The inheritance of 672 AFLP markers was determined in 28 progeny clones derived from two crosses between P. c. chabaudi AJ and either clone AS (3CQ) or AS (30CQ). The majority of the AFLP markers showed independent assortment in the 28 progeny clones, as illustrated previously [11]. However, 66 markers failed the chi-square test at 5%, 15 of which failed it after the Bonferroni correction. Markers were allocated to linkage groups using the Map Manager program and groups assigned to chromosomes using 42 previously mapped RFLP markers as anchors [1]. Estimated numbers of recombination events, genetic lengths of chromosomes and recombination frequencies were also determined for the identified chromosomes using Map Manager.
Allocation of markers to linkage groups
The 672 AFLP markers formed a total of 22 linkage groups with a final p-value of 0.001. Additional file 1 summarises the numbers of AFLP and RFLP markers assigned to each chromosome or to unassigned linkage groups, the estimated physical size of each chromosome [19] and the number of AFLP markers per Mb. 400 AFLP markers in 10 linkage groups could be assigned to P. c. chabaudi chromosomes 1 and 5–13, by virtue of their linkage to RFLP markers previously assigned to specific chromosomes by physical mapping [1]. 272 AFLP markers could not be assigned to a specific chromosome. 214 were placed in 12 unassigned linkage groups, each with between 2 and 51 AFLP markers. At least four of these linkage groups are likely to map to chromosomes 2, 3, 4 or 14. The failure to assign these linkage groups occurred because RFLP markers previously mapping to chromosomes 2, 3, 4 and 14 were not allocated to linkage groups. This was probably due to insufficient characterization of the inheritance patterns of these RFLP anchors which were determined in a small number of recombinant clones [1]. For instance, the inheritance of a RFLP marker assigned to chromosome 2, Ca-ATPase, was only determined for 7 out of the 28 recombinant clones. No independent physical mapping of unassigned linkage groups was attempted here.
58 AFLP markers, 21 of which segregated in a non-random fashion, could not be allocated to any linkage groups. Physical mapping of these markers would be required to assign them to specific chromosomes. Alternatively, unassigned linkage groups or unallocated markers might map to the small mitochondrial or apicoplast genomes, although these combined represent only 0.2% of the genome. Several RFLP markers could not be allocated to linkage groups by the Map Manager software, probably due to the small numbers of clones analysed for these markers. These markers were added to assigned linkage groups according to their chromosomal assignment, as previously determined by physical mapping [1].
With the exception of chromosomes 2, 3, 4 and 14 (discussed above), chromosomes 9 and 10 showed the lowest density per Mb of AFLP markers. Chromosome 7 showed the highest density. This may simply reflect natural variation in the frequency of AFLP polymorphisms on particular chromosomes. However, for chromosomes with a low apparent density of AFLP markers such as chromosomes 9 and 10, it is likely that some of the markers in unassigned linkage groups would physically map to these chromosomes. These unassigned groups may not show genetic linkage with (groups of) assigned markers because of factors such as a high rate of recombination between two linkage groups (one linked to the RFLP anchor) or because of an intervening region with a low density of AFLP markers. Both factors, or a combination of the two, may prevent two physically linked groups from being identified as genetically linked. Conversely an apparent unusually high frequency of AFLP markers (as in chromosome 7) may arise from strong linkage disequilibrium between loci on two different chromosomes. Some markers located on one chromosome may thus appear to be genetically linked to markers on another. This could arise where one locus exerts a strong constraint on another unlinked locus. For instance, the AJ allele of an enzyme in a metabolic pathway may only function with the presence of the product of the AJ allele encoding another enzyme in the same pathway. This constraint might be structural or functional. The same may be true of AS alleles of the same enzymes. In this case, the genes encoding these enzymes, and markers strongly linked to them, may appear in the same genetic linkage group.
Order of markers in the linkage groups
AFLP markers were initially ordered on the linkage groups as described in Materials and Methods. After inspection of the predicted marker order, occasional manual adjustments were made to correct markers which appeared to be inappropriately positioned.
Because Map Manager failed to allocate some RFLP markers to an assigned linkage group, these were positioned manually. The final distribution of the markers on the 10 linkage groups assigned to chromosomes is shown in Fig. 1, 2, 3 (See also Additional file 2, Additional file 3. and Additional file 4). 7 unassigned linkage groups containing 9 or more markers each are shown in Fig. 4 (see also Additional file 5).
Number of recombination events per chromosome
If markers and recombination events were both uniformly distributed across the genome, then we would expect the number of predicted recombination events (totalled from 28 clones characterised here) in each chromosome to correlate with its physical size. The predicted total number of recombination events occuring in each linkage group is shown in the Additional file 1. Uniformity was evaluated by comparison of the frequency of recombination events (from the 28 clones) in each chromosome. This varies between 6.7/Mb (chromosome 10)) and 31.1/Mb (chromosome 7), with an overall value of 13.3/Mb. Chromosomes 1 and 11 also showed low frequencies. These differences may reflect natural variation in recombination rates across the genome. However, other factors may also contribute. For instance, there is likely to be a systematic underestimation of recombination frequency because the physical extent of the linkage groups assigned to particular chromosomes will be less than their actual size. For instance, if the chromosome 10 linkage group extends across only half of the chromosome, then the real density of markers and recombination events (per Mb) is likely to be about twice the apparent value given. Indeed, Additional file 1 shows that when data from the unassigned linkage groups are included, the recombination frequency (across the whole genome) increases to 15.9 events per Mb. Regardless of the number of markers assigned to each chromosome, we would expect the number of recombination events per AFLP marker to remain relatively constant, if both the frequency of polymorphism and the rate of recombination vary little between chromosomes. This is indeed the case. For the different chromosomes, the measure varies only between 0.4 and 0.9 recombination events per AFLP marker (see Additional file 1).
Assuming that the P. c. chabaudi genome consists of about 20 Mb [19] the value of 13.3 recombination events in 28 recombinant clones/Mb converts to about 9.5 recombination events/clone/genome which is very close to the value of 10 estimated for Plasmodium falciparum [21].
A significant number of double crossover events around a single AFLP marker were observed in many linkage groups. Any such occurrences were re-evaluated on the original X-ray film to detect any possible errors or ambiguous bands. Many distinct double crossover events were observed. The same phenomenon was also commonly observed in P. falciparum, and were interpreted as being due to non-reciprocal conversion events [20].
Genetic length of linkage groups
The apparent genetic length of each linkage group in cM was calculated by Map Manager based on the number of recombination events (Additional file 1). When added together, the linkage groups assigned to chromosomes combined to give a total genetic length of 1180 cM and the unassigned linkage groups a further 497 cM, giving a total for the genome of 1676 cM. Due to the limited number of clones available (28), the smallest genetic distance that could be measured between two markers was approximately 3.6 cM, corresponding to the presence of a single recombination event between two markers in 28 clones. In general, the estimated genetic lengths of the linkage groups assigned to chromosomes 1, 5–13 increased with the estimated physical sizes of the chromosomes (Figure 5). The Pearson correlation coefficient was 0.794 (p < 0.005). The estimated sizes of map unit for chromosomes 1 and 5–13 are shown in the Additional file 1. These values varied from 8.9 kb/cM (chromosome 5) to 24.1 kb/cM (chromosome 11) with an overall estimated mean of 15.1 kb/cM. It is likely that there is some overestimation of physical size of a map unit because individual identified linkage groups are unlikely to cover the full extent of any chromosome. Indeed, when the genetic lengths of the unassigned linkage groups are included in the analysis, the overall map unit size is reduced to 13.7 kb/cM. The inclusion of additional unallocated markers may reduce this value even further.
However, some overestimation of genetic length in linkage groups is also possible, which leads to an underestimation of map unit size. For instance, Figure 2 shows that chromosomes 9 and 10 both show two abnormally large sections bounded by RFLP anchors without intervening AFLP markers. Specifically, chromosome 9 shows, at one end, an RFLP marker, ran, 40 cM from its nearest AFLP marker. At its other end, an RFLP marker, Ag 3027, lies about 35 cM from its nearest AFLP marker. Chromosome 10 has RFLP marker cDNA121 about 70 cM from its nearest AFLP marker and RFLP marker, 5S rRNA, a further 70 cM distant. These large gaps may be artefacts which arise for two reasons. Firstly, some unreliability in the typing of clones using RFLPs was previously noticed [4]. Secondly, the inheritance patterns of these markers were not determined in all 28 recombinant clones. Markers ran, Ag3027, cDNA121 and 5S rRNA were typed for only 9, 17, 9 and 16 clones, respectively. The characterisation of inheritance of RFLP markers in all of the 28 progeny clones, and the correction of possible mistakes may reduce the estimated genetic length. This would lead to an increase in map unit size. Nevertheless, it is notable that the value reported above (13.7 kb/cM) is close to estimates made for P. falciparum (15–30 kb/cM [21] or 17 kb/cM [20]), although slightly smaller. It is likely that the recombination rate may vary within as well as between chromosomes or genomic loci [22].
Estimate of potential alleles due to indel mutations
Of the 400 AFLP markers placed on chromosomes, 37 AS-AJ pairs (74 markers i.e. 18% of the total) shared the same selective bases at both primer ends and showed complementary segregation in the cross-progeny clones. Most of these markers also differed in size by only a few base pairs. They are likely to be alleles at the same loci. This was confirmed by sequencing two such pairs, namely ASTA01AC and AJTA01AC (chromosome 5), and ASTT02CA and AJTT02CA (chromosome 13) (sequence data not shown). This suggests that a significant proportion of the polymorphisms observed between AS and AJ may be due to small insertions or deletions. In fact, it was observed that small indels tend to occur in introns or intergenic regions (data not shown).
Reliability of the AFLP markers in the progeny clones
A few AFLP markers which were originally identified between clones AJ and AS [11] were not found in the progeny clones. Also, a few bands appeared in the progeny clones that were absent in the parents. All of these markers were ignored during the generation of the linkage map. It is possible that such markers could indicate genetic re-arrangements in the drug-resistant clones AS (3CQ) and AS (30CQ). However, they did not segregate with chloroquine resistance phenotype (data not shown). Some other markers were difficult to investigate because of their proximity to other bands or their location at the bottom of the gel, where bands tend to be fuzzier and more difficult to interpret.
Effect of typing mistakes in the markers
Ongoing work on linkage between chloroquine resistance and markers on chromosome 11 [4] suggested that AFLP and/or RFLP markers were occasionally incorrectly characterised in one ore more of the 28 clones. To test the effect of incorrect typing in AFLP markers, some deliberate mistakes were introduced by changing the parent from which a particular marker was inherited in a particular recombinant clone. The effect of such changes ranged from the appearance or disappearance of predicted double-crossover events and consequent change in the estimated genetic length, to a larger scale change in the order of markers within a linkage group. Occasionally markers were reallocated to a different linkage group. It was concluded that patterns of linkage may be sensitive to errors in genotyping individual clones.