Accuracy and differential bias in copy number measurement of CCL3L1 in association studies with three auto-immune disorders
© Carpenter et al; licensee BioMed Central Ltd. 2011
Received: 23 March 2011
Accepted: 18 August 2011
Published: 18 August 2011
Copy number variation (CNV) contributes to the variation observed between individuals and can influence human disease progression, but the accurate measurement of individual copy numbers is technically challenging. In the work presented here we describe a modification to a previously described paralogue ratio test (PRT) method for genotyping the CCL3L1/CCL4L1 copy variable region, which we use to ascertain CCL3L1/CCL4L1 copy number in 1581 European samples. As the products of CCL3L1 and CCL4L1 potentially play a role in autoimmunity we performed case control association studies with Crohn's disease, rheumatoid arthritis and psoriasis clinical cohorts.
We evaluate the PRT methodology used, paying particular attention to accuracy and precision, and highlight the problems of differential bias in copy number measurements. Our PRT methods for measuring copy number were of sufficient precision to detect very slight but systematic differential bias between results from case and control DNA samples in one study. We find no evidence for an association between CCL3L1 copy number and Crohn's disease, rheumatoid arthritis or psoriasis.
Differential bias of this small magnitude, but applied systematically across large numbers of samples, would create a serious risk of false positive associations in copy number, if measured using methods of lower precision, or methods relying on single uncorroborated measurements. In this study the small differential bias detected by PRT in one sample set was resolved by a simple pre-treatment by restriction enzyme digestion.
Within the last few years there have been significant developments in our understanding of copy number variation (CNV) and its contribution to human genetic variation. It is now apparent that CNVs are both a frequent form of variation throughout the human genome, and are implicated in disease susceptibility [1, 2].
In general, three forms of CNVs can be defined; insertion, deletion and multi-allelic, but accurate and standardised measurement of individual copy numbers is technically challenging. In particular the measurement of multi-allelic CNVs is the most problematic, more specifically those with higher copy numbers, where it is essential that the technical challenges of copy number genotyping are robust for all copy number classes and do not result in the systematic rejection of samples with high copy number. Repeated exclusion of particular genotypes will lead to mis-representation of copy number variation at a particular locus in a given population due to an artificially reduced apparent mean copy number. To define the full variation at a particular copy number variable locus it is critical that these technical challenges are overcome successfully.
Furthermore, and perhaps most importantly, accurate CNV measurement is essential in the context of case control association studies. Critically, inaccurate copy number measurements can lead to differential bias between cases and controls and result in false positive findings . In SNP based studies differential bias can lead to differences in allele frequency estimates between batches of samples. In the context of CNVs bias can be seen as a systematic shift in the raw measurement around integer values for cases and controls, leading to differential mis-classification of samples by either over- or under-estimation of the integer copy number. Differences in DNA source and quality can cause small changes in the measured copy number distribution of case and control samples . Furthermore, the power to detect disease associations in case control studies is also reduced with inaccurate genotyping of copy number . Thus accurate determination of copy number is limited by the precision of the methodology used, which needs to be robust and replicable to give maximum power and reduce spurious disease associations.
New whole-genome genotyping platforms now incorporate probes to interrogate multiple CNVs , and have begun to yield associations between such variants and disease phenotypes . However, even these methods are sensitive to DNA quality  and are limited in their accuracy to genotype multiallelic loci. The recent report by the Wellcome Trust Case Control Consortium  discusses in detail the potential artefactual influences on copy number determination and demonstrates the potential for reporting spurious associations with CNVs, especially multiallelic regions. To complement whole genome approaches, particularly for complex diseases where effect sizes are likely to be small, locus specific methods are also required for investigating multiallelic CNV loci. Such targeted methods can provide sufficiently accurate data to limit the potential for false positive or negative results as a result of differential bias. Locus specific methods are generally PCR-based, and perhaps the most widely used is Real-Time quantitative PCR, which depends upon the quantification of a copy variable test locus in comparison with an unrelated reference locus . However, differential amplification efficiencies between test and reference loci can generate inaccuracies, particularly for high copy number measurements where the relative difference in ratio between test and reference products is smaller for neighbouring copy number classes. An alternative locus specific method with improved reliability for CNV determination is the paralogue ratio test (PRT) . This method uses a single pairs of primers to amplify specifically two products simultaneously, one from a copy-constant reference locus and the other from the copy variable test locus of interest. The copy number of the test locus is then estimated from the ratio of test to reference products. Recently, the accuracy of the PRT method has been directly compared with a Real-Time quantitative PCR method of copy number measurement for β-defensins and shown to be the more accurate and robust approach . The PRT method has previously been used to successfully analyse both β-defensins and CCL3L1/CCL4L1 copy numbers [10, 11].
The copy variable genes CCL3L1 and CCL4L1 located on chromosome 17q12 lie in a 90 kb repeat unit, neighbouring the paralogous, but copy invariant, genes CCL3 and CCL4. CCL3L1 and CCL4L1 show 96% sequence similarity at both the nucleotide and protein level with their respective paralogues [13, 14]. All four genes (CCL3, CCL3L1, CCL4 and CCL4L1) encode chemokines, chemotactic cytokines, which play an important role in the immune response by attracting lymphocytes and macrophages to sites of infection and inflammation. Furthermore they are all natural ligands for CCR5, the co-receptor used by HIV-1 for cell entry [15, 16], with CCL3L1 being the most potent . As such there have been a number of association studies focused on CCL3L1 copy number variation and HIV-1 susceptibility [7, 18, 19]. However, the reported associations are under dispute and the accuracy of CCL3L1 measurement is a major factor in the debate [20–23].
Here we report a modification of a previously described PRT method to measure CCL3L1/CCL4L1 copy number making it more efficient, cost effective and convenient. As CCL3L1 and CCL4L1 function as an attractant of inflammatory mediators we have performed three case control studies with autoimmune phenotypes (Crohn's disease, rheumatoid arthritis and psoriasis). We subsequently discuss in detail the accuracy of the PRT methodology used, the precision in CCL3L1/CCL4L1 copy number measurement and highlight the implications of differential bias with copy number variation.
CCL3L1 copy number measurement
Breakdown of total numbers and concordance of PRT systems for the European samples by disease cohort
Random UK controls
The different PRT systems (CCL3A/C, CCL4A and LTR61A) show a high level of agreement in all the European samples. This supports the previous observation by our lab  and suggests that discordancy between copy numbers of CCL3L1 and CCL4L1 as previously reported is rare . Furthermore the agreement between all PRT systems indicates that variation in the copy number of the reference genes CCL3 and CCL4 is also uncommon in Europeans.
Accuracy and error of PRT measurements
Germline copy numbers are integers, and thus we assume that precision and accuracy of measurement will be shown by a discrete distribution reflecting the underlying integer values . The distributions of the raw data shown in Figure 1 clearly demonstrate PRT values clustering around the inferred integers. The overall standard deviation (normalised for copy number) of the full dataset was 0.075; and for each cohort separately the standard deviations were 0.069 for the European controls, 0.066 for Crohn's disease samples, 0.077 for rheumatoid arthritis samples and 0.081 for Dutch psoriasis and control samples. These values are both small and consistent between datasets, indicating that the variance inherent in the CCL3L1/CCL4L1 PRT measurement system is low, and the precision of the PRT measurements in estimating integer copy number is high and reproducible between datasets.
Mean, deviations and predicted probability of error for the unrounded copy number values for 1581 European samples
Normalised Standard deviation
Probability of error
7.2 × 10-07
4.9 × 10-04
4.9 × 10-03
1.9 × 10-02
Basis of differential bias
Interestingly, two different DNA extraction methods were used for the psoriasis cases and controls. All the cases were extracted using a Qiagen kit, whereas the controls were extracted using a salting-out procedure. It has previously been suggested that different methods of DNA extraction between cases and controls can lead to differential bias due to differences in DNA quality and purity . We envisaged that one potential physicochemical difference between genomic DNA preparations could be the level of residual protein, and in particular locally bound proteins that might be unusually tenacious in particular regions. We therefore tested whether separating our PRT amplicon from neighbouring DNA by restriction digestion might remedy the differential bias. Digestion with the restriction enzyme Bcc I leaves all PRT amplification sites used here intact, but separates them from neighbouring sequences that may potentially influence PCR efficiencies differentially. The discrepancy between cases and controls was rectified by pre-digestion of DNA samples with Bcc I: after digestion with Bcc I the cumulative frequency plot showed no significant difference between cases and control samples (data not shown), (two sets of control experiments were also performed; one set of samples was incubated at 37°C but lacked BccI and the other set of samples was incubated at 37°C with enzyme buffer present but not BccI. Differential bias was still observed with all these control samples). The differential bias did not alter the integer copy number calling of the samples. Although it may not be the cause of all or even most cases of differential bias, the observation that restriction digestion appears to mitigate differences between the case and control DNA collections does suggest that residual local protein binding might underlie the bias initially observed in this study.
Distribution of integer copy numbers in Crohn's disease cases and controls
Distribution of integer copy numbers in rheumatoid arthritis cases and controls
Distribution of integer copy numbers in psoriasis cases and controls
Over the last few years there has been a surge in interest in copy number variation in relation to human variation and disease and accurate copy number measurements are thus essential. Here we describe a more efficient modification of an already published method of locus specific copy number measurement for CCL3L1/CCL4L1 to ascertain copy number for a total of 1581 independent European samples.
We present here three association studies of CCL3L1/CCL4L1 measured with PRT, and find no association for variation in copy number with the autoimmune phenotypes Crohn's disease, rheumatoid arthritis or psoriasis. It is interesting to note however, that, for the Crohn's disease cohort, whilst the sample size is small, there are 7 cases with an integer copy number of 5 and greater, and yet there were no controls observed to have copy numbers greater than 4 (table 3). It has previously been reported that higher copy numbers of CCL3L1/CCL4L1 are associated with increased protein levels and enhanced chemotactic activity  which could suggest that higher copy numbers lead to increased recruitment of immune mediators and potential pathology. The role of both CCL3L1 and CCL4L1 predicts an influence of copy number on autoimmune disorders, but this was not supported by our data.
There is a previous report of an association between rheumatoid arthritis and CCL3L1/CCL4L1 copy number in a New Zealand cohort . It is possible that the association study presented here for rheumatoid arthritis lacks power, but this association was also not supported by the recent report from the Wellcome Trust Case-Control Consortium (WTCCC) on CNV associations in eight common diseases .
Detailed analysis of the raw data generated from the PRT methodology shows the copy number genotyping to have a high degree of accuracy in integer copy number prediction, and that the level of accuracy is reproducible between European datasets. Inaccurate genotyping can lead to differential bias, a proportional shift in raw measurements that differ between sample batches, which we were able to investigate here due to the accuracy of the copy number measurement assay. Whilst evidence for differential bias was detected here with one of the case control cohort, this was resolved. This observation, however, does highlight the importance of examining raw data thoroughly prior to association analysis. Fortunately, the analysis presented here was with lower copy numbers and so the small differential bias observed did not alter the integer copy number calling. However it is clear that at higher copy numbers, or with a less precise measurement assay, differential bias could have easily caused frequent misclassification of integer copy number, potentially leading to false positive associations .
In conclusion, the data presented here failed to find any association between disease and CCL3L1 copy number on three auto-immune disorders (Crohn's disease, rheumatoid arthritis and psoriasis) but did identify an example of differential bias. This was resolved by simple pre-treatment by restriction enzyme digestion; emphasising the importance of preparing case and control DNA together and encouraging the presentation of raw data to allow assessment of accuracy or bias in published association studies.
Control DNA samples
192 random control samples from the UK were used for direct comparison between the "CCL3A" and "CCL3C" PRT systems. These samples were from the European Collection of Cell Cultures (ECACC) Human Random Control (HRC) panels 1 and 2 (http://www.hpacultures.org.uk).
These ECACC samples were also used for the Crohn's disease and rheumatoid arthritis association studies, along with the parents (n = 60) from the International HapMap phase I CEPH samples (http://ccr.coriell.org), making a total of 252 random European control samples. All DNA provided was extracted from lymphoblastoid cell lines.
Samples from patients with Crohn's disease, rheumatoid arthritis and psoriasis
All samples were collected after informed consent and with appropriate ethical approval; details can be found in [9, 10, 26]. There were 657 Crohn's disease patient samples of UK origin, collected from patients attending clinics in London or Newcastle . DNA was obtained from 274 rheumatoid arthritis patient samples of UK origin, although due to limited DNA volumes only 252 samples were successfully genotyped. The psoriasis samples and matched controls were from a cohort of samples from Nijmegen, Holland and have already been described . In brief, there were 276 control samples from the Nijmegen Biomedical Study (NBS), and 202 unrelated psoriatic cases. All patients were diagnosed with psoriasis vulgaris, and classified as having moderate to severe psoriasis.
In order to address whether there is any population stratification of CCL3L1 copy number across Europe a t-test was performed between the Dutch control samples (n = 202) and the UK ECACC control samples (n = 192). There was no significant difference between the means of the two control datasets suggesting that that there is no population stratification across Europe and that the control samples are appropriate for the association studies.
Copy number measurement
The paralogue ratio test (PRT)
CCL3L1 copy number of the psoriasis samples and the random control samples was measured using a minor modification of the PRT method previously described . In the modified method the system "CCL3A" was changed to a similar system, termed "CCL3C", where the primer pairs for amplification of both CCL3/CCL3L1 were re-designed. The primers were specifically designed using DNA sequence flanking exon 1 of CCL3 and CCL3L1, and therefore will not amplify from the CCL3L1 pseudogene, which lacks exon 1. For the "CCL3C" PRT 1 μM each of primers FAM-labelled CCL3CF (GGC TAA GAC CCC TTC TAG AG) with CCL3CR (AAT CAT GCA GGT CTC CAC T) were used which gives products of 252 bp for CCL3 and 258 bp for CCL3L1. The "CCL3C" system PCR is more efficient than the "CCL3A" system and could be multiplexed with both the "CCL4A" and "LTR61A" systems in a single PCR reaction with a reaction mixture of 0.5 U Taq DNA polymerase (NEB) in a buffer with final concentrations of 50 mM Tris-HCl pH8.8, 12.5 mM ammonium sulphate, 1.4 mM magnesium chloride, 7.5 mM 2-mercaptoethanol, 125 μg/ml BSA and 200 μM each dNTP. PCR cycle conditions were 24 cycles of 95°C for 30s, 55°C for 30s, 70°C for 60s followed by a final hold at 70°C for 40 minutes. CCL3L1 copy number of Crohn's disease and rheumatoid arthritis samples were measured from genomic DNA using the paralogue ratio test (PRT) method previously described .
In order to reduce variation due to batch effects or plate position artefacts, case and control samples were interspersed within 96-well plates and genotyping was performed blind to the clinical status of each sample. Furthermore, all plates included some duplicate samples that were measured independently on separate plates to ensure consistency. For all PRTs fragment analysis of the test and reference loci was carried out by electrophoresis on an ABI3100 36 cm capillary using POP-4 polymer with an injection time of 30 s at 1 kV. Products from the PRT reactions were mixed with 10 μl HiDi formamide with ROX-500 marker (Applied Biosystems).
Two microsatellite PCRs were also performed, as described previously , for any sample showing inconsistencies between the PRT systems.
Copy number measurement of the PRT products
The peak areas were extracted using GeneMapper software (Applied Biosystems) and the ratio of test locus to reference locus was calculated for each sample. In order to calibrate and standardise unknown DNA samples, DNA samples of established copy number were included in each PCR in triplicate (ECACC samples, C0075 with a copy number (CN) = 1; C0150 with CN = 2; C0007 with CN = 3; and C0877 with CN = 4). A linear regression was applied to the cluster of ratios generated from the standard reference DNA samples and used to calibrate the inference of copy number from the ratios of the unknown samples. Copy number values from each of the independent PRT systems were compared and an arithmetic mean value was calculated from all PRT measurements to generate a single unrounded copy number value for each sample.
In order to test the difference in means between the copy number distributions of cases and control samples a two-tailed t-test was performed. Analyses were performed between European control samples and Crohn's disease or rheumatoid arthritis samples (both UK), and between Nijmegen psoriasis cases and Dutch controls.
Accuracy and error measurements
The accuracy of the PRT measurements to ascertain correctly the copy number for each sample was assessed in each dataset from the unrounded copy number data. All analyses were performed in Excel (Microsoft 2007) and SPSS (Version 16). For each dataset the mean and standard deviation of the unrounded data were measured for each copy number group. To obtain a general standard deviation of the whole dataset, and for each copy number class, the data were normalised by dividing the unrounded copy number value by the predicted integer copy number.
To estimate the probability of error inherent in a PRT system it is assumed that measurements of copy number values are normally distributed around each integer. From the observed data the mean and standard deviation for each copy number can be estimated and used to calculate the probability of mis-assigning a sample to the appropriate integer class.
To support the probability of misclassification approximation, a Kolmogorov-Smirnov test for normality was performed with the random control samples for copy number classes 1, 2 and 3. All 3 groups showed no significant deviation from normality (see additional file 3 figure S3 for Q-Q plots).
Differential bias was first examined for each PRT system (CCL3C, CCL4A and LTR61A) using the random control samples, to ensure that the different PRT systems are not differentially sensitive to DNA quality nor are there differences in PCR efficiencies that could lead to bias. Together this ensures that all the PRT data can be collated into a single value. Differential bias was also investigated between cases and control samples in the clinical cohorts, by comparing a normalised cumulative frequency plot of the unrounded copy numbers around an integer copy number of 2, as this is the most frequent copy number group, and therefore with greatest power to show any discrepancies. Where a difference was observed, a two-tailed t-test was performed to examine the significance of the deviation and other integer classes were investigated.
Copy number variation
single nucleotide polymorphism
paralogue ratio test.
The authors would like to thank Dr. Jane Worthington for contributing rheumatoid arthritis samples. DC is supported by a Wellcome Trust grant (number 083929) awarded to JALA, and SW was supported by a Medical Research Council studentship.
- Redon R, Ishikawa S, Fitch KR, Feuk L, Perry GH, Andrews TD, Fiegler H, Shapero MH, Carson AR, Chen W, et al: Global variation in copy number in the human genome. Nature. 2006, 444 (7118): 444-454. 10.1038/nature05329.PubMed CentralPubMedView Article
- Wellcome Trust Case Control Consortium: Genome-wide association study of CNVs in 16,000 cases of eight common diseases and 3,000 shared controls. Nature. 2010, 464 (7289): 713-720. 10.1038/nature08979.View Article
- Clayton DG, Walker NM, Smyth DJ, Pask R, Cooper JD, Maier LM, Smink LJ, Lam AC, Ovington NR, Stevens HE, et al: Population structure, differential bias and genomic control in a large-scale, case-control association study. Nat Genet. 2005, 37 (11): 1243-1246. 10.1038/ng1653.PubMedView Article
- McCarroll SA, Altshuler DM: Copy-number variation and association studies of human disease. Nat Genet. 2007, 39: S37-S42. 10.1038/ng2080.PubMedView Article
- Stranger BE, Forrest MS, Dunning M, Ingle CE, Beazley C, Thorne N, Redon R, Bird CP, de Grassi A, Lee C, et al: Relative Impact of Nucleotide and Copy Number Variation on Gene Expression Phenotypes. Science. 2007, 315 (5813): 848-853. 10.1126/science.1136678.PubMed CentralPubMedView Article
- Ionita-Laza I, Rogers AJ, Lange C, Raby BA, Lee C: Genetic association analysis of copy-number variation (CNV) in human disease pathogenesis. Genomics. 2009, 93 (1): 22-26. 10.1016/j.ygeno.2008.08.012.PubMed CentralPubMedView Article
- Gonzalez E, Kulkarni H, Bolivar H, Mangano A, Sanchez R, Catano G, Nibbs RJ, Freedman BI, Quinones MP, Bamshad MJ, et al: The Influence of CCL3L1 Gene-Containing Segmental Duplications on HIV-1/AIDS Susceptibility. Science. 2005, 307 (5714): 1434-1440. 10.1126/science.1101160.PubMedView Article
- Armour JAL, Palla R, Zeeuwen PLJM, den Heijer M, Schalkwijk J, Hollox EJ: Accurate, high-throughput typing of copy number variation using paralogue ratios from dispersed repeats. Nucl Acids Res. 2007, 35 (3): e19-10.1093/nar/gkl1089.PubMed CentralPubMedView Article
- Aldhous MC, Abu Bakar S, Prescott NJ, Palla R, Soo K, Mansfield JC, Mathew CG, Satsangi J, Armour JAL: Measurement methods and accuracy in copy number variation: failure to replicate associations of beta-defensin copy number with Crohn's disease. Human Molecular Genetics. 2010, 19 (24): 4930-4938. 10.1093/hmg/ddq411.PubMed CentralPubMedView Article
- Hollox EJ, Huffmeier U, Zeeuwen PLJM, Palla R, Lascorz J, Rodijk-Olthuis D, van de Kerkhof PCM, Traupe H, de Jongh G, den Heijer M, et al: Psoriasis is associated with increased [beta]-defensin genomic copy number. Nat Genet. 2008, 40 (1): 23-25. 10.1038/ng.2007.48.PubMed CentralPubMedView Article
- Walker S, Janyakhantikul S, Armour JAL: Multiplex Paralogue Ratio Tests for accurate measurement of multiallelic CNVs. Genomics. 2009, 93 (1): 98-103. 10.1016/j.ygeno.2008.09.004.PubMedView Article
- Irving SG, Zipfel PF, Balke J, McBride OW, Morton CC, Burd PR, Siebenlist U, Kelly K: Two inflammatory mediator cytokine genes are closely linked and variably amplified on chromosome 17q. Nucl Acids Res. 1990, 18 (11): 3261-3270. 10.1093/nar/18.11.3261.PubMed CentralPubMedView Article
- Nakao M, Nomiyama H, Shimada K: Structures of human genes coding for cytokine LD78 and their expression. Mol Cell Biol. 1990, 10 (7): 3646-3658.PubMed CentralPubMedView Article
- Modi WS: CCL3L1 and CCL4L1 chemokine genes are located in a segmental duplication at chromosome 17q12. Genomics. 2004, 83 (4): 735-738. 10.1016/j.ygeno.2003.09.019.PubMedView Article
- Choe H, Farzan M, Sun Y, Sullivan N, Rollins B, Ponath PD, Wu L, Mackay CR, LaRosa G, Newman W, et al: The [beta]-Chemokine Receptors CCR3 and CCR5 Facilitate Infection by Primary HIV-1 Isolates. Cell. 1996, 85 (7): 1135-1148. 10.1016/S0092-8674(00)81313-6.PubMedView Article
- Alkhatib G, Combadiere C, Broder CC, Feng Y, Kennedy PE, Murphy PM, Berger EA: CC CKR5: A RANTES, MIP-1alpha, MIP-1beta Receptor as a Fusion Cofactor for Macrophage-Tropic HIV-1. Science. 1996, 272 (5270): 1955-1958. 10.1126/science.272.5270.1955.PubMedView Article
- Menten P, Struyf S, Schutyser E, Wuyts A, De Clercq E, Schols D, Proost P, Van Damme J: The LD78b isoform of MIP-1a is the most potent CCR5 agonist and HIV-1-inhibiting chemokine. The Journal of Clinical Investigation. 1999, 104 (4): R1-R5. 10.1172/JCI7318.PubMed CentralPubMedView Article
- Dolan MJ, Kulkarni H, Camargo JF, He W, Smith A, Anaya J-M, Miura T, Hecht FM, Mamtani M, Pereyra F, et al: CCL3L1 and CCR5 influence cell-mediated immunity and affect HIV-AIDS pathogenesis via viral entry-independent mechanisms. Nat Immunol. 2007, 8 (12): 1324-1336. 10.1038/ni1521.PubMedView Article
- Kulkarni H, Marconi VC, Agan BK, McArthur C, Crawford G, Clark RA, Dolan MJ, Ahuja SK: Role of CCL3L1-CCR5 Genotypes in the Epidemic Spread of HIV-1 and Evaluation of Vaccine Efficacy. PLoS ONE. 2008, 3 (11): e3671-10.1371/journal.pone.0003671.PubMed CentralPubMedView Article
- Bhattacharya T, Stanton J, Kim E-Y, Kunstman KJ, Phair JP, Jacobson LP, Wolinsky SM: CCL3L1 and HIV/AIDS susceptibility. Nat Med. 2009, 15 (10): 1112-1115. 10.1038/nm1009-1112.PubMedView Article
- Field SF, Howson JMM, Maier LM, Walker S, Walker NM, Smyth DJ, Armour JAL, Clayton DG, Todd JA: Experimental aspects of copy number variant assays at CCL3L1. Nat Med. 2009, 15 (10): 1115-1117. 10.1038/nm1009-1115.PubMed CentralPubMedView Article
- Urban TJ, Weintrob AC, Fellay J, Colombo S, Shianna KV, Gumbs C, Rotger M, Pelak K, Dang KK, Detels R, et al: CCL3L1 and HIV/AIDS susceptibility. Nat Med. 2009, 15 (10): 1110-1112. 10.1038/nm1009-1110.PubMed CentralPubMedView Article
- He W, Kulkarni H, Castiblanco J, Shimizu C, Aluyen U, Maldonado R, Carrillo A, Griffin M, Lipsitt A, Beachy L, et al: Reply to: "CCL3L1 and HIV/AIDS susceptibility" and "Experimental aspects of copy number variant assays at CCL3L1". Nat Med. 2009, 15 (10): 1117-1120. 10.1038/nm1009-1117.PubMedView Article
- Townson JR, Barcellos LF, Nibbs RJ: Gene copy number regulates the production of the human chemokine CCL3-L1. European Journal of Immunology. 2002, 32 (10): 3016-3026. 10.1002/1521-4141(2002010)32:10<3016::AID-IMMU3016>3.0.CO;2-D.PubMedView Article
- McKinney C, Merriman ME, Chapman PT, Gow PJ, Harrison AA, Highton J, Jones PBB, McLean L, O'Donnell JL, Pokorny V, et al: Evidence for an influence of chemokine ligand 3-like 1 (CCL3L1) gene copy number on susceptibility to rheumatoid arthritis. Annals of the Rheumatic Diseases. 2008, 67 (3): 409-413.PubMedView Article
- Orozco G, Hinks A, Eyre S, Ke X, Gibbons LJ, Bowes J, Flynn E, Martin P, Wellcome Trust Case Control C, consortium Y, et al: Combined effects of three independent SNPs greatly increase the risk estimate for RA at 6q23. Human Molecular Genetics. 2009, 18 (14): 2693-2699. 10.1093/hmg/ddp193.PubMed CentralPubMedView Article
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.