Great Lakes Resource at Risk

For centuries the Great Lakes have been treated callously These five magnificent lakes—Superior, Michigan, Huron, Erie, and Ontario—located along the eastern half of the Canadian–U.S. border have served as a virtual sewer catching waste from industry, agriculture, commercial shipping, and households. Their natural barriers to other water systems have been breached, exposing indigenous ecosystems to aggressive invaders. They’ve been used as a highway for colossal ships that require deepened and broadened channels to crisscross the lakes, and that import exotic species along with their intended cargo. At times it could seem that this long-suffering water system will see no end of indignities. But recent renewed focus on the unique and tremendous value of this resource by governments and communities surrounding the lakes may turn the tide of neglect and abuse. 
 
According to the U.S. Environmental Protection Agency (EPA), the Great Lakes contain 21% of the Earth’s and about 84% of United States’ surface freshwater. That’s about 22,000 cubic kilometers of water spread over 94,250 square miles. Each year the lakes provide more than 6.7 million cubic meters of water to municipalities and quadruple that to industry. They support a commercial fishery worth about $13 million as of 2002, according to the U.S. Geological Survey (USGS), and a sport-fishing industry of nearly $1.3 billion as of 2001, according to the U.S. Fish and Wildlife Service. Today about 25% of Canadians and 10% of Americans—a total of more than 33 million people—live in the Great Lakes watershed. 
 
“The whole industrial expansion that took place during the ‘robber baron’ era [of the late nineteenth century] expanded along the shores of the Great Lakes,” says Deborah Swackhamer, a University of Minnesota professor of environmental chemistry. Soon ships were carrying iron ore, coal, and limestone from mines and quarries to steel mills and later steel to factories and products to markets. In addition to serving as a transportation system, the lakes provided a place to discharge manufacturing by-products. 
 
Unlike a sewer, however, whatever enters this lake system stays awhile. On average, less than 1% of the five lakes’ water turns over each year, which means that many pollutants stay in place. They settle in sediments, adhere to other surfaces, become suspended in water, and bioaccumulate in organisms. Similarly, with the exception of migratory birds, most wildlife in the basin spend their entire life cycle in or near the lakes. 
 
As a result of all these stressors, the lakes now house fish that are dangerous to eat, water that can be unsafe to drink, anoxic “dead zones”—areas in which virtually no plants or animals can survive—that appear each summer like clockwork, and an ever-growing population of unwanted species from other parts of the world.


Introduction
Alfalfa (Medicago sativa subsp. sativa) is one of the most important forage crops in the world. Often called the "Queen of the Forages," it is the fourth most widely grown crop in the United States behind corn (Zea mays L.), wheat (Triticum aestivum L.) and soybean (Glycine max L.). Alfalfa not only has a high nutritional quality as animal feed, but has potential to be used in soil-bioremediation, bio-fuel for the production of electricity, and for the production of industrial enzymes. An estimated 30-40% of the world's arable soils have a pH below 5.5 and therefore possess aluminum (Al) toxicity hazards (von Uexkull and Mutert 1995). Soil acidity is common to soils where rainfall is high enough to leach appreciable amounts of exchangeable bases from the soil surface layers which lower the soil pH (Brady 1974). In low pH, Al, the third most abundant element, becomes soluble in its toxic forms and is absorbed by plants. The root tips of plants show the most sensitivity to Al toxicity, inhibiting cell division and cell elongation, and thereby also inhibiting the root and shoot growth (Ryan et al. 1993;Sivaguru and Horst 1998). Alfalfa productivity is highly limited by Al toxicity related to acid soils across the world and in the United States (Rechigl et al. 1988). Even in the areas where liming is used to raise the soil pH, the subsoil may still remain acidic, reducing root growth, thus resulting in plants that are stunted and drought susceptible (Sumner et al. 1986). A cost eVective alternative is growing Al-tolerant cultivars in problem soils combined with soil amendments (Foy 1988).
In several crop species, the genetic variability for Al tolerance has been exploited to develop Al-tolerant varieties and to explore the function of the genes involved in Al tolerance. The tolerance strategies involve either a mechanism of exclusion of Al from the root apex by excretion of organic acids that chelate Al, or a mechanism that allows the plants to tolerate Al within the cells. In wheat, barley (Hordeum vulgare L) and sorghum (Sorghum bicolor L. Moench), Al tolerance was attributed to the action of a single dominant gene, involved in Al-activated malate transportation (Delhaize et al. 1993;Minella and Sorrells 1997;Magalhaes et al. 2004). However, in the wheat cv. Chinese Spring, three quantitative trait loci (QTLs) that enhanced root growth under Al stress were identiWed, suggesting that inheritance of Al resistance is polygenic . Approximately nine Al tolerance QTLs in rice (Oryza sativa L) and Wve QTLs in soybean have been identiWed (Nguyen et al. 2001;Bianchi-Hall et al. 2000). Two QTLs for Al tolerance were identiWed in Arabidopsis thaliana, both of which cosegregated with Al-activated malate release from roots (Kobayashi and Koyama 2002;Hoekenga et al. 2003).
Extensive eVort has been directed to screening and selecting alfalfa for resistance to acidic, Al-containing soil (Baligar et al. 1989;Bouton 1996;Dall'Agnol et al. 1996). Nevertheless, no M. sativa subsp. sativa cultivar or plant introduction is currently available that shows Al tolerance and does not suVer a decline in biomass under acid conditions. This lack of Al tolerance in primary alfalfa germplasm dictates the need for identifying genes or QTLs for Al tolerance in relatives of alfalfa that could be transferred to cultivated alfalfa. Genotypes with Al tolerance have been identiWed among wild diploid M. sativa subspecies (Bouton 1996) and in M. truncatula germplasm (Sledge et al. 2005a). Recent QTL mapping using a diploid alfalfa, M. sativa subsp. coreulea, a relative of cultivated alfalfa, has provided insights into the inheritance of Al tolerance. Three F 2 populations were developed from a cross between tolerant and sensitive diploid alfalfa accessions (Sledge et al. 2002). Four restriction fragment length polymorphism (RFLP) markers, in conjunction with callus growth bioassay, were found to be associated with Al tolerance using a single marker analysis. Two of the RFLP markers were conWrmed to be present in genotypes of high Al tolerance in a backcross population derived from the same tolerant accession, using both a callus growth bioassay and a soil-based assay. RFLP markers, however, are problematic for high throughput genotyping eVorts for eVectively introgressing the QTL. Polymerase chain reaction (PCR) based markers linked to Al tolerance QTL that are useful for marker assisted selection (MAS) have not yet been reported, nor have markers Xanking the QTL been identiWed in the diploid M. sativa subsp. coreulea population. In addition, since single marker analysis does not provide precise QTL locations, it is possible that the QTL can be lost through recombination between the marker and the QTL. Thus it is critical to do a whole genome coverage to identify additional Altolerance QTLs and to Xank the previously identiWed QTL using simple sequence repeat (SSR) markers to be used for eYcient marker-assisted introgression of the QTLs into commercially useful populations.
The ultimate goal underlying QTL mapping is often to identify the speciWc genes responsible for phenotypic variation. One method of doing this is the placement of candidate genes that are associated with desirable phenotypes from other species on to genetic maps to look for coincidence of map position. Several studies to identify genes associated with Al tolerance have been carried out in wheat, maize, Arabidopsis, sugarcane and other grass species. However, in Medicago, no speciWc candidate gene(s) conditioning Al tolerance have been reported. Genes involved in Al-tolerance mechanisms from other species may also be involved in tolerance mechanisms in Medicago species. Drummond et al. (2001) used this reasoning and identiWed candidate genes by prospecting the sugarcane expressed sequence tag (SUCEST) data bank for sugarcane genes with homology to known Al tolerance genes from other plant species and yeast. In this study we used the same approach to search the M. truncatula database to identify genes and DNA sequences with high homology to Al tolerance genes identiWed in other plant species, to be used as candidate genes for genetic mapping in alfalfa.
The objectives of this study were to (a) develop a genetic linkage map of a diploid alfalfa backcross population segregating for Al tolerance using PCR based SSR markers; (b) identify additional Al-tolerance QTLs and Xank the previously identiWed QTL using SSR markers to be used for MAS in alfalfa breeding; (c) to map candidate genes associated with Al tolerance from other plant species; and (d) to test for co-localization of candidate genes with mapped QTLs.

Population development and screening
The backcross population (BC 1 F 1 ) developed by Sledge et al. (2002) was used in this study. It was derived from a cross between diploid alfalfa (M. sativa subsp. coreulea) genotype PI 440501-2 (Al-sensitive) designated as AL1 and PI 464724-25 (Al-tolerant) designated as AL2 from the USDA collections. The sensitive parent was crossed as the recurrent female parent to F1 (designated as AL4) to make the Wrst backcross. The parents and the 130 backcross progenies were screened for Al tolerance by means of a callus growth bioassay by Sledge et al. (2002). BrieXy the calli from the leaf tissues of the parents and population were established in modiWed Blaydes medium with and without Al. A ratio of growth in Al relative to growth without Al was used as an Al tolerance score. The higher values for the ratios indicate increased tolerance to Al toxicity. Analysis of variance (ANOVA) was performed in this study to demonstrate the statistical diVerences between the genotypes.
A two-step approach was devised to select primers from the above sources for the polymorphism survey on the parents. In the Wrst step, bulks of the population were prepared based on the presence or absence of the RFLP marker associated with Al tolerance identiWed in Sledge et al. (2002). Three bulks were prepared for each of two RFLP markers by pooling equal amounts of DNA from Wve to eight genotypes. One bulk consisted of individuals homozygous for the positive allele of the RFLP marker, one was heterozygous, and the last was homozygous for the negative or neutral allele. The DNA bulks and the two parental lines were initially screened with a random set of SSR markers. Markers that were polymorphic among the bulks and parents were screened on the backcross population to establish linkage with the RFLP markers previously identiWed as being associated with Al tolerance. In the second approach, all of the markers that were mapped in the genetic linkage map of tetraploid alfalfa (Sledge et al. 2005b) were screened for polymorphism. A total of 314 EST-SSR primer pairs were chosen for screening the parents for polymorphism. The DNA probes for the four RFLP markers that were associated with increased Al tolerance in Sledge et al. (2002) were obtained from the UGA alfalfa genetic map (Brummer et al. (1991).

Candidate gene primers
The list of candidate genes published by Drummond et al. (2001) for Al tolerance in sugarcane was used as a guide in selecting potential candidate genes. Selected genes included those coding for proteins involved in organic acid and signal transduction pathways, and those involved in alleviating oxidative stress. For these groups of genes, the M. truncatula database was prospected to identify orthologous regions for the chosen candidate genes. The TIGR database was searched using two algorithms from the BLAST (basic local alignment search tool) family of programs, TBLASTN and TBLASTX, for amino acid or nucleotide sequence queries, respectively. Clusters were accepted as related to a particular gene when e-values fell below 10 ¡5 . Fifteen genes known to be involved in organic acid pathways, oxidative stress and pathogen defense mechanisms in response to Al stress were selected for candidate gene mapping (Table 1).
Two groups of EST primer pairs were designed from the identiWed orthologous regions of M. truncatula. Group one consisted of one primer pair for each of two genes, 'malate dehydrogenase' (Genbank accession: AF020272) and 'citrate synthase' ( Table 1). The full length genomic sequence of the M. truncatula ortholog for malate dehydrogenase and partial sequence of M. truncatula ortholog for citrate synthase were obtained from M. truncatula sequence data (http://www.mtgenome.ucdavis.edu/index.html). The primers in this group were designed based on the 'exon/intron' junction site information, inferred from the pairwise alignments (http://www.ncbi.nlm.nih.gov/BLAST/) between the alfalfa (M. sativa) ESTs, and their M. truncatula ortholog full length genomic sequences. They were designed using the Primer3 software program (http:// www.frodo.wi.mit.edu/cgi-bin/primer3.html) to anneal in exon sequences and amplify across intron regions. This strategy of focusing on non-coding regions could maximize the opportunities for detecting single-nucleotide polymorphism (SNP) variation. In the second group of primers, SSRs within each of the identiWed ESTs were found using the SSR identiWcation tool (http:// www.bioinfo.noble.org/phpssrminer/). Primers were designed using the same software to amplify fragments that contained the SSR region. Primers were designed to have an average length of 20 nucleotides, melting temperatures of 58-64°C, and theoretical PCR amplicons of 150-500 bp. A total of 68 candidate gene primer pairs were developed.

PCR reaction, genotyping and sequencing
The forward primers were synthesized with an additional 18 nucleotides from the M13 universal primer appended to the 5Ј end of the forward primer (Schuelke 2000). The M13 universal primer was labeled either with blue (6-FAM), green (VIC), yellow (NED), or red (PET) Xuorescent tags. PCR products (3 l) with diVerent fragment size and diVerent Xuorescent labels were pooled and combined with 10 l deionized formamide and 0.5 l GeneScan-250LIZ internal size standard and analyzed on an ABI PRISM ® 3730 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The reactions were visualized and scored with GeneMapper 3.7 software. The PCR were performed as explained in Sledge et al. (2005b). For sequencing, PCR products were prepared in a reaction volume of 25 l which contained 50 ng template DNA, 2.5 mM MgCl 2 , 1 £ PCR buVer II (Applied Biosystems, Foster City, CA, USA), 0.15 mM dNTPs, 1.0 pmol of each forward and reverse primer, and 0.5 U AmpliTaq Gold ® DNA polymerase (Applied Biosystems, Foster City, CA, USA). Reaction products were sequenced on the ABI PRISM ® 3730 Genetic Analyzer.

Detection and evaluation of polymorphism pattern
Primer pairs that ampliWed polymorphic fragments were visualized and scored using the GeneMapper software. The alleles generated by the candidate gene primers that were monomorphic in size were sequenced to search for SNPs. The primers that produced multiple bands were separated on PAGE gels and each band was reampliWed by PCR. All PCR products were gel-puriWed using QIAquick spin columns (QIAGEN, Valencia, CA, USA) and sequenced from both ends. The resulting genomic sequences were aligned with the corresponding EST sequences via BLAST (http://www.ncbi.nlm.nih.gov/blast/bl2seq/wblast2.cgi) to conWrm the ampliWcation of the required region. DNA sequences from the two parental alleles were aligned using the software program DNASTAR to identify SNPs.
Single-nucleotide polymorphisms were converted to cleaved ampliWed polymorphic sequences (CAPS) by identifying SNPs that showed diVerential restriction enzyme sites between the two parental lines. In cases in which a suitable restriction enzyme site was not identi-Wed, a single nucleotide mismatch was designed adjacent to the polymorphic position, such that a restriction site was created in the PCR product of one parent, but not in the other (dCAPS; see NeV et al. 1998). The PCR product was digested at 37°C for 2-4 h and the cleaved DNA fragments were analyzed on 2.3% SFR agarose gels (Genemate, ISC Bioexpress) stained with ethidium bromide.

Linkage analysis
The BC 1 F 1 population and the parents were genotyped using 177 primers, of which 165 were Xuorescently labeled primers and four were RFLP markers. For the seven CAPS and dCAPS primers, the population was genotyped using 2% agarose gel electrophoresis, and was manually scored. JoinMap 3.0 (van Ooijen and Voorips 2001) was used for constructing a linkage map. Linkage groups (LOD = 5.0) were identiWed on the basis of the assignment of known markers mapped in a previously published tetraploid alfalfa map (Sledge et al. 2005b). The data were analyzed as population type CP and map distances were calculated using the Kosambi function. Since the parents were non-inbred, four single-locus segregation ratios were tested: (1:1 3:1, 1:2:1, and 1:1$ and 1:1#) (Join-Map 3.0) using a chi-square goodness of Wt. The charts of genetic linkage and the overall chromosomal maps were drawn using the MapChart (Voorrips 2002).

QTL analysis
All the components of the QTL analysis were performed with the MapQTL 5.0 software program (van Oojien 2004). First, the non-parametric Kruskal-Wallis (KW) test was employed to detect association between markers and traits individually. In a second step, interval mapping (IM) analysis was performed to select markers signiWcantly associated with the trait to constitute an initial set of cofactors. A backward elimination procedure was applied to the initial set of cofactors.
Only signiWcant markers at P < 0.02 were used as cofactors in the multiple QTL method (MQM) (Jansen and Stam 1994) analysis for QTL detection. A mapping step size of 2 cM was used for both the IM and MQM analyses. Log of odds (LOD) thresholds for genomewide (P < 0.05) were empirically determined for the trait using the PERMUTATION test of MAPQTL with 5,000 iterations. Based on the permutation tests, a threshold LOD value of 3.1 was used to declare the presence of a QTL. Regions with a LOD score above 2.5 were considered as suggestive of a QTL.

Phenotypic data
The sensitive parent (AL1) had an Al tolerance score of 0.48 and the tolerant parent (AL2) had a score of 0.72. Among all genotypes of the BC 1 F 1 population, mean Al tolerance score values ranged between 0.32 and 1.15, with an average of 0.69 (Sledge et al. 2002).
The histogram exhibited a continuous and normal distribution consistent with polygenic inheritance, characteristic of a quantitative trait (Fig. 1). The ANOVA showed signiWcant variation (P < 0.001) among genotypes for Al tolerance score. High heritability (0.93) for Al tolerance using the callus bioassay was previously reported (Sledge et al. 2002).

Marker analysis and polymorphism
From among the 314 EST-SSR selected primers and the four RFLP markers, 162 (51.3%) detected polymorphism between the two parents. In addition, there were 68 candidate gene primers designed from 15 candidate genes selected for this study. In the Wrst group of primers (13 primer pairs) designed from intron sequences, ten primer pairs ampliWed products that ranged from 140 to 520 bp with GT-AG sequences at the intron junctions. Out of the ten intron-based primers, six of them (60%) were polymorphic. In the second group of primers (55 primer pairs) designed to amplify SSR repeat regions, 32 resulted in PCR ampliWcation with a product size that ranged from 125 to 346 bp. In this group, 12 primer pairs gave a product that was diVerent by sequence and size than what was expected.
In the second group of primers 14 out of 32 pairs (44%) Fig. 1 Distribution of the Al tolerance score evaluated using a tissue culture method across the backcross population of diploid alfalfa including the parental lines Al tolerance score No. of individuals were polymorphic. Of the 42 primers that gave a product from groups 1 and 2, nine primer pairs were mapped in the population by virtue of their inherent fragment size or presence/absence criteria, between parental alleles (Table 2). Sequence information obtained from the ampliWcation of the remaining 33 primer products from both groups showed the presence of six SNPs and four indels ( Table 2). The primers that showed an insertion/deletion pattern of polymorphism were not genotyped. Five of the SNP alleles were converted to CAPS markers by using the appropriate restriction enzyme and one SNP was converted to a dCAPS primer to obtain polymorphism between the parents. Thus, a total of 15 out of the original 68 candidate gene primer pairs, representing ten diVerent genes, were informative and were used for genotyping (Table 2). In summary, a total of 177 primers including the candidate gene primers were used for genotyping the entire BC 1 F 1 population. There were 15 markers that did not produce distinct segregation patterns in the population, and these were eliminated from the analysis.

Linkage map
Five segregation patterns, 1:1 $, 1:1 #, 3:1, 1:2:1 and 1:1:1:1 were tested for Wt in the backcross population developed from non-inbred parents. A genetic linkage map of the BC 1 F 1 population (Fig. 2) was constructed using 162 segregating loci. A total of 132 loci were placed into ten linkage groups and matched to the eight linkage groups of the previously published EST-SSR tetraploid alfalfa map (Sledge et al. 2005b). Of these 132 loci, 121 loci were placed at LOD¸5.0, and eight were placed at LOD¸3.0. There were three loci that were placed on the map in LG I at a LOD¸2.5 and 30 loci that remained unlinked. The linkage groups LG II and LG V were each represented by two distinct linkage groups. The marker assignments to the eight linkage groups were mostly similar to the tetraploid alfalfa map, except for LG III, VI and VIII. This genetic linkage map spanned a distance of 764 cM covering 84% of the estimated genome length (Chakravarti et al. 1991) with coverage ranging from 25 markers in LG I to 5 markers in LG V (Fig. 2). Coverage of the individual linkage groups was 92% (LG I), 90% (LG III), 86% (LG IIA, IV, VI, VII and VIII) and 66% for LG IIB, LG VA and VB (data not shown). SigniWcantly skewed segregation ratios (P < 0.01) were observed in 43.2% of the loci of which 48% of the distorted markers were due to the presence of an excess of Al-sensitive parent alleles and 52% of the distorted alleles were due to the presence of heterozygotes. The majority of the distorted loci were spread across LG I, LG III and LG VII. There were only three skewed markers in LG IV and LG VIII, and Wve markers in LG VI. LG II and LG VI each had two non-contiguous

95.4
LGVIII distorted markers. Out of the four RFLP markers, UGAc502 mapped to LG I, UGAc471 and UGAc191 to LG II and UGAc141 to LG VIII.

QTL associations
The non-parametric (Kruskal-Wallis) test suggested the existence of QTLs for Al tolerance in three linkage groups (LG I, LG II and LG III) using P · 0.05 as a threshold criterion for QTL detection (Table 3). Two regions were associated with the Al tolerance phenotype in LG I. One region was spread across several markers in the middle of the linkage group, extending approximately 46 cM. In addition, there were two markers at the proximal region of this linkage group that were sig-niWcantly associated (P · 0.05) with the trait. Four markers in the proximal region of LG II were signiWcantly associated with Al tolerance (P · 0.05). In the case of linkage group LG III, three markers in the proximal region of the group showed a signiWcant association (P · 0.05). The two markers in linkage group IV were less statistically supported (P · 0.05) for their association with Al tolerance. Four additional markers, one in LG V and three from the unlinked set of markers, also showed a signiWcant association with the trait (P < 0.01) using the Kruskal-Wallis test.
The permutation test to obtain empirical signiWcant threshold levels resulted in a LOD interval of 3.1 or above (genome-wide, P < 0.05) which was used to declare the presence of a segregating QTL in the interval and MQM analyses. The genetic location, corresponding LOD scores, and the variation for each QTL identiWed using the interval and MQM analysis, is shown in Table 4. Interval mapping identiWed at least four genomic regions linked to Al tolerance in linkage groups I, II and III (Table 4). LG I had two genomic regions separated by 60 cM. The LOD score for the QTL on LG II (2.86) using interval mapping analysis was slightly below the signiWcance threshold. The Multiple QTL model included background markers in the QTL analysis as cofactors. The Wrst QTL spanned a region of 50-71 cM in LG I between BE93-2 and UGAc502 and explained 37.9% of the variation using marker BE103 as a co-factor in the MQM model (Table 4; Fig. 2). This major locus on LG I increased the Al tolerance score over the sensitive parent by 17%. For the two linked markers in this region, the increased Al tolerance score was due to the absence of the allele from the sensitive parent (Table 4). A second locus aVecting the trait at the proximal region of the linkage group was suggested from the Kruskal-Wallis and interval analysis. However inclusion of the marker in this region as a co-factor in the MQM analysis revealed that this second QTL had a slightly lower LOD score (2.83) than the signiWcance threshold level (Table 4; Fig. 2). The absence of the allele from the sensitive parent for marker MTR58 and the presence of the allele in a heterozygous state for marker 122161_4 contributed to increased Al tolerance ( Table 4). The Al tolerance due to this minor QTL was increased by 12%.
In LG II, the LOD score of marker AL99 did not reach the LOD threshold value of 3.1 (genome-wide, P · 0.05), but the inclusion of this marker as a co-factor in the MQM model raised the LOD value to 3.47. The QTL associated with marker AL99 at map position 70 cM explained 16% of the phenotypic variation. The absence of the allele from the sensitive parent at this locus increased the Al tolerance score (Table 4). This QTL in LG II increased the Al tolerance by 14%. There were two potential markers associated with Al tolerance in the proximal region of LG III. Inclusion of marker AW369 alone as a co-factor in the MQM analysis revealed a signiWcant QTL in this region which explained 27% of the phenotypic variation. The QTL region, between BF220 and AW369 at 43-50 cM, 1 5.0 ** BF28 1 5.85 ** explained 26% of the phenotypic variation. The absence of the allele from the sensitive parent for marker BF220 and the presence of allele in the heterozygous state for marker AW369 contributed to higher Al tolerance in the population (Table 4). The eVect of this QTL in terms of increase in Al tolerance score was by 9%. Further analysis with marker AA05 in LG V as a cofactor in the MQM analysis did not reveal any more QTLs.

Phenotypic data collection
In this study, the Al tolerance score for the population was obtained by the callus growth assay method. The advantage of this approach is that it minimizes environmental variation and G X E interactions, which increases the eYciency of the study. Such in vitro selection must correspond to practical selection procedures, such as the soil based assay or solution culture methods. In a study to compare the cell culture and soil based selection procedures, the cell culture technique, was as eYcient as the soil assays in selecting the tolerant germplasms (Dall'Agnol et al. 1996). Sledge et al. (2002) identiWed an RFLP marker associated with Al tolerance using the tissue culture method and con-Wrmed the presence of the same marker while phenotyping using the soil assay method. The parents chosen for this study varied greatly for Al tolerance using the tissue culture method, and the population developed from these parents segregated for Al tolerance (Fig. 1). Sledge et al. (2002) reported a high heritability value (93%) for this trait using a tissue culture method, which suggests that the QTLs identi-Wed in this study could be eYciently transferred to another population by marker-assisted backcrossing procedures.

Linkage map construction
The linkage map developed in this study showed good agreement with a tetraploid alfalfa genetic map (Sledge et al. 2005b). Despite the overall similarity, there were diVerences in a few loci in LG III, VI and VIII. There were two distinct groups for linkage groups II and V. Recently we established a genetic linkage map of Medicago truncatula from an F 2 population obtained from a cross between Jemalong A17 and PI 566890 (unpublished). In our study the distinct grouping in LG II and V and the marker content in these two groups followed a pattern similar to that observed in the unpublished M. truncatula genetic linkage map. The marker density and genome coverage of this map is high for all groups except LG V. In particular, the 87-92% genome coverage of linkage groups I, IIA and III which carry the QTL, was high and therefore adequate for QTL identiWcation. The polymorphism rate in this study (51%) using the EST-SSRs was comparable to that reported for an alfalfa genetic linkage map (Sledge et al. 2005b), and a M. truncatula genetic linkage map (unpublished). In all three genetic maps, the genetic distance between the Xanking markers for a marker linked to the QTL was less than 20 cM. In a simulation study to determine the eVect of marker spacing, number of markers and population size, Darvasi et al. (1993), reported that the power for detecting a QTL is virtually the same for a marker spacing of 10 cM as for an inWnite number of markers, and was only slightly decreased for marker spacing of 20 cM.
Outcrossing species often exhibit a high level of segregation distortion, as observed in pine (Kubisiak et al. 1995), peach (Dettori et al. 2001), apple (Liebhard et al. 2002), willow (Hanley et al. 2002) and rose (Yan et al. 2005). Diploid alfalfa, an out-crossing species, carries a high proportion of heterozygous loci, and when inbred, shows evidence of inbreeding depression. Previous genetic mapping studies in diploid alfalfa reported distorted segregation (15-63%) due to the excess of heterozygotes that survived in F 2 segregating populations (Brummer et al. 1993;Kiss et al. 1993;Kalo et al. 2000) or a backcross population (Echt et al. 1994). Thus, the high level of segregation distortion observed in this study (43%) may be related to the association between heterozygosity and plant vigor in alfalfa. The proximal portion of LG I, LG VII and a major portion of LG III had markers showing signiWcant segregation distortion.

QTL detection and comparison
This study identiWed at least three putative QTLs in linkage groups I, II and III. The most important QTL with the largest eVect (R 2 = 38%) was observed in LG I. In this linkage group, the QTL was centered on marker BE103 and spanned 20 cM. Using interval mapping, multiple markers in this region (BE90, BE103 and BE93-2) were signiWcantly associated with Al tolerance, with LOD scores between 3.12 and 5.47, well above the threshold level of LOD 3.1. Using MQM mapping, however, a single major QTL inXuencing the trait was revealed in this region (Fig. 2). In the proximal region of this linkage group, a second QTL associated with two markers (MTR58 and 122161_4) was identiWed using interval mapping analysis. With MQM mapping, however, the QTL exhibited a LOD of 2.83, slightly below the threshold level. It could be that the gap of approximately 30 cM and the segregation distortion in this region may have caused underestimation of the eVects and that the mapping of more markers in this region may better resolve this QTL. The presence of multiple loci in the same linkage group does not strictly imply the presence of multiple QTLs; however, the possibility of the presence of multiple QTLs cannot be ruled out in this study. Sledge et al. (2002) had earlier identiWed the RFLP marker (UGAc502) in a segregating F 2 population associated with Al tolerance trait using a single marker analysis which was conWrmed in the backcross population. The BC 1 population used in this study was derived from the same parents of that used to develop the F 2 population. It is noteworthy that the same RFLP marker UGAc502 lies proximal at a distance of 13.4 cM to the most sig-niWcant marker BE103 in this QTL region and slightly outside the QTL region and signiWcant using single marker analysis. Another marker, BE90, maps at a distance of 6.8 cM proximal to BE103, and is signiWcantly linked to this QTL. A second signiWcant QTL located on the proximal portion of LG II accounted for 16% of the phenotypic variation. The likelihood of the presence of this QTL (LOD = 3.47) using interval-mapping analysis was slightly below the threshold level. However, the sparse coverage around the QTL region may have underestimated the eVects of this QTL. Better marker coverage in this region could help to resolve the QTL and its eVect. The RFLP marker, UGAc471, previously reported to be loosely linked with Al tolerance and UGAc191 (Sledge et al. 2002) maps 33 cM away from this QTL and is not linked to Al tolerance trait. The third most signiWcant QTL was identiWed on linkage group III, between markers BF220 and AW369, and explained 27% of phenotypic variation. There were no prior reports of the existence of an Al tolerance QTL on this linkage group. On linkage group V, one marker was highly signiWcant using the Kruskal-Wallis test, but the interval mapping analysis did not support it. However linkage group V had consisted of two linkage groups, and marker coverage was less than for the other groups.
EVorts to saturate this linkage group using markers from other genetic linkage maps, such as that of tetraploid alfalfa (Sledge et al. 2005b) have thus far been unsuccessful. Linkage group IV had two markers that showed a signiWcant eVect using the Kruskal-Wallis test. Marker-trait associations were not detected with other analyses, even though the genome coverage in the linkage group was high (86%), with 14 markers. Thus, the results from the Kruskal-Wallis test could be a false positive in LG IV. A signiWcant association was observed between Al tolerance and three unlinked markers using the Kruskal-Wallis test of signiWcance (Table 3). These, three markers, MTIC 447, AL108 and BE28, map to LG I in the tetraploid alfalfa linkage map (Sledge et al. 2005b). It is possible that these three markers could be linked to LG I if additional markers are added to this group.

Candidate gene mapping
The co-localization of candidate genes with a QTL can support the hypothesis that the gene is responsible for variation in the phenotype. However, this approach is limited by the understanding of the physiology and biochemistry of the trait of interest and by the previous identiWcation of genes potentially involved in trait expression. In our study we used the extensive sequence data available for M. truncatula to choose three groups of candidate genes that have been associated with Al tolerance in other plant species. The three groups included the set of genes that coded for proteins responsible for organic acid synthesis, genes involved in signal transduction, and genes that code for enzymes that alleviate oxidative stress. Choi et al. (2004) developed intron-targeted primers for EST-based markers by aligning the Medicago truncatula ESTs with the genomic sequences of Arabidopsis orthologs to take advantage of the DNA polymorphisms in introns. We adopted a similar strategy for two M. sativa genes, cytosolic malate dehydrogenase and citrate synthase, using their EST alignments with genomic sequences of M. truncatula. Of the few organic acids produced by plants in response to Al stress, citric acid has the highest binding capacity, followed by malate, oxalate and succinate (Hue et al. 1986). Al tolerant lines of wheat and corn release citric and malic acid in response to Al stress Zhang et al. 2001). Sasaki et al. (2004) cloned the wheat gene Al-activated malate transporter (ALMT1) using the near-isogenic wheat lines that diVer in Al tolerance. Kochian et al. (2005) summarized several reports showing high levels of Alactivated release of carboxylates correlated with Al resistance in a large number of plant species. When all the evidence in support of Al-activated root carboxylate release as a major resistance mechanism is examined, a very strong case in support of this concept is seen. For this reason, the closest M. truncatula ortholog was used to infer intron position for malate dehydrogenase and citrate synthase, and thereby aid primer design.
The full-length genomic sequence for the malate dehydrogenase gene and a partial sequence for the citrate synthase gene were available. For primers designed from intron/exon junction site information, it was anticipated that PCR primers designed to anneal in conserved exon regions and to amplify across the more highly diverged intron regions would result in a higher rate of polymorphism. Choi et al. (2004), while analyzing 47 EST markers, found that the intron targeted mapping strategy was an eYcient method to map transcribed genes because of a higher polymorphism rate. Wei et al. (2005) reported 63% polymorphism rate using an intron based marker strategy to develop a Rhododendron genetic linkage map. We developed a small set of intron based primers which exhibited a high (60%) rate of polymorphism. Two of the intron based primers designed from the sequences of the M. truncatula ortholog for malate dehydrogenase mapped to LG I. One of them, 122161_4, co-localized with a putative QTL that was identiWed using the interval mapping analysis, but which fell slightly short of the threshold values when using the MQM analysis (Fig. 2). Although the candidate gene marker mapped to a second locus on LG I, proximal to the major QTL, it is worth examining this region thoroughly in order to conWrm that the QTL resides in this region. Similar results were found in Wne-scale mapping of a QTL for Al tolerance in A. thaliana in which a gene named AtALMT1 was found located proximal to the major QTL on chromosome 1 (Hoekenga et al. 2006). The physical and genetic location for AtALMT1 while not consistent with the location for the principal Al tolerance locus, indicated that AtALMT1 is an essential factor for Al tolerance in Arabidopsis but does not represent the major Al tolerance QTL.
Another candidate gene, 149493_2, also an intron based marker designed from the sequences of M. truncatula ortholog encoding for malate dehydrogenase, mapped at the distal end of the linkage group and showed a signiWcant association (P · 0.01) with the Kruskal-Wallis analysis (Table 3). However, this marker did not reach signiWcance in the interval mapping analysis. The other candidate genes mapped to other linkage groups (LG IV, V and VI) but were not associated with Al-tolerance in our population. In addition, six of the candidate gene primers remained unlinked and were not associated with Al tolerance in this population. Although the most compelling evidence has focused on a resistance mechanism based on chelation and exclusion of extracellular Al via Al-activated root organic acid release, a number of other potential Al-resistance mechanisms have begun to receive attention. Mechanisms such as Al-chelating ligands and internal detoxiWcation of Al (Ma et al. 1997;Ma and Hiradate 2000), Al-induced alkalinization of rhizosphere pH (Degenhardt et al. 1998), and root exudation of phenolic compounds (Kidd et al. 2001) need to be investigated.

Conclusion
This study revealed three genomic regions in LG I, LG II and LG III in wild diploid alfalfa that increased the Al tolerance score using a tissue culture assay. The putative QTLs identiWed in this study are Xanked by PCRbased SSR markers, which can aid in marker assisted transfer of these QTLs to cultivated alfalfa. These QTLs are being isolated in near isogenic lines by crossing and backcrossing to tetraploid alfalfa, and selecting for the lines carrying the QTL using Xanking markers. The colocalization of a candidate gene marker designed from malate dehyrogenase with a putative QTL suggests that this gene may be responsible for part of the variation in Al tolerance observed in this study. The association between candidate genes and Al tolerance in this mapping population are only suggestive, however, and require more rigorous validation in a cultivated alfalfa background to evaluate their utility.