I. BLOOD PRESSURE AS A QUANTITATIVE TRAIT

Top Previous Next References

Quantitative traits are those that show continuous variation from low to high values. This natural variation is present in all populations of eukaryotes and is caused by environmental or genetic factors, or by both genes and the environment. A recent short review emphasizes the interaction between genes and environment in hypertension (75). The basis for the genetic component of quantitative traits is often due to the cumulative effects on the phenotype of many genetic loci, i.e., the trait is polygenic. The loci that control a quantitative trait have been termed quantitative trait loci (QTL). The term QTL is used to describe a broad chromosomal region that may contain one or more loci controlling the quantitative trait and also to refer to one of the individual specific genetic loci participating in controlling the trait.

Many traits of biological, medical, and agricultural importance are quantitative in nature, e.g., intelligence, body weight, body length, plant and animal disease resistance, crop yields, and seed oil content. Until the advent of molecular biology and the development of large numbers of easily typeable genetic markers, the dissection of quantitative traits into their individual genetic components was impossible except in model organisms such as Drosophila that have unique genetic properties.

Blood pressure (BP) in humans is well known to have a genetic component based on studies in families, twins, and adopted children as reviewed by Ward (266). In animals, selective breeding (reviewed below for rodents) for high or low BP from outbred stock proves that genes controlling variation in BP segregate in essentially all such stocks. Blood pressure at the high end of the distribution (hypertension) predisposes humans (and rats) to cardiovascular disease (stroke, coronary heart disease, heart failure, peripheral vascular disease, and renal failure), and the higher the BP, the worse the cardiovascular disease (154). High BP above 140 mmHg systolic or 90 mmHg diastolic is generally defined as hypertension; this classification, although useful, is an arbitrary truncation of a continuous distribution. Because genetic hypertension (so-called essential hypertension) has a high prevalence that increases with age in Westernized populations (271), it represents a major health problem. For these reasons, it is desirable to describe the genes that account for the natural variation of BP in populations as a requisite for understanding the genetic causes of hypertension.

The physiological systems altering BP are well known, and there are potent pharmacological agents for lowering or increasing BP designed around these known physiological systems. More recently, transgenic experiments in mice allow deletion or addition of specific genes that influence BP (109, 168, 177, 235, 236, 237). Such experimental manipulations of physiological systems are very informative on a physiological level because they provide information on how quantitative variation in candidate genes influences BP in the context of the whole animal with all its regulatory and compensatory systems intact. It needs to be emphasized, however, that such manipulations (with one possible exception, Ref. 84) give no evidence as to which genetic loci actually do harbor naturally occurring alleles that are present in populations in high enough frequency to account for the high prevalence of hypertension. Because this review deals with detecting naturally occurring alleles altering BP, detailed review of the transgenic literature on candidate genes is not included. It is noted, however, that once naturally occurring candidate allelic variation is detected, the ultimate test of its BP effect will almost certainly be transgenic experiments to substitute very specific DNA variation into endogenous genes by homologous recombination to evaluate their BP effects in vivo. Such experiments have not been done in the rat, although this may be feasible in the future (see sect. IX).

Although BP behaves in populations as a polygenic quantitative trait, several specific syndromes increasing or decreasing BP have been recognized in humans (19, 76, 141, 142, 216, 223, 230, 232). These are caused by rare mutants that result in drastic physiological perturbations that are penetrant enough to be observable as Mendelian traits segregating in human families. Such syndromes are usually associated with large quantitative effects on BP. Although many of these syndromes were known for decades, their molecular genetic basis has only recently been elucidated. Many of the genes identified involve either renal electrolyte transport molecules or adrenal steroid metabolic pathways controlling electrolyte balance (141, 272).

The question arises as to whether the same loci involved in monogenic variation in BP in humans are responsible for quantitative variation in BP through more subtle genetic variants. For the loci of the epithelial sodium channel involved in Liddle's syndrome, the answer appears to be no in humans (18, 181, 182) and rats (69, 88, 123). In contrast, it could be argued that the allelic variation in 11-hydroxylase in Dahl rats does provide a positive example of subtle quantitative variation in a steroid biosynthetic pathway causing an increment of BP (23, 26, 162, 190-193), but the point has not been proven by precise gene substitution experiments.

The possible Mendelian segregation of BP was observed by Tanase (250) in a cross of spontaneously hypertensive rats (SHR) and Donryu rats and would seem to have been confirmed in a four-way cross by the same author (251). Such a phenomenon has not been observed in extensive work by others using SHR, and such a powerful genetic locus in rats has not been mapped. Its existence remains problematic.

	II. ANIMAL MODELS OF HYPERTENSION

Top Previous Next References

Because human essential hypertension is a complex disease, researchers have selectively bred rats for high BP to provide animal models for the disease. In some cases, animals were also selected for low BP and/or a random-bred unselected stock from which the hypertensive strain was developed, was also maintained. These models and their initial references are listed in Table 1. Brief reviews on development of most of the rat strains and the mouse strains are available (65, 152). Because these strains are critical to any genetic analysis, brief comments on their genetic status and availability are given below.

The genetically hypertensive (GH) rats are a fully inbred strain (20 or more generations of brother × sister matings). Although they are not widely available, a colony is maintained by Eugenie Harris at the Department of Surgery, University of Otago (PO Box 913, Dunedin, New Zealand; e-mail: jean.harris{at}stonebow.otago.ac.nz). The rats were selectively bred on the basis of BP without any dietary or environmental provocative stimuli.

Dahl salt-sensitive (S) and Dahl salt-resistant (R) rats were bred on the basis of their BP after being fed a high-salt (8% NaCl) diet. There is a strong interaction on BP between strains and dietary NaCl. The original selected rats were maintained as outbred stocks but were continually selected for divergence of BP on a high-salt diet (93). These outbred stocks are commercially available (Harlan Sprague-Dawley, PO Box 29176, Indianapolis, IN; www.harlan.com) and are designated DS (Brookhaven) for Dahl salt sensitive and DR (Brookhaven) for Dahl salt resistant. The outbred stocks are of little value because inbred strains are available. Inbred strains were developed from Dahl's outbred (Brookhaven) stocks (194) and are designated SS/Jr for the Dahl salt sensitive (Rapp) and SR/Jr for Dahl salt resistant (Rapp). The definitive colonies of these strains are maintained by myself (John Rapp, Dept. of Physiology and Molecular Medicine, Medical College of Ohio, 3035 Arlington Avenue, Toledo, OH 43614-5804; e-mail: jrapp{at}mco.edu). The SS/Jr and SR/Jr strains are commercially available (Harlan Sprague-Dawley, see above; and Møllegaard Breeding and Research Centre, DK-4623 LI, Skensved, Denmark; e-mail: molgene{at}inet.uni-c.dk). Another set of inbred strains was developed from Dahl's outbred stocks at Brookhaven National Laboratory (Upton, NY) by Iwai and Heine (94). These strains have been designated Dahl-Iwai S and Dahl-Iwai R and are apparently only available in Japan (Tsukuba Research Laboratories, Aisai, 5-1-3 Tokodai, Tsukuba, Ibaraki 300-26, Japan) (280).

Unfortunately, the inbred SS/Jr rats commercially available from Harlan Sprague-Dawley were genetically contaminated in 1992-1993 (140, 209). A test of recent commercial stocks indicated that this problem was apparently corrected (264).

The SHR was selectively bred for high BP without any provocative dietary or environmental stimuli by Okamoto and Aoki (175) in Kyoto, Japan. The SHR and "control" stock from Wistar-Kyoto rats (WKY) were imported by the National Institutes of Health (NIH) in the United States before either strain was fully inbred and distributed to commercial suppliers. This created the undesirable situation of genetic differences among various colonies of SHR and WKY from commercial suppliers in the United States (130, 131, 165, 214) and between the colonies in the United States and Japan (169).

Before full inbreeding, a substrain of SHR with exceptionally high BP was found to be more susceptible to stroke than other SHR substrains. This substrain was subsequently selectively bred for a high incidence of stroke (176, 281, 282); the stroke-prone SHR are designated SHRSP. Although exceptionally high BP is clearly a factor in stroke of SHRSP, genetic factors in addition to those operating through BP also are likely to be involved. This was shown by producing F₁ and both backcross populations from SHR and SHRSP. Thus the three populations are F₁(SHRSP × SHR) × SHR, F₁(SHRSP × SHR), and F₁(SHRSP × SHR) × SHRSP, and they have 25, 50, and 75% SHRSP genes, respectively. All the populations were fed a high salt intake to remove BP differences among populations (all BP became very high), but there was still a positive relationship between the average percentage of SHRSP genes and the incidence of cerebrovascular lesions across populations (170).

The SHR and WKY are widely commercially available in the United States (e.g., Harlan Sprague-Dawley, see above; Charles River Laboratories, 251 Ballardvale St., Wilmington, MA 01887; www.criver.com; Taconic, 273 Hanover Avenue, Germantown, NH 12526; www.taconic.com) or in Europe (Møllegaard Breeding and Research Centre, see above). The SHRSP are not commercially available, but colonies are maintained around the world (Genetic Resource Section, NIH, Building 14F, Room 101, Bethesda, MD 20892; www.nih.gov/od/ors/dirs/vrp/nihagr.htm; Dr. Y. Yamori, Graduate School of Human and Environmental Studies, Kyoto University, Yoshida Nihonmatsu-cho, Sakyo-ku, Kyoto 606-8501, Japan; Fax: 81 75 7532997; Dr. R. C. Webb, Dept. of Physiology, Univ. of Michigan, 7812 Medical Science Building II, Ann Arbor, MI 48109-0622; Fax: 734 936 8813; Dr. D. Ganten, Max-Delbrück-Centrum für Molekulare Medizin, Robert-Rössle-Strasse 10, D-13125 Berlin-Buch, Germany, Fax: 49 30 949 7008; Dr. A. F. Dominiczak, Dept. of Medicine and Therapeutics, Gardiner Institute, Western Infirmary, Glasgow G11 7T, UK; Fax: 44 141 211 1763; e-mail: ad7e{at}clinmed.gla.ac.uk). Genetically defined SHRSP/Izm, SHR/Izm, and WKY/Izm are available from the Disease Model Cooperative Research Association (86-2 Jodojishimobabacho, Sakyo-ku, Kyoto 606-8413, Japan; Fax: 81 75 761 2382).

The SHR are hyperactive in an open-field environment compared with WKY. Blood pressure and hyperactivity were not correlated in an F₂ population derived from SHR and WKY, implying independent genetic determinants (81). To confirm that there was no obligate genetic association between hyperactivity and hypertension, two strains with contrasting properties were selectively bred from an F₃ population derived from a cross of SHR and WKY. One strain was selected for hyperactivity and the other for high BP. In fact, a hypertensive nonhyperactive strain (WKHT) and a normotensive hyperactive (WKHA) strain were successfully developed (48, 82). These strains have been inbred for at least 25 generations of brother-sister mating and are maintained by Dr. C. F. Deschepper (Institut de Recherches Cliniques de Montreal, 110 Pine Ave. West, Montreal, Quebec, Canada H2W 1R7; Fax: 514 987 5585; e-mail: deschec{at}ircm.qc.ca).

Sabra hypertension-prone (SBH) and Sabra hypertension-resistant (SBN) rats were bred on the basis of the BP response to unilateral nephrectomy, treatment with deoxycorticosterone acetate, and 1% NaCl to drink (7). The SBH and SBN colonies have recently been rederived from Ben-Ishay's stock and are currently fully inbred (276). The colony has been moved from Hebrew University in Jerusalem to Ashkelon, Israel (Dr. Y. Yagil, Laboratory of Molecular Medicine, Ben-Gurion University, Barzilai Medical Center Campus, Ashkelon 78306, Israel; Fax: 972 7 674 5824; e-mail: labmomed{at}bgumail.bgu.ac.il).

The Lyon strains of rats were selected for high (LH) or low (LL) BP, and a third line (LN) was selected for normal BP without special dietary or environmental challenges (53). At last report, the BP of the LL and LN rats were similar (215), and both of course were lower than LH rats. The strains are presumably inbred, but a report in 1993 found a few alleles at genetic markers still to be segregating (49). The strains are only available in France (Dr. J. Sassard, Département de Physiologie et Pharmacologie Clinique, CNRS ESA 5014, Faculté de Pharmacie, 8 Avenue Rockefeller, 69373 Lyon Cedex 08, France; Fax: 33 478 777118).

The Milan strains were selected for high (MHS) and low (MNS) BP without special dietary or environmental challenges (8). The strains are fully inbred and are maintained in the United States (Genetic Resource Section, NIH, see above) or in Italy (G. Bianchi, Div. of Nephrology, Universit di Milano, Ospedale San Rafaele, Via Olgettina 60, 1-20132 Milan, Italy; Fax: 39 02 2643 2384; Patrizia Ferrari, PRASSIS Research Institute, Via Forlanini 3, 20199 Settimo Milanese, Milan, Italy; Fax: 39 02 3350 0408).

The fawn-hooded (FH) rat was obtained initially from a cross of German-brown and white Lashley rats. It is characterized by a platelet abnormality resulting in a mild bleeding disorder (255) and by glomerular sclerosis (118). A colony was established at Unilever Research Laboratories, Vlaardinger, The Netherlands, which was maintained as a closed outbred colony until the mid 1980s, at which point it was recognized that the rats had spontaneous hypertension and renal lesions resulting in proteinuria (124-126). Selective inbreeding was performed for high BP (FHH strain) and for normotension (FHL strain) (188, 258). The FHH strain is homozygous recessive for three coat color genes: red-eyed dilution (r), nonagouti (a), and hooded (h) (14). The FHH and FHL strains are maintained by A. P. Provoost (Dept. of Pediatric Surgery, Erasmus University Rotterdam, The Netherlands; e-mail: provoost{at}heel.fgg.eur.nl).

A strain called inherited stress-induced arterial hypertension (ISIAH) was selected based on BP of conscious rats restrained for 30 min (158, 159). It seems doubtful if this selection environment is meaningfully different from other selection experiments (SHR, LH, MHS strains) that were also done on conscious restrained rats. The ISIAD rats are maintained by A. L. Markel (Institute of Cytology and Genetics, Russian Academy of Science, Siberian Branch, 630090, Novosibirsk 90, Russia; e-mail: markel{at}cgi.nsk.su).

Rats have been selectively inbred for high or normal BP in Prague producing the Prague hypertensive rat (PHR) and the Prague normotensive rat (PNR) (80). Both strains originated from the same single pair of breeders. The strains are maintained by J. Heller (Institute of Clinical and Experimental Medicine, Vídensk 800, CS-146 22, Prague 4, Czech Republic).

Mice have also been selectively bred for BP. In the case of the mice, the selection experiment was more rational than much of the rat work. With some of the rat breeding, the strains were started from a narrow genetic base (sometimes one pair of rats) that obviously limits the alleles entering the experiment, and selection and inbreeding were practiced concomitantly. The latter procedure is not optimal for separating contrasting alleles into the divergent strains. The mouse selection experiment started from an 8-way cross of inbred strains, and selection was applied for 23 generations before the lines were inbred by at least 40 generations of brother-sister mating as of 1997. Selection was done for both high and low BP, and a random-bred line was also maintained and also subsequently inbred (217-219). The mouse strains are designated BPH/2, BPL/1, and BPN/3 for the high, low, and normotensive strains, respectfully, and are available from the Jackson Laboratory (Animal Resources, 600 Main Street, Bar Harbor, ME 04609-1500; www.jax.org).

Cursory review of Table 1 indicates that there is not much genetic diversity in the origin of hypertensive rat strains; most were developed from Wistar-related stocks. Two strains were developed from Sprague-Dawley stock, but Sprague-Dawley and Wistar have a common origin (146). The fawn-hooded rats apparently have a more distant origin and thus may be of unique value. The commonality of the hypertensive strains can be partially circumvented, however, by crossing a hypertensive strain to more distantly related strains in genetic analysis of BP. This has the virtue of possibly introducing unique contrasting alleles at BP loci, and at the marker loci needed for the genetic analysis (197). Genealogic trees for rat stocks constructed from historical (146) and genetic marker (17) data are available.

	III. GENOMIC RESOURCES

Top Previous Next References

Most of the resources required in a molecular genetic analysis of QTL are available for the rat. The most useful genetic markers are short tandem repeated sequences of DNA consisting of mono-, di-, tri-, or tetranucleotide repeated sequences, or so-called microsatellites, also referred to as short tandem repeats (STR) or simple sequence length polymorphism (SSLP) markers, or simple sequence repeats (SSR). These repeated sequences are very numerous and are scattered throughout the noncoding DNA of eukaryotes. They are also highly polymorphic with regard to length and are easily genotyped using the polymerase chain reaction (PCR) (4, 239, 268, 269). Initially, the oligonucleotide primers for the PCR amplification and genotyping of rat microsatellites were obtained from sequences that were in databases as a consequence of sequencing known genes or limited screening for microsatellites (67, 129, 203, 225, 288). Subsequently, several laboratories developed larger numbers of markers by screening small-insert genomic libraries for microsatellites and prepared genetic linkage maps of all the rat chromosomes using such markers (10, 13, 29, 99, 163, 184, 263) (www.genome.wi.mit.edu; www.well.ox.ac.uk/pub/genetics/ratmap). Such linkage maps with readily available markers are certainly the single most important genomic resource for genetic analysis of any quantitative trait. Polymorphism data for microsatellites on 48 strains of rats is available at www.informatics.jax.org/rat/.

Rat chromosomes have been flow sorted by Hoebee et al. (87) and Shepel and co-workers (228, 229). Chromosome-specific libraries were used to develop microsatellite markers for chromosomes (chr) 1 (70), chr 2 (43, 227), chr 3 (38), chr 5 (42), and chr 10 (51, 52). Because flow sorting did not separate certain chromosomes, such libraries also sometimes yielded markers on multiple chromosomes (47, 228, 244). There are two reports of chromosome dissection on chr 10 to obtain genetic markers (6, 112).

Genome resources are more highly developed in the mouse than in the rat. It has been possible therefore to utilize mouse microsatellite PCR primer pairs in the rat. About 20% of such mouse markers amplify a specific product from rat DNA, but only ~4% of the mouse markers provided polymorphic markers (46, 115).

Somatic hybrid cell panels are a useful resource for localizing genes onto chromosomes. In this technique, rat and mouse cultured somatic cells are fused. For unknown reasons, one or more rat chromosomes will be stably retained at random in some of the mouse cells. A panel of such cell lines known by cytologic techniques to retain one or more rat chromosomes can be tested for the presence of specific rat genes or markers to establish linkage to rat chromosomes. Two rat/mouse hybrid-cell panels are available for the rat, one developed in Belgium by Szpirer and co-workers (245-247) and one developed in Japan at Kyoto University (285, 286).

An additional potentially important technique for preparing genetic maps utilizes radiation hybrid-cell (RH) panels (137, 265). In this technique, a donor (e.g., rat) cultured somatic cell line is lethally irradiated with X-rays and then fused to somatic cell lines of another species (e.g., hamster). Rat chromosome fragments will be integrated into the hamster chromosomes, and a series of relatively stable RH cell lines is established. The closer two rat loci are to each other on a rat chromosome, the more likely they will be retained in the same RH cell line. If the rat loci are far apart, or on different chromosomes, they will be much less likely to be retained in the same RH cell lines. Data on the presence or absence of a rat gene/marker usually detected by PCR in the RH panel can be used to construct genetic maps. First-generation RH maps are available for the mouse (62, 166). A rat/hamster RH panel developed by Peter Goodfellow's laboratory at Cambridge University in the United Kingdom is currently available from Research Genetics (Huntsville, AL; www.resgen.com) or the Resource Centre of the German Human Genome Project (www.rzpd.de). Extensive radiation hybrid cell maps of rat chromosomes containing 5,255 markers have recently been published (267) (www.well.ox.ac.uk/rat_mapping_resources). An additional extensive set of radiation hybrid cell and linkage maps for the rat are also available (242) (http://goliath.ifrc.mcw.edu/LGR/research/rhp/index.html; http://curatools.curagen.com/ratmap).

Once a gene of interest is very well localized on a genetic map, a cloning project for that gene will almost certainly require the use of large-insert DNA libraries for the rat prepared using a yeast artificial chromosome (YAC) and/or a P₁ bacteriophage vector, the so-called P₁-derived artificial chromosome (PAC) (http://bacpac.med.buffalo.edu/rat_pac.html). These important reagents have recently become available for the rat (16, 71, 273), and an anchored YAC framework map is available (15).

The rat linkage maps have been oriented with the cytogenetic maps by placing clones of known genes, cosmid, or YAC clones containing known microsatellite markers on the chromosomes by fluorescent in situ hybridization (FISH) (3, 248). Also, comparisons of the chromosomal maps for mouse, human, and rat are available (138, 279). See also www.otsuka.genome.ad.jp/ratmap for comparative maps. Synteny means that two loci are in the same linkage group; conserved synteny means that two linked loci in one species have homologous loci that are also linked in another species. Often, the order of the linked loci is also conserved between species. This information is very helpful in locating candidate loci in QTL regions or in developing genetic markers in specific regions of rat chromosomes by knowing what loci are in the homologous target region of human or mouse.

A recent new technology is the ability to construct microarrays of thousands of different DNA molecules for the purpose of hybridization to labeled DNA. The technology and its use have recently been reviewed (293). The technique is useful, for example, to array sequences from known genes to detect the relative abundance of mRNA molecules in samples (11, 50). Again, such reagents for the human and mouse genome are more available than for the rat, but a recent article utilized a microarray of 10,000 cDNA clones, from a normalized library of rat heart (1).

	IV. STRATEGIES FOR DEFINING QUANTITATIVE TRAIT LOCI FOR BLOOD PRESSURE

Top Previous Next References

Given strains of rats with markedly different BP, it is obvious that there must be a genetic basis for these differences that are passed from one generation to the next. It follows logically that genetic techniques must be employed to define and understand these genetic differences. This has only become a practical possibility with the advent of large numbers of genetic markers throughout the genome.

For example, consider the four individuals from a segregating population with genotypes shown below. Let A, B, C, D, and E be loci on five different chromosomes that influence BP, i.e., so-called BP QTL. Assume there are two alleles at each locus such that an allele with a subscript 1 (A₁, B₁, C₁, etc.) lowers BP by 5 mmHg (minus allele) and an allele with subscript 2 (A₂, B₂, C₂, etc.) increases BP by 5 mmHg (plus allele) around an overall mean of 150 mmHg. Assume as a first approximation that the effects at all loci are additive. For example, individual 1 has eight subscripts 1 (8 × 5 = 40 mmHg) and two subscripts 2 (2 × +5 = +10 mmHg), so the net effect of the plus and minus effects is 40 + 10 = 30 mmHg about an overall mean of 150 mmHg yields 120 mmHg for individual 1. The BP for the other individuals are calculated similarly
Obviously knowing that individuals 1 and 2 have the same BP does not define the genotype at, for example, QTL A which is A₁A₁ in individual 1 and A₂A₂ in individual 2. Similarly, individuals 1 and 4 have markedly different BP but have the same genotype, A₁A₁, at QTL A.

Unambiguous genotyping at a genetic locus is required to place the locus on genetic maps. For simple Mendelian traits, the genotype can be inferred from the phenotype, and the locus involved can be placed on the genetic map as the first step in cloning and identifying a previously unknown locus causing the Mendelian trait. This is the usual strategy for cloning of genes causing genetic diseases inherited in a simple Mendelian fashion. Although such accurate map localization is impossible for (essentially all) BP QTL by studying BP per se, DNA-based genetic microsatellite markers are inherited as simple Mendelian traits and are readily placed on genetic linkage maps as noted in section III. It is logical then to determine which genetic markers cosegregate with BP to find the approximate location of the QTL.

It is worth understanding how this cosegregation works in the context of experimental mammalian genetics. Consider a marker locus M, which is a microsatellite marker locus on the same chromosome and closely linked to BP QTL A. Suppose the alleles M₁ and M₂ at locus M and alleles A₁ and A₂ at QTL A are organized as follows in two inbred parental strains P₁ and P₂. Let A₁ be a minus allele that lowers BP and A₂ be a plus allele that increases BP.
In the diagram above, each line represents one chromosome of a pair of chromosomes. Thus marker allele M₁ is linked to the QTL A minus allele A₁, and marker allele M₂ is linked to the QTL A plus allele A₂. Because P₁ and P₂ are inbred strains, both the marker and QTL loci are homozygous for their respective alleles.

P₁ is crossed to P₂ and the F₁ rats are intercrossed to produce a large F₂ population.
The F₂ rats are phenotyped for BP and genotyped at the microsatellite marker locus by PCR. The marker alleles will segregate in Mendelian fashion 1M₁M₁:2M₁M₂:1M₂M₂. The BP of the rats in the three marker classes can be compared by a one-way ANOVA. In the example constructed above, there will be BP differences among the marker classes because the marker is linked to a BP QTL. In the extreme case where locus M is very closely linked to QTL A, the marker class M₁M₁ will consist of rats with genotype A₁A₁ at the QTL, M₁M₂ will represent rats with QTL genotype A₁A₂, and M₂M₂ will represent rats with QTL genotype A₂A₂. Because alleles at QTL on other chromosomes (loci B, C, D, and E in the example above) will assort at random with respect to M (and A), there will be no consistent enrichment of the alleles on other chromosomes in one marker classes at M compared with another marker class at M, and the plus and minus BP effects at the QTL other than at locus A will cancel each other out. The segregation of alleles at the other QTL will, however, increase the BP variance within a marker class at M, thus making the detection of BP effects associated with M more difficult statistically. Examples of single marker linkage to BP are given in Table 2; the marker on chr 2 is linked to BP, but the one on chr 3 is not.

As presented above, the marker M was very close to the BP locus A. In general, this will not be the case. As the chromosomal map distance between M and A increases, the chances for recombination by chromosomal crossing over between M and A increases. This will degrade the accuracy with which the genotype at M predicts the genotype at A, and consequently, the BP cosegregating with M will decrease as the distance between M and A increases.

Figure 1 shows meiosis with or without recombination in an F₁ heterozygote obtained by crossing the strains P₁ and P₂ described above. In Figure 1, top, where there is no recombination, M₁ stays coupled with A₁ and M₂ with A₂. In Figure 1, bottom, recombination occurs by chromosomal crossing over between loci M and A, and as a consequence, M₁ is as likely to be coupled to A₁ as it is to A₂, and similarly M₂ is as likely to be coupled to A₁ as A₂. More realistically, the meiotic products (and thus the population of rats produced from these gametes) will be a mixture of the events depicted in Figure 1, top and bottom. The closer M and A are to each other, the more likely the original parental coupling of M₁ to A₁ and M₂ to A₂ is to be retained in a high percentage of gametes, and the further the distance the less likely this coupling is preserved.

View larger version (24K): [in this window] [in a new window]   Fig. 1. Gametes produced from meiosis without (top) and with (bottom) chromosomal crossing over (recombination). A single pair of chromosomes is shown which have replicated into 2 sister chromatids that are still joined at centromere during synapse. M is a marker locus with alleles M1 and M2. A is a quantitative trait locus (QTL) with alleles A1 and A2.

If BP linkage to a single marker has been found, the problem arises: Where is the QTL in relation to the marker and how large is the effect of the QTL on BP? With reference to the data in Table 2, the marker at Nak1 on chr 2 is associated with a BP effect of 21 mmHg (the difference between BP of rats with SS genotype and those with WW genotype at Nak1). This could arise from a QTL very closely linked (or identical) to Nak1 with an effect of 21 mmHg, or it could arise from a QTL of larger effect at a further distance from the Nak1 locus. Without precise information on the location of an unknown locus on a chromosomal map, it is impossible to identify it.

It is, of course, possible to study loci along a chromosome for linkage to BP considering each locus separately. Representative data are given in Figure 2. The BP effect is at a maximum around the Ae3 locus (anion exchanger 3) and drops off on either side of Ae3 as the linkage map distance from Ae3 increases. Nevertheless, such data do not provide any formalized statistical estimate of QTL location and do not localize the QTL well enough by inspection to attempt to identify it by positional cloning.

View larger version (20K): [in this window] [in a new window]   Fig. 2. Linkage analysis of genetic markers with blood pressure (BP) on chromosome (chr) 9 in an F2(S × R) population. Genetic linkage map is shown at left. Rats are categorized by marker genotype at each marker. S, allele from Dahl salt-sensitive S rats; R, allele for Dahl salt-resistant R rats; n, number of rats; cM, centiMorgan. BP of 3 genotypic classes are compared by a 1-way ANOVA at each marker. [From Rapp et al. (198). Copyright 1998 Academic Press.]

The twofold problem of proving the existence of a QTL and of finding its map position has received much attention by statistical geneticists. There are good textbooks on the subject (148, 153), and several different techniques have been developed, all of which are mathematically intense and require computer implementation. Two popular and readily available programs are MAPMAKER/QTL for PC and other platforms (contact Eric Lander, e-mail: mapmaker{at}genome.wi.mit.edu) (133, 134, 143, 144, 180) and Map Manager QT for MacIntosh computers (contact Kenneth Manly, e-mail: kmanly{at}mcbio.med.buffalo.edu). Others are listed by Paterson (179) and have recently been reviewed (156).

One approach to QTL analysis utilizes maximum likelihood techniques and calculates a statistic, the LOD score (which is the log of the ratio of the likelihood of there being a QTL present vs. the likelihood of no QTL being present at a particular map position). The LOD scores are calculated at many selected points in an interval between markers and plotted versus map position. The QTL effect enters the likelihood equations through the observed BP data and the QTL map position enters the equations through the probability of a given QTL genotype given the flanking marker genotypes for each individual and the assumed QTL position in the interval. Although the mathematical development of such methods is hardly appropriate here, it is worth presenting how to interpret such calculations and to understand their limitations.

Figure 3 gives an LOD plot for the same data given in Figure 2. The peak of the LOD plot gives the most likely location of the QTL, and the height of the peak is a measure of statistical significance. The height of the LOD peak in Figure 3 (LOD = 4.8) tests for the existence of a BP QTL. As a rough approximation, a LOD >3 has often been considered to be significant. A more precise evaluation of how to interpret LOD values, however, shows that factors such as the size of the genome, the underlying assumed genetic model for the QTL effects (i.e., additive, dominance/recessive, or unconstrained model), and the breeding paradigm (e.g., backcross or F₂ population) all influence interpretation of the LOD plot. Lander and Kruglyak (135) have made calculations accounting for such factors and give threshold values for LOD scores under various conditions. The appropriate thresholds are shown in Figure 3, where the dotted line indicates the threshold for suggestive significance (LOD = 1.9), and the solid line is the threshold for significance (LOD = 3.3) under the conditions of the particular experiment in Figure 3. Because the peak value (LOD = 4.8) in Figure 3 is greater than the threshold for significance, the data are interpreted to mean that a QTL for BP on chr 9 is segregating in the specific cross made.

View larger version (23K): [in this window] [in a new window]   Fig. 3. Logarithm of the odds favoring linkage of BP to genetic markers (LOD plot) is shown for the same data as in Figure 2. Dotted line at LOD = 1.9 is threshold for suggestive statistical significance, and solid line at LOD = 3.3 is threshold for significant linkage to BP. Construction of 1-LOD and 2-LOD support intervals to right of linkage map are shown. A congenic strain constructed by introgressing the Dahl R rat region into the Dahl S rat is shown by larger solid bar to left of linkage map; open regions at ends of the congenic bar are intervals in which recombination occurred. The congenic strain had a BP 19 mmHg lower (P < 0.0001) than S rats, implying that a minus BP allele was successfully moved from R to S strain. [From Rapp et al. (198). Copyright 1998 Academic Press.]

The position of the peak in Figure 3 gives the most likely position for the QTL, but its exact location remains unknown. The support interval defined by one LOD unit around the peak gives an ~95% confidence interval for the QTL location. For QTL of small effect, this interval is too short as the one-LOD rule yields an interval with 60-95% probability of containing the QTL (155); simulation studies suggest that support intervals based on two LOD units should be used (256). Construction of these LOD support intervals is shown in Figure 3.

The length of the confidence interval will be a function of the size of the experimental population, the magnitude of the QTL effect, and marker density. It has been shown (35) that increasing marker density beyond one marker every 5-10 cM does not significantly improve the QTL localization. The larger the phenotypic effect of a QTL, the easier it will be to detect and localize it. Most BP QTL described so far have a modest effect individually accounting for only 5-15% of the total BP variance in experimental populations. Unfortunately, this means that with experimental populations of reasonable number (200-300 backcross or F₂ rats), BP QTL are likely to be localized only to about a 20- to 35-cM interval by linkage analysis (90, 256). In actual practice, this has been the case. A 20- to 35-cM interval is much too large to attempt positional cloning of the QTL, so other techniques are needed. As will be seen in section VB, construction of congenic strains and congenic substrains using the information obtained from linkage analysis has been a popular approach. Positional cloning means the identification of loci causing a phenotype by the use of linkage analysis and study of the DNA structure and function in the chromosomal interval defined by the linkage.

In addition to the inherent statistical imprecision with which a QTL can be localized, the possibility exists that two or more QTL may be linked on the same chromosome. This may, or may not, result in multiple peaks in the LOD plot, depending on how close the QTL are and the phase of the alleles. If two QTL are far enough apart to segregate independently (e.g., 80 cM apart), then two peaks can be observed (133). If two QTL are close and their alleles are in repulsion (see Fig. 4), the effects of the two QTL may cancel each other out, and neither may be observed. If the alleles of the two linked QTL are in coupling phase (see Fig. 4), then 1) a single LOD-plot peak or a broad plateau may be observed if the QTL are close, 2) two peaks may be observed if the QTL are far apart, or 3) a "ghost" peak between the two QTL may be observed if the QTL are ~40 cM apart and the markers are widely spaced (~20 cM) (72, 161, 274). An example of a LOD plot with multiple peaks is shown in Figure 5. Obviously, this is difficult to interpret with regard to the number of linked QTL and their location.

View larger version (9K): [in this window] [in a new window]   Fig. 4. Diagram of "repulsion" and "coupling" phase for plus (+) and minus () alleles at 2 polymorphic QTL on same chromosome. Both chromosomes of a pair are shown. If chromosomes of 2 parental inbred strains (P1 and P2) to be crossed in a linkage analysis are arranged in repulsion, they are likely to have their effects cancel each other; if they are arranged in coupling, their effects may enhance QTL signal. Degree to which 2 QTL interfere with each other is a function of distance between them and relative size of their effects.

View larger version (15K): [in this window] [in a new window]   Fig. 5. LOD plot for linkage of BP to chr 1 in an F2(S × LEW) population. Dotted line is LOD threshold for suggestive linkage, and solid line is threshold for statistically significant linkage. Multiple peaks that are difficult to interpret with regard to number and location of BP QTL are illustrated. Only representative genetic markers are indicated on map. [From Garrett et al. (66). Copyright 1998 Cold Spring Harbor Laboratory.]

It needs to be emphasized that a given linkage result arising from a segregating population obtained from crossing two inbred strains is often unique to these two strains. This is because crosses of other pairs of strains may not be functionally polymorphic at a given QTL. Functionally polymorphic means that at the locus accounting for a QTL there is an allelic DNA difference that causes functional differences in expression or activity of the locus-encoded product. Obviously, if a pair of inbred strains are homozygous for the same allele at a QTL, there will be no effect of this QTL on the quantitative trait in segregating populations derived from these strains. Other considerations causing QTL results to be cross-specific include the genetic background that can influence the expression of a given QTL effect and that is unique to each pair of parental strains used to form the segregating population. Often the experimental conditions (diet, age of rats, etc.) vary between experiments, which may alter expression of a given QTL.

Recombinant inbred (RI) strains are produced by 1) crossing two inbred strains, 2) producing an F₂ population from the F₁ heterozygotes, and 3) selecting breeding pairs at random from the F₂ population, each pair to be founders for an inbred strain produced by 20 generations of brother-sister matings. The process is illustrated in Figure 6. Recombination of the parental chromosomes occurs at meiosis in the F₁ animals and in subsequent generations, but at the same time recombinant chromosomal regions are becoming fixed (homozygous) by brother-sister inbreeding. The resulting inbred strains are each a mosaic of the genetic material from the founding parental strains as illustrated in Figure 6, bottom, for a hypothetical chromosome. If a large enough panel of RI strains is produced, it can be used to produce genetic linkage maps and for detecting and locating QTL.

View larger version (21K): [in this window] [in a new window]   Fig. 6. Derivation of recombinant inbred (RI) strains from 2 parental strains [spontaneously hypertensive (SHR) and Brown-Norway (BN)] is illustrated. Unique chromosomes that are a mosaic of two founding strains are formed by recombination during early generations after F1, but these eventually become fixed (homozygous) in each RI strain due to inbreeding. One pair of SHR chromosomes is represented by open bars, and homologous pair of BN chromosomes is represented by solid bars. Markers M1, M2, and M3 at different positions on chromosome are shown. Genotypes at each marker for representative RI chromosomes illustrated are shown at bottom. S, SHR allele; B, BN allele.

Linkage information from RI strains is fundamentally similar to that in a segregating population as illustrated in Figure 6, bottom. If two loci (M₁ and M₂) are close together on the same chromosome, they are likely to have fixed the same parental allele because the probability of a chromosomal crossover in a small distance is low. Thus, in Figure 6, each of the four RI strains illustrated have alleles from the same parental strain at M₁ and M₂ (the strains are 100% concordant for M₁ and M₂). The probability that two loci on the same chromosome are concordant among RI strains decreases as the distance between the loci increases. In Figure 6, only 50% of the RI strains are concordant at loci M₁ and M₃ because the loci are far apart. Information on the concordance of a panel of RI strains at various loci can of course be used to construct genetic maps (231). It is emphasized that the larger the number of RI strains available, the more accurate the genetic linkage and that the four strains illustrated in Figure 6 are hardly sufficient. Most panels of RI strains consist of 20-30 strains.

Animals within an individual RI strain are genetically identical. One of the advantages of RI strains is that the phenotype of each strain, for example, BP, can be carefully and repeatedly studied on many individuals from each strain. The RI strains can then be used for QTL analysis. In its simplest form, the genotype of each RI strain at a marker locus is determined, and the phenotypic values of the two groups are compared by a t-test; an individual strain at any locus is either homozygous for one parental allele or the other. Following the example in Figure 6, the strains that were BB or SS at the marker of interest would have their BP compared. If the two groups of strains had significantly different BP, this would be evidence for a QTL in the vicinity of the marker used.

Pravenec et al. (185) have developed a panel of 31 RI strains derived from spontaneously hypertensive rats (SHR/Ola) and a strain of Brown-Norway (BN.lx/Cuba) carrying genes for polydactyly. This panel of RI strains is maintained in two places: Michael Pravenec, Institute of Physiology, Czech Academy of Sciences, Vídensk 1083, CS-14220, Prague 4, Czech Republic; Fax: 42 02 475 2297; e-mail: pravenec{at}biomed.cas.cz; and Morton Printz, Department of Pharmacology-0636, 9500 Gillman Drive, University of California at San Diego, La Jolla, CA 92093-0636; Fax: 619 534 0464; e-mail: mprintz{at}ucsd.edu. Figure 7 illustrates the range of BP obtained in the RI strains and the principle for the use of this RI strain panel in analysis of BP. Hamet et al. (74) showed that for a restriction fragment length polymorphism in 70-kDa heat shock protein (HSP70) the RI strains carrying the SHR allele had higher BP than the RI strains carrying the BN allele (mean arterial pressures of 143 ± 6 vs. 128 ± 3 mmHg, P < 0.02; see Figure 7). The HSP70 is located in the major histocompatibility complex (RT1) on chr 20 (see sect. VB20 for corroborating evidence for a BP QTL in this region).

View larger version (25K): [in this window] [in a new window]   Fig. 7. Association of BP and a BamH I restriction fragment length polymorphism (RFLP) in RI strains obtained from SHR and BN rats. RI strains are arranged in order of increasing mean arterial pressure. RFLP allele carried by each strain at 70-kDa heat shock protein (HSP70) locus is indicated diagrammatically below BP bar for each strain as either a 4.4- or a 3.0-kb fragment. Rats were also typed immunologically for haplotype at major histocompatibility complex (RT1) as either RT1K or RT1N. Note 100% concordance between HSP70 RFLP fragment and RT1 haplotype because they are linked closely on chr 20. (Some RFLP data are missing because some RI strains were not available for RFLP analysis.) Note also that most of strains carrying RFLP 4.4-kb fragment from BN fall to left (lower BP) and most of strains carrying 3.0-kb fragment from SHR fall to right (higher BP) of diagram. Average of mean BP of strains carrying BN allele was lower (128 ± 3 mmHg) than BP strains carrying SHR allele (143 ± 6 mmHg). [Redrawn from Hamet et al. (74).]

Recombinant inbred strains have advantages and disadvantages compared with analysis of standard segregating F₂ or backcross populations. In addition to the obvious advantage of being able to repeatedly, and thus more accurately, determine the value of a quantitative phenotype for each strain (genotype), it is also possible to study the panel of strains in different environments and thus describe genotype × environmental interactions. Also, the ability to euthanize genetically identical animals of various ages for study could reveal genetically controlled temporal effects. A disadvantage of RI strains is the cost of developing and maintaining the strains. Each panel of RI strains is of course limited to the QTL that are functionally variant between the parental strains. In QTL analysis of BP and many other traits, it has proven advantageous to study several crosses to introduce different QTL alleles and genetic backgrounds into the experiments. This is easier to do by producing F₂ populations from various parental strains than it is by developing panels of RI strains from various parental strains.

Regardless of the experimental design or the method of analysis, linkage data yield only a chromosomal region containing a gene (or genes) influencing BP that is too large for positional cloning. In addition to that, the statistical support for the existence of the BP QTL is often problematic. The limiting factors in linkage analysis are the modest phenotypic effects of most BP QTL and the limited number of chromosomal recombinant events that can be observed in segregating populations of practical size. One way around this impasse is the construction of congenic strains and then congenic substrains (136, 197).

The construction of a congenic strain is a standard procedure of experimental mammalian genetics (231) originating from the work of Snell (238) on histocompatibility loci. Figure 8 outlines the procedure for moving a polymorphic marker gene from one inbred (donor) strain to another inbred (recipient) strain. In Figure 8, the donor at marker locus M has genotype M₁M₁, and the recipient is M₂M₂. The two strains are crossed, and the F₁ heterozygotes, M₁M₂, are backcrossed to the recipient strain. The backcross offspring will segregate 1M₁M₂:1M₂M₂. Heterozygote M₁M₂ are selected and backcrossed again to the recipient strain. This procedure is repeated for at least eight cycles of backcrossing at which point two heterozygotes are intercrossed to yield offspring in the ratio of 1M₁M₁:2M₁M₂:1M₂M₂. Two offspring homozygous M₁M₁ are bred to fix the M₁ allele on the recipient background.

View larger version (24K): [in this window] [in a new window]   Fig. 8. Breeding scheme for construction of a congenic strain from 2 inbred strains, one of which is donor and the other recipient. Selection is made at microsatellite marker locus M through a series of backcrosses such that the M1 allele of donor replaces the M2 allele of the recipient. Rats are genotyped at M by PCR reaction using DNA obtained by tail biopsy. Genome of donor strain is represented by an open symbol, and genome of recipient strain by a solid symbol. Increasing shades of gray from light to dark represent increase in percentage genetic background of recipient that occurs with each backcross (BC). [Diagram from Rapp and Deng (197). Copyright 1995 Lippincott-Williams & Wilkens. Diagram is based on that by Klein (110).]

At each backcross, 50% of the loci outside of the region linked to the M locus that are segregating will become fixed for the recipient allele by chance. Thus the genetic background becomes progressively enriched for the recipient strain genes until the background is >99% recipient genes after the eighth backcross. Of course, donor genes linked to M₁ will be "dragged" along with the M₁ allele in this procedure. The congenic strain that results from selecting only one locus M will have flanking donor chromosomal segments on average equal to 100/N cM on each side (231), where N is the number of backcrosses. After eight backcrosses, the flanking DNA is on average 12.5 cM on each side of the selected marker, or 25 cM in size.

If M is a marker near a QTL, the above procedure may or may not move the donor QTL allele to the recipient strain depending on whether or not a recombination event occurred between M₁ and the donor QTL allele during any of the backcrosses. To largely ensure that a linked QTL allele is moved to the recipient strain, one can easily select for donor alleles at markers along the chromosomal region of interest, always rejecting individuals with recombination in the target region and selecting for individuals that are heterozygous for all markers in the region. Thus the strategy is to select a large region in the construction of the initial congenic strain to ensure moving the QTL allele, the position of which is known only as a statistical approximation.

An example of a congenic region moved from Dahl R to Dahl S rats is shown in Figure 3. In this case, only the marker Inha was used because the other markers were not available when the work began several years ago. Fortuitously, the congenic strain did move a low BP allele into the S strain decreasing BP 19 mmHg (P < 0.001) in the congenic compared with S rats when the rats were fed 2% NaCl diet for 24 days (198). In planning such an experiment with the availability of markers it would have been appropriate to select several markers from D9Rat4 to D9Uia10 in Figure 3.

Congenic strains can be constructed in two ways with a given pair of parental strains. The low-BP strain can be the donor and the high-BP strain the recipient, or vice versa. The main difference between the two is the genetic background. In the Dahl rats, for example, the genetic background of the R rats is not very permissive for expressing BP differences (25, 26, 195) and so most congenic strains using Dahl rats have been made on the S genetic background. On the other hand, successful congenic strains have been made in both directions with SHR and WKY (64).

Once a congenic strain is constructed that includes a QTL, the next logical step is to construct congenic substrains with smaller and smaller donor fragments to localize the QTL. One strategy for construction of congenic substrains is shown in Figure 9. The required recombinant chromosomes can be obtained by crossing the congenic strain to the donor strain to produce an F₁. Recombinant events in gametes produced by these F₁ rats can be recognized by genotyping the markers in the congenic region in rats produced from either an F₁ × recipient strain cross (i.e., backcross) or in an F₂ (i.e., F₁ × F₁ intercross). The F₂ cross is preferable because two meioses are scorable in each rat, whereas only one meiosis can be scored in the backcross. In either case, to fix a new recombinant chromosome the rat carrying that chromosome has to be backcrossed to the recipient strain to duplicate the new recombinant chromosome and to obtain male and female rats heterozygous for this chromosome that can be bred to fix the recombinant chromosome in the homozygous state.

View larger version (20K): [in this window] [in a new window]   Fig. 9. Diagram shows substitution mapping of a QTL for BP with the use of a set of congenic strains. Thick line represents a chromosome from donor normotensive strain, and thin line represents a chromosome from recipient hypertensive strain [assumed to be Dahl salt-sensitive (S) rats]. A map of informative markers (A to H) is assumed to be known along a chromosomal region thought to contain a blood pressure QTL. In the example, QTL is located between markers D and E as indicated by arrow. Strain 1 is a congenic strain in which entire chromosomal region (A to H) has been substituted into recipient S strain from the donor strain. This substituted region contains a minus QTL allele different from plus QTL allele present in recipient S strain, and thus substitution lowers blood pressure of congenic strain 1 relative to S. Strains 2-7 are congenic substrains carrying various donor fragments as indicated. If donor fragment contains QTL, then congenic substrain carrying it will have its blood pressure lowered compared with S rats. If QTL is absent in the donor fragment, congenic strain will have the same blood pressure level as parental recipient S strain. This information on the presence or absence of a blood pressure effect, combined with location of substituted chromosomal fragment, can be used to map blood pressure QTL to a smaller region. [From Rapp and Deng (197). Copyright 1995 Lippincott-Williams & Wilkens.]

Because the original congenic strain produced contains a large segment of chromosome by design (to be sure to include the QTL), it is likely that more than one round of congenic substrains shown in Figure 9 will be required to narrow the QTL location to a workable range. It is also possible (or even likely) that a QTL region contains more than one locus influencing BP. This complicates the construction of congenic substrains as the BP effect seen in the original congenic strain becomes smaller (and thus harder to identify) if it is divided between two (or more) loci in the target region. Fine mapping of the QTL to a resolution of at least 1 cM is required for positional cloning. In mice (and presumably in rats), 1 cM is equivalent to 2 × 10⁶ bp of DNA (231). One can expect ~40 genes in such a region (60), so identification of a base sequence variant that accounts for a (potentially subtle) functional change in a gene is a formidable task (28).

It is, however, easy to test individual candidate genes in a congenic region for effects on BP by making a congenic strain that includes only the candidate gene and a few centiMorgans of flanking DNA. If such a congenic strain shows no BP effect, then the candidate gene has been eliminated as the gene accounting for the BP QTL in the particular pair of rat strains involved in the original linkage test and subsequent congenic construct. This has been done, for example, with inducible nitric oxide synthase (Nos2) on chr 10 (51). Thus, if a candidate gene is in fact not "the" gene accounting for the QTL, it is easy to rule it out. It is, however, much more difficult to prove that a candidate gene in a small congenic region that is positive for a BP effect is in fact the gene of interest because 1) it becomes more and more difficult to make the congenic strain really small (<0.5 cM) around a candidate gene because a high density of markers in a smaller and smaller region is required and 2) finding recombinants in the small region means screening potentially large numbers of rats (in a 1-cM interval one recombinant is expected for every 100 products of meiosis).

In the preceding sections, strategies for defining QTL that are in use for BP analysis were given. In this section other strategies that are either in use in other areas of investigation or that have been suggested on a theoretical basis are briefly given. Additional advanced theoretical considerations are given by Darvasi (33).

As outlined in section IVC, the construction of congenic strains in rats or mice requires 2.5-3 yr of breeding. Most of this time is spent backcrossing to the recipient strain to dilute out the unwanted donor genome outside of the desired congenic region. The elimination of the donor genome can be expedited by selecting against donor strain alleles at markers strategically placed throughout the genome, at the same time one selects for donor alleles at markers in the congenic region. The term speed congenics has been coined for this procedure (136). Theoretical considerations and detailed practical advice on the construction of speed congenics have been presented (160, 270). Such methods allow the production of a congenic strain in 15-18 mo. One caveat is that in moving a large QTL segment of chromosome, recombinant events will occur frequently within the desired region, making it more difficult to find animals positive for the complete donor congenic region and negative for a large proportion of the rest of the donor genome. If, on the other hand, the congenic strain is being made by selection for one specific locus, the procedure should proceed quite well.

The suggestion has also been made (5) that the breeding cycle can be shortened by superovulating (and breeding) 3-wk-old mouse (or rat) pups followed by embryo transfer to mature females for production of the next generation. This might shorten the generation time to 6 wk and the whole congenic procedure to 1 yr. Theoretically, the advantages of superovulation and selection against the donor genome could be combined for the production of a congenic strain in under a year. Although one gains time, the work and technical expertise to implement rapid congenic strain production obviously increases.

Darvasi (32) has suggested a theoretically efficient alternate method of producing congenic strains, interval-specific congenic strains (ISCS). Given a pair of parental strains for which the broad intervals containing QTL are known, an F₂ population is produced and screened simultaneously for recombinants in all the QTL-containing intervals. In this way, recombinants are efficiently accumulated at all QTL. Congenic strains are then produced from these recombinants by backcrossing to one parental (recipient) strain with selection for the recombinant in one QTL interval and against all the other donor alleles at all other QTL. This is analogous to the procedure for speed congenics, but only the other known QTL are selected against, not the entire genome. After two or three generations of selection, donor segments at the other QTL should be eliminated, and the congenic region is then fixed in the usual manner. Of course, the production of congenic strains at each QTL proceeds is parallel.

In theory, it is possible to produce a congenic strain by a series of backcrosses but with selection for the phenotype. For example, one might cross a high and low BP strain and make repeated backcrosses to the low strain, selecting as breeders those rats with the highest BP at each generation. In such a procedure, one would expect to lose the genes with small effects on BP, but the hope is to isolate a strain with a gene(s) having a large effect. Theoretical considerations on how this might work are given by Hill (86).

In fact, such an experiment was tried once with SHR and Wistar/Lewis rats (103), but a successful congenic strain was not obtained. The work is of considerable historical interest because it was one of the earlier attempts at a rational genetic analysis of BP in SHR. Tanase (250) had described Mendelian segregation of BP in crossing SHR × Donryu rats. Construction of congenic strains on both parental backgrounds appeared successful based on BP segregation in series of backcrosses, but a definitive report on the phenotypically selected congenic strains does not exist, so by inference neither do the congenic strains exist. With the advent of molecular genetic markers, congenic strain construction by phenotypic selection for BP would seem to offer no advantages.

Recombinant congenic (RC) strains draw on properties of congenic and RI strains (37). Two inbred strains, one of which serves as the donor and the other as the background (recipient) strain, are crossed, and two backcrosses to the background strain are made without selection. At least 20 inbred strains are then developed from this second backcross population by 20 generations of brother-sister mating. Each strain will on average contain 12.5% of the donor genome and 87.5% of the background strain genome. The expectation is that some of the strains will carry different alleles at only one or a limited number of QTL when compared with the background strain. The parental strains chosen to start the RC panel may (or may not) be divergent for a quantitative trait(s) under study. In the case where the parental strains have a similar phenotype for a quantitative trait, they may nevertheless carry different sets of genes resulting in similar phenotypes. Thus, if an RC panel is developed from such strains, genetic information may still be obtained, since QTL will be segregating in the cross. This was the case in at least one RC panel used in a tumor susceptibility study (257).

Each RC strain has to be characterized phenotypically and genotyped at markers throughout the genome to identify regions containing donor alleles (68). Individual RC strains that differ phenotypically from the background strain, or from each other, can be used to produce segregating populations for linkage analysis using microsatellite markers. An individual RC strain is crossed to the background strain to produce an F₁ generation and then either backcrossed to the background strain or intercrossed to produce segregating backcross or F₂ populations, respectively, for linkage analysis. The advantages of these segregating populations over those produced from the parental strains directly is that only one (or very few) QTL are likel