple or quadruple congenics? It is quite likely that some genes carry a more profound effect on blood pressure than others, favouring the use of single congenics, but gene-to-gene interaction is hardly likely to be detected by using only single congenics. Thus, the use of single congenics in the attempt to confirm the functional role of a QTL and narrow down its span is problematic and risky. Acknowledging this limitation of single congenics, Rapp's group has studied in the past gene-to-gene interaction using double congenics, which is an even more laborious process [⁵]. And yet, one cannot argue with success, as Joe et al. [¹] appear to have accomplished the narrowing down of the span of their QTL to 1.4 cM using multiple single congenic strains and substrains.

Interestingly, in the study by Joe et al. [¹], transferring the QTL from the Lewis rat onto the genomic background of the hypertensive Dahl strain achieved a reduction in blood pressure of up to 28 mmHg. When reviewing additional publications of Rapp and colleagues [^3,6–8] we notice a similar reduction in blood pressure, in terms of order of magnitude, with other QTLs as well. If we were to add up the blood pressure reductions attributed to all QTLs, the animals would have had a negative blood pressure. How is this possible? One explanation is that each of the genes may act through a common pathophysiological pathway, but working at different levels along the same axis. Thus, this pathway can be activated at several levels by different genes located in different QTLs, achieving a similar order of magnitude in terms of the hypotensive effect. Another possibility is that indeed gene-to-gene interaction occurs and that the magnitude of the QTL effect on blood pressure might have been lower in the presence of other genes. Such interaction cannot be detected when a single gene or set of genes is transferred in single congenics, but might have been detected had double or triple congenics been used.

Finally, did the study by Joe et al. [¹] lead to positive and definitive identification of genes that are implicated in hypertension? The answer is no. However, they did come out with a list of 19 annotated rat candidate genes. This list may well be incomplete because sequencing of the rat genome has only very recently been achieved, and annotation of genes within the rat genome is still ongoing. Has the study by Joe et al. [¹] been cost effective? Apparently so, for a lack of alternatives. Joe et al. [¹] have unquestionably been able to reduce the number of candidate genes within their QTLs from several hundreds of genes within a 17 cM QTL to a mere 19 genes within the 1.4 cM QTL. But how useful is it to reduce the number of candidate genes within a QTL from over 100 genes to 19? The task to pinpoint the culprit gene remains enormous because the study of individual candidate genes is remains laborious and some researchers have spent lifetime careers each focusing on very few candidate genes at a time. On the other hand, if we accept the critical perspective recently put forth by Pravenec et al. [⁹] suggesting a continuing algorithm in the search for the genetic basis of hypertension that combines genomics with transcriptomics, and if the authors of the current study will try to determine which of these 19 candidate genes are also differentially expressed, as can be found using DNA microarrays, then the yield of their efforts with the congenics may well be worth the cost. But the proof the pudding is yet to be made, as congenics have not led so far to identification of novel candidate genes. However, this does not preclude that this might occur in the very near future.