| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
ANIMAL GENETICS |
* Department of Agricultural, Food and Nutritional Science, University of Alberta, Edmonton, Alberta, T6G 2P5, Canada;
and
Igenity Livestock Production Business Unit, Merial Ltd., Duluth, GA; and
Agriculture and Agri-Food Canada, Lacombe Research Centre, Lacombe, Alberta, T4L 1W1, Canada
 |
Abstract |
|---|
 Top Abstract INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
Key Words: beef cattle • feed efficiency • feed intake • quantitative trait loci • residual feed intake
|
INTRODUCTION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
). It is also important for the environmental viability of beef production because an improvement in feed efficiency reduces methane production (Nkrumah et al., 2006; Hegarty et al., 2007).
One measure of feed efficiency is feed conversion ratio (FCR), which is the ratio of feed intake to BW gain and is therefore highly correlated to the growth of the animal. An alternative measure of feed efficiency is residual feed intake (RFI), which is the difference between the actual feed intake of an animal and its predicted feed intake based on growth and BW of an animal (Koch et al., 1963; Archer et al., 1999). This makes RFI phenotypically independent of the traits used to measure it and may therefore reveal variations in metabolic processes that determine efficiency (Kennedy et al., 1993; Archer et al., 1999; Arthur et al., 2001b).
The moderate heritability of RFI and the high costs of measuring feed intake to calculate RFI make it a target for genetic improvement through marker-assisted selection (Archer et al., 1999; Liu et al., 2000; Arthur et al., 2001a). Two studies have made progress in identifying genetic markers associated with the traits. A whole-genome association study has identified SNP throughout the bovine genome with effects on RFI (Barendse et al., 2007). A preliminary genome scan for QTL identified QTL for RFI, FCR, and feed intake (Nkrumah et al., 2007b), but had low marker coverage and consequently poorly defined QTL regions. Therefore, the objective of this study was to map the QTL using a greater density of markers. The RFI QTL results of BTA 2, 5, 10, 20, and 29 have been reported previously, and SNP associated with RFI have been identified within those QTL regions (Moore et al., 2006; Sherman et al., 2008). Therefore, these chromosomes will not be mentioned further.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
). Animals and Phenotypic Data
A total of 400 steers from 20 sire families with an average of 20 progeny per sire and a range of 8 to 56 progeny per sire were used in this analysis. Other available animals from families with fewer than 8 progeny were not used. The sires used in this study were Angus, Charolais, or University of Alberta Hybrid bulls. The dams were produced from 3 composite lines, namely Beef Synthetic 1, Beef Synthetic 2, and Dairy x Beef Synthetic (Goonewardene et al., 2003). Steers were produced over 3 yr from multiple-sire breeding groups on pasture. The sire of each calf was later determined using a panel of microsatellite markers.
Phenotypic data were collected over 3 yr from feedlot tests using the Growsafe automated feeding system (Growsafe Systems Ltd., Airdrie, Alberta, Canada; Basarab et al., 2003) at the University of Alberta Kinsella Research Station. The test procedures and diets used are described elsewhere (Nkrumah et al., 2007a). Briefly, animals were, on average, 252 (SD = 42) d old with BW of 353 (SD = 61) kg at the start of test. Tests were done in 6 batches of 80 animals over 3 yr. The diet in yr 1 was composed of 80% dry-rolled corn, 13.5% alfalfa hay pellet, 5% feedlot supplement, and 1.5% canola oil, which was 88.9% DM supplying 2.90 Mcal of ME/kg and 12.5% CP. The test diet in yr 2 and 3 contained 64.5% barley grain, 20% oat grain, 9.0% alfalfa hay pellet, 5.0% beef feedlot supplement (32% CP beef mineral supplement containing 440 mg/kg of monensin, trace minerals, and vitamins), and 1.5% canola oil, which was 90.5% DM supplying 14.0% CP and 2.91 Mcal of ME/kg. The diet in yr 1 was different because of a barley grain shortage.
The traits considered in this study include RFI corrected for production, daily DMI, FCR, and RFI corrected for ultrasound backfat thickness in addition to production (RFIbf). Average daily gain, metabolic BW (MWT), and ultrasound backfat thickness (BF) were also used in this study for the calculations of RFI and RFIbf. Body weight measurements were taken weekly, and a linear regression against days was used to calculate ADG and MWT (midtest BW0.75). Measurements of BF were predicted using a linear regression of measurements gathered every 28 d with an Aloka 500V real-time ultrasound (Overseas Monitor Corporation Ltd., Richmond, British Columbia, Canada). The daily DMI was calculated from the total feed intake over the test period. Residual feed intake was calculated as the actual DMI of each animal minus the expected feed intake, which is predicted using regression of DMI on ADG, MWT, and test batch (Arthur et al., 2001b). The calculation for RFIbf was the same as for RFI, except it also included BF as an independent variable in the regression (Basarab et al., 2003; Schenkel et al., 2004).
Genetic Markers
The genetic markers used in this study included 101 microsatellites and 2,093 SNP that covered the 24 BTA, with an average coverage of 91 markers per chromosome. The average spacing between markers was 1.01 cM but the median was 0.40 cM. The range of spacing was from 0 to 17.3 cM with 14 gaps greater than 10 cM. The locations of the markers were obtained from a composite map produced from 2 genetic maps and 3 whole-genome radiation hybrid panels (Snelling et al., 2007). Microsatellite maker genotypes were determined by an automated fragment analysis using the ABI Prism 377 and ABI 3730 DNA sequencers (Applied Biosystems, Foster City, CA). Genotyping of the SNP was performed using the Illumina GoldenGate assay on the BeadStation 500G Genotyping System (Illumina Inc., San Diego, CA), which simultaneously genotypes 1,536 SNP loci on 96 samples (Oliphant et al., 2002).
QTL Analysis
The QTL mapping method used in this study was the multiple-marker interval mapping approach for half-sib families as described by Knott et al. (1996). This method was also used by Nkrumah et al. (2007b) in beef cattle, by Spelman et al. (1996) and Schnabel et al. (2005) in dairy cattle, and by de Koning et al. (1999) in pigs. This study used 20 paternal half-sib families, and the detailed QTL mapping method is described by Nkrumah et al. (2007b). Briefly, this method uses a conditional probability for each animal of inheriting a putative QTL allele from its sire, which was calculated at 1-cM intervals along each chromosome based on the informative flanking markers. These probabilities were obtained from the online software QTL Express (http://qtl.cap.ed.ac.uk/; Seaton et al., 2002). The analysis was performed across families with the QTL effects nested within each sire family, because the sires are assumed to be unrelated; therefore, the linkage phase between the marker and QTL alleles can differ between families. A mixed model was used with a single fixed QTL effect and random sire effect, which allows one to account for the expected covariances among paternal half-sibs (Nagamine and Haley, 2001; Van Eenennaam et al., 2007). The mixed model used was
 |
where Y is a vector of observations on the progeny of each sire, X is a known incidence matrix relating observations to their fixed effect levels, β is a vector of fixed effects, which included test batch and age of animal, G relates observations to sires, s is a vector of random additive polygenic effects of sires, Q is a vector of the conditional probabilities, at each interval, that a calf inherited the first allele of a putative QTL from the sire based on the flanking marker information, 
is the regression coefficient corresponding to the fixed allele substitution effect for a putative QTL within half-sib family, and e is a vector of random residuals. This mixed model analysis was conducted using SAS software (SAS Institute Inc., Cary, NC) to obtain an F-statistic at each position (cM). The location with the largest F-statistic was considered to be the most likely location of the QTL (Spelman et al., 1996). The QTL results within individual families were also obtained for the traits on chromosomes showing an across-family QTL to determine which sires were potentially segregating for the QTL.
Significance thresholds were determined using permutations using a modified version of the method described by Churchill and Doerge (1994) and was performed using SAS as described by Nkrumah et al. (2007b). The chromosome- and genome-wise thresholds were determined using 20,000 permutations per chromosome at the 5, 1, and 0.1% levels. Because the objective of this study was to fine map the QTL from the previous genome scan for feed intake and efficiency (Nkrumah et al., 2007b), in this study we chose to report QTL above the 5% chromosome-wise threshold level rather than the 10% level that the previous genome scan used. To obtain an estimate of the width of a QTL, a 5% chromosomal threshold interval was determined using regions with F-statistics above the 5% chromosomal significance level surrounding the QTL location. Within-family QTL results were only reported for across-family QTL >5% genome-wise threshold.
|
RESULTS AND DISCUSSION |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
The most likely positions for each QTL, the chromosome- and genome-wise significance levels, and the QTL interval exceeding the 5% chromosome-wise significance threshold for the across-family analysis are shown in Table 1. Both chromosome-wise and genome-wise significance levels were calculated from 20,000 permutations per chromosome. In total, on the 24 chromosomes presently fine mapped, 53 QTL of the 4 traits were detected at the chromosome-wise threshold (at least P < 0.05); among these, 4 QTL exceeded the genome-wise threshold of P < 0.001, 3 exceeded at P < 0.01, and 17 at P < 0.05 (Table 1). No QTL were detected on BTA 8, 16, and 27 above the 5% chromosome-wise significance threshold for any of the traits. Examples of across-family QTL F-statistic profiles on BTA 3, 7, 11, and 24 are depicted in Figure 1.
|
|
). Nineteen QTL regions were identified for RFIbf at the chromosome-wise threshold of 5% on 18 chromosomes. Of the 19 QTL regions, 9 reached the 5% genome-wise threshold and 2 (BTA 1 and 3) reached the genome-wise significance level of P < 0.001. Both DMI and FCR are highly correlated with RFI [rp = 0.64 (phenotypic correlation) and rg = 0.73 ± 0.18 (genetic correlation) for DMI, rp = 0.52 and rg = 0.62 ± 0.09 for FCR; Nkrumah et al., 2007a], but both DMI and FCR had fewer QTL than RFI, showing that, although they are genetically correlated, they do not share all of the same QTL locations. For FCR there were 12 QTL on 12 chromosomes above the chromosome threshold, 2 of which were above the genome-wise threshold on BTA 13 (P < 0.05) and 24 (P < 0.01; Figure 1D). Four DMI QTL were detected above the chromosome threshold and only 1 on BTA 7 was above the genome-wise threshold (P < 0.001; Figure 1B). Within-Family QTL Analysis
The across-family analysis allowed us to pool many half-sib families together to estimate the location of QTL within the population. However, this means that the sires used could be both homozygous and heterozygous for the QTL, and they are pooled together, which can reduce the ability to detect QTL if only a small number of sires are heterozygous for the QTL (Hiendleder et al., 2003). Additionally, because the sires used are assumed to be unrelated, it is possible that they carry different QTL alleles, which will also affect the across-family results. Moreover, the QTL location and size can be influenced by a few families that are highly significant for the QTL or by many families contributing moderate to weak effects to the QTL (Nkrumah et al., 2007b). For these reasons it is important to examine the within-family results in addition to the across-family results. As an example, the within-family results for the QTL of RFI and RFIbf on BTA 3 and 11, DMI QTL on BTA 7, and FCR QTL on BTA 24 are shown in Tables 2 to 5. The other within-family results for QTL above the genome-wise threshold of P < 0.05 are available upon request. For each table, only the families that are segregating for the QTL (P < 0.15) are shown. In the across-family portion of the table, the estimate of the QTL effect within each family at the across-family QTL location is shown. In the within-family portion of the table, the most likely position of the QTL within the individual families is shown along with the estimated QTL effect.
|
|
|
|
), in which sire families 2 and 3 are very significant at the location of the QTL (82 cM), but in which 8 other families are not significant at 82 cM but are segregating at other locations along the chromosome ranging from 1 to 107 cM. Because more individual tests are done to perform the within-family QTL analysis, it is also possible that some of these QTL at other locations may not be real. For the FCR QTL on BTA 24 (Table 3) the opposite is true, where multiple families (6 of the 9 families) segregating on this chromosome have the level of significance of P < 0.05 around the QTL location of 59 cM, and all contributed to the consensus QTL locations as determined by the across-family analysis. The within-family analysis can also influence the width of the QTL. For example, on BTA 24 for FCR, the chromosome-wise 5% interval of the across-family analysis stretched from 43 to 78 cM with the greatest point at 59 cM. When examining the within-family QTL positions, it was found that the families with the significant QTL had locations that ranged from 0 to 78 cM with only 3 families (sires 2, 10, and 18) being within a few centimorgans of the across-family QTL position. This could be because the marker heterozygosity for the sires differs, and, therefore, the QTL location would change. Another possibility is that the sires are assumed to be unrelated, and, therefore, might be segregating for different QTL. Furthermore, the opposite direction of the QTL effect shows that the linkage phase between the markers used and the QTL might be different. Conversely, the DMI QTL on BTA 7 (Table 4) has a narrow interval of 53 to 59 cM, which is again reflected by the within-family positions, where 7 of the 10 families segregating for this QTL are clustered around 54 cM in the same location as the QTL (range 53 to 59). It is of interest to note that even though these families are around the same position, 2 of the families (sires 12 and 19) show a negative QTL effect, whereas the other 5 show a positive QTL effect, again revealing the issues of phase between markers and QTL differing in families. Another possible issue may arise because of some of the smaller families (fewer than the average of 20 progeny) having QTL effects that are overestimated, which would cause a greater incidence of false positives. Comparison of RFI with RFIbf
It has been suggested in earlier studies that RFI may have a correlation with BF and that selection for RFI may have a correlated response to BF (Arthur et al., 2001a; Crews et al., 2003) so it has been suggested that RFI can be corrected for ultrasound backfat thickness (Basarab et al., 2003; Schenkel et al., 2004). This prompted us to include RFI corrected for BF (RFIbf) in this study to reveal if a difference exists between RFIbf and RFI only corrected for production. Generally, RFIbf and RFI had very similar QTL profiles, and the most significant QTL for both RFI and RFIbf was on BTA 3 (Figure 1A). This agrees with the high phenotypic correlation found between these traits (r = 0.97), which is consistent with the previously reported phenotypic correlation of r = 0.98 (Schenkel et al., 2004). However, some differences were detected between RFI and RFIbf. The most common difference was in the level of QTL significance. On BTA 6 and 19, QTL of RFI were more significant than that of RFIbf, whereas on BTA 11, 14, 17, and 21, QTL of RFIbf were more significant than that of RFI. Although the levels of significance were different on these chromosomes, locations of the QTL were generally consistent between the 2 versions of RFI. The greatest difference between the QTL profiles of RFI and RFIbf was on BTA 11, where the QTL of RFIbf (P = 0.0009) was significant at the 5% genome-wise level, but the QTL of RFI was only significant at the chromosome-wise level (P = 0.0086). As well, the location of the QTL peaks differed by 27 cM with the RFI QTL at 31 cM and RFIbf at 58 cM (Table 1), although this may be misleading because both shared similar significance threshold intervals of 13 to 61 cM and 22 to 63 cM for RFI and RFIbf, respectively (Figure 1C). The within-family analysis (Table 5) reveals that the shift in peak between RFI and RFIbf is due mainly to one family, sire 2, which was very significant for the QTL for both traits, but the position switched from 59 cM for RFIbf to 31 cM for RFI, reflecting the change in the across-family analysis QTL position. Although some differences in QTL profiles were detected between RFI and RFIbf, the fact that the results were generally similar shows that the correction of RFI with BF does not greatly influence the QTL results.
Comparison with Previous Studies
A few studies have been conducted to examine RFI QTL in cattle (Nkrumah et al., 2005, 2007b; Moore et al., 2006; Barendse et al., 2007). The aim of this study was to refine the QTL locations throughout the genome by using a greater density of markers than the previous RFI QTL genome scan (Nkrumah et al., 2007b). This study used 1,841 more SNP than the previous QTL genome scan, which is an improvement of 85 more markers per chromosome, and decreased the average spacing by 5.36 cM to 1.01 cM. An updated composite map was also used for the present study (Snelling et al., 2007).
In total, 15 QTL were repeated from the primary genome scan (Nkrumah et al., 2007b). There were 9 chromosomes in common that showed RFI QTL in both studies: BTA 1, 7, 12, 17, 18, 19, 21, 24, and 26. Five chromosomes shared FCR QTL in both studies: BTA 7, 11, 17, 24, and 28; only BTA 18 had a DMI QTL in both studies. For the 15 QTL that were confirmed, the regions were narrowed down, on average, by 21.9 cM, from an average of 34.5 cM in the primary scan (Nkrumah et al., 2007b) to 12.6 cM in the present study. Of the 15 QTL, only the RFI QTL on BTA 7, which had the largest interval, was not narrowed from the previous study.
Seventeen QTL were not repeated from the primary genome scan (Nkrumah et al., 2007b), but 19 new QTL were detected in the current QTL scan. For RFI, 4 QTL on BTA 8, 14, 16, and 28 that were identified in the previous scan were not repeated in this study, but 9 new QTL for RFI were detected in this study, which includes BTA 3, 4, 6, 9, 11, 13, 22, 23, and 25. A similar pattern was also seen for FCR, where 3 QTL (BTA 3, 16, and 22) were not detected from the previous scan, but 7 new QTL (BTA 1, 4, 6, 13, 15, 18, and 26) were identified for FCR. This could be attributable to the increased density of markers used in the current study, allowing greater power to detect more QTL. However, for DMI, the current study only detected 3 QTL, but the genome scan detected 11 QTL. This could be because of the greater stringency used to declare a QTL in the present study, as 6 of 11 of the DMI QTL in the previous study were only suggestive with P < 0.10. Indeed, the strongest DMI QTL in the previous study and only one to reach the P < 0.001 level was on BTA 18, which is the one chromosome to show a DMI QTL in both studies. It is noteworthy that the BTA 7 QTL for DMI, which was very significant in our study, was not present in the previous primary genome scan. In addition, the current study used an updated bovine composite map (Snelling et al., 2007) and a revised marker order, which may also explain the inconsistencies between the 2 studies. This study has statistically strengthened the previously identified QTL.
The whole-genome association analysis (Barendse et al., 2007) can also be compared with the present study for RFI, but not for DMI or FCR, as the earlier study did not test those traits. Because not all of the SNP used in that study were mapped and only the physical locations of the SNP were available (Snelling et al., 2007), the approximate physical locations of the QTL in the present study were estimated. This was done using a SNP that mapped at the QTL location and then identifying the physical location of that SNP in the genome sequence (Btau 3.1). The closest SNP to a QTL from the whole-genome association analysis was rs29020547:A > G on BTA 25, which was approximately 200 kbp from the QTL identified in this study. One other SNP, rs29022381:C > T was within 1 Mbp from the QTL on BTA 18 detected in this study. Six SNP, rs41255288:C > T (BTA 1), rs29013885:A > G (BTA 7), rs29024192:C > G (BTA 11), rs29020027:C > G (BTA 19), rs29020542:T > C (BTA 22), and rs29021596:C > T (BTA 23), were within 5 Mbp of RFI QTL reported in the current study. Of these 8 chromosomes, BTA 1, 7, 18, and 19 also contained RFI QTL from the primary genome scan (Nkrumah et al., 2007b), and, therefore, have the most consistent results across the 3 studies. It is of note that the most significant QTL for RFI in our study on BTA 3 was not present in the preliminary genome scan (Nkrumah et al., 2007b), but 1 SNP (rs29013888:C > T), which is within 6 Mbp of the QTL detected in this study, was significantly associated with RFI in the whole-genome association analysis (Barendse et al., 2007).
We have reported the fine mapping and identification of QTL for RFI, DMI, and FCR in beef cattle using 2,194 markers. A contrast was made between RFI and RFI corrected for BF showing that very little difference exists in our QTL results between these 2 measures of feed efficiency. Comparison to the preliminary genome scan (Nkrumah et al., 2007b) and the whole-genome association study (Barendse et al., 2007) shows that the most consistent QTL for RFI are located on BTA 1, 7, 18, and 19. The identification of these QTL provides further evidence for areas in the bovine genome that should be further analyzed for genes and markers that can be used for marker-assisted selection and management.
|
Footnotes |
|---|
2 Corresponding author: stephen.moore{at}ualberta.ca
Received for publication January 16, 2008. Accepted for publication September 8, 2008.
|
LITERATURE CITED |
|---|
|
TopAbstractINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONLITERATURE CITED |
|---|
This article has been cited by other articles:
|
|
 |
|
  E. L. Sherman, J. D. Nkrumah, and S. S. Moore Whole genome single nucleotide polymorphism associations with feed intake and feed efficiency in beef cattle J Anim Sci, January 1, 2010; 88(1): 16 - 22. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
