Transcript system/files/public/presentations/EHodcroftCROI2012posterx

Population based study indicates viral genetic effect on HIV virulence is small but significant
Emma Hodcroft1, Esther Fearnhill2, Andrew Phillips3, David Dunn2, Deenan Pillay4, Jarrod Hadfield5 and Andrew J. Leigh Brown1
on behalf of the UK HIV Drug Resistance Database
Previous studies have estimated the
genetic effect or heritability (h2) of
virulence in HIV at between 5-50%1-5
By analysing the viral phylogeny as a
pedigree, h2 can be estimated using wellestablished quantitative genetic methods
Of the 3 subtypes analysed (A, B, C), the
B subtype produced a significant h2
estimate of 7.7% (CI 4.3-11.0%, p<0.0025)
Collapsing poorly-supported nodes in the
B-subtype tree did not change the
estimate greatly (7.1%; CI 3.6-10.5%,
p<0.0025)
Sub-sampling the B-subtype tree failed to
produce a significant h2 estimate
The heritability of virulence in HIV is small
but significant, needs a large sample size
to be detected, and may be coming from
the deeper tree structure
data:
55,556 sequences were available from the
UK HIV Drug Resistance Database
• 11,096 of these had clinical data including viral load
from UK CHIC
• 539 subtype A, 1,821 subtype C, and 8,483 subtype B
sequences and viral loads were analysed
• 39 reference sequences from LANL were used as an
outgroup
Question: How much of the variation in viral load is due to the viral genetic effect?
introduction:
Pedigree
Fig 1 - Pedigree
Phylogeny
Set-point viral load is an important predictor of virulence in
HIV and varies greatly between individuals6
Previous studies have estimated the genetic effect
(heritability) of virulence in HIV at between 5-50%1-5
However, these studies have had restricted inclusion and
small sample sizes, and many may have been prone to the
confounding effects of transmission pair studies
Quantitative genetic techniques have been widely used for
years to estimate heritability by connecting trait values to
the degree of relatedness in pedigrees (Fig 1)
By analysing phylogenies constructed from HIV sequences
as pedigrees7 (Fig 2) and using these techniques, a large
number of samples can be used and potential biases
avoided
Fig 2 – Phylogeny
methods:
• Initial viral load was taken as an estimate of setpoint, after excluding potential acute-stage or postART viral load measures
• Phylogenies for each of the three subtypes were
reconstructed using FastTree8
• Viral load data and phylogenies were linked to form
a ‘pedigree’ which was analysed in ASReml9
• Each analysis was compared to a model with no
phylogeny to test for significance
Emma Hodcroft
D-274
[email protected]
emmahodcroft.com
m.emmahodcroft.com
1University of Edinburgh, UK; 2Medical Research Council Clinical Trials Unit, London, UK; 3Royal Free Hospital, London, UK; 4University College London, UK; 5University of Oxford, UK.
summary:
contact
HIV DRD Steering Committee: tiny.cc/w2xwk
Leigh Brown group: www.hivbio.org
results:
Collapsed Mean Viral Genetic Effect
(Conf Interval)
11.9% (-6.8 - 30.6%)
No
9.6% (-2.9 - 21.0%)
No
7.7% (4.3 - 11.0%)
No
Significance of Trees
(p<0.0025)
A
C
B
Dataset
Size
539
1,821
8,483
B
B (sub-sample)
8,483
2,120
Yes
No
7.1% (3.6 - 10.5%)
3.5% (-2.4 - 9.4%)
**
B (sub-sample)
4,241
No
5.5% (1.3 - 9.8%)
5/10 were significant†
Subtype
†If
**
only the significant trees are considered: h2 = 7.6% (2.8 - 12.3%)
• Only the B-subtype gave a significant h2 estimate, of 7.7%
• Collapsing poorly-supported nodes eliminated tip structure but did not affect the estimate
• Sub-sampling failed to consistently yield a significant estimate
conclusions:

• Fig 3 - In the B-subtype tree, poorly-supported branches
(bootstrap-equivalents < 0.9) were then collapsed to
polytomies and re-analysed

• Fig 4 - To investigate the effect of sample size,
approximately one-quarter and one-half (N = 2120 & 4241,
respectively) of the B-subtype data set was randomly
sampled and re-analysed
• The heritability of virulence in HIV, at 7.7%, is small, but significant,
matching the lower end of previous estimates
• The deeper tree structure seems to be the source of the heritability,
rather than the tips
• A large sample size is needed to detect the heritability of virulence
• The sub-samples that were significant were perhaps those that
retained the deep tree structure
1. Tang J, et al. AIDS Res Hum Retroviruses 2004;20(1):19-25,
references
2. Hollingsworth TD, et al. PLoS Pathog 2010;6:e1000876.
6. Mellors JW, et al. Science 1996;272(5265):1167-70.
3. Alizon S, et al. PLoS Pathog 2010;6(9).
7. Hadfield JD, Nakagawa S. J Evo Biol 2010;23(3):494-508.
4. Hecht FM, et al. AIDS 2010;24(7):941-5.
5. van der Kuyl AC, et al. AIDS 2010;24(10):1607-8. 8. Gilmour AR, et al. www.vsni.co.uk 2009.
9. Price MN, et al. PLoS ONE 2010;5:e9490.