60 Etiopathogenesis of Cruciate Ligament Rupture
of the extracellular matrix (ECM) of cartilage
(Schwartz & Domowicz 2002). A significant
association was identified with a SNP found in
CSPCP, but notVCAN, and the trait, suggesting
a role for aggrecan in CR pathogenesis.
More recently, an updated GWAS of CR was
performed in a population of Newfoundlands
using SNP genotyping (Bairdet al. 2014b). Ini-
tially, 96 dogs (48 cases, 48 controls) were
genotyped using a SNP array with>170,000
SNP markers evenly spaced across the canine
genome. Clinical diagnoses were confirmed
by a veterinary surgeon. Control dogs were
defined as dogs over the age of 5 years with no
history of stifle injury, stifle instability, or pelvic
limb lameness. After quality control, the sam-
ples were analyzed using the Cochran–Mantel–
Haenszel (CMH) test as well as Efficient Mixed-
Model Association Expedited (EMMAX) analy-
sis. Both the CMH test and EMMAX are used
to correct for relatedness that may exist in the
population, an important step for canine GWAS
studies. The 65 most significant SNPs from this
analysis were re-genotyped in a larger group
of 271 Newfoundlands (96 from the prelim-
inary GWAS and 175 new) using Sequenom
genotyping, and case-control association was
re-analyzed. Three main associations were iden-
tified on chromosomes 1, 3, and 33. The associ-
ation on chromosome 1 included SNPs within
theRNF152gene, the association on chromo-
some 3 included SNPs within theSORCS2gene,
and the association on chromosome 33 included
SNPs withinSEMA5B,DIRC2,andZDHHC23.
SEMA5B,SORCS2,andZDHHC23all have var-
ious roles in the nervous system. Other nervous
system genes were also identified in regions
that did not maintain statistical significance
after correction for multiple testing. This pro-
vides evidence for the potential role of neuro-
logical pathways in CR disease risk. The authors
cited the importance of mechanoreceptors for
appropriate proprioception, as reduced pro-
prioception may prevent appropriate response
to mechanical loading, placing the CrCL at
increased risk of matrix damage and fiber rup-
ture (Bairdet al.2014b). These regions did not
overlap with the regions identified in the earlier
Newfoundland CR GWAS (Wilkeet al. 2009).
Most recently, a GWAS of CR was performed
in another high-risk breed, the Labrador
Retriever (Bakeret al. 2017). Purebred Labrador
Retrievers (237 dogs consisting of 98 cases,
139 controls) were genotyped using an array
with>170,000 SNPs evenly spaced across the
canine genome. Clinical diagnosis of CR was
confirmed by a veterinary surgeon. Controls
were over the age of 8 years (Reif & Probst
2003) with a normal orthopaedic examination.
Additionally, lateral standing stifle radiographs
were taken of all control dogs to confirm that
there were no signs of stifle pathology before
enrollment (Chuanget al. 2014). This level of
phenotyping is more stringent than in previous
studies (Wilkeet al. 2009; Bairdet al. 2014b).
After quality control,∼119,000 SNPs remained
for linear mixed model association using two
algorithms, GEMMA (Zhou & Stephens 2012)
and GCTA (Yanget al. 2011), as well as Penal-
ized Unified Multiple-locus Association using
PUMA (Hoffmanet al. 2013). Genome-wide
significance was defined separately for each
algorithm using a permutation procedure
(Bakeret al. 2017). To facilitate pathway anal-
ysis, an additional set of candidate loci was
identified with a cutoff ofP<5E-04 (Karlsson
et al. 2013), and results from all three algorithms
were combined. This approach identified 129
SNPs associated with CR. These SNPs were
grouped into 99 regions based on linkage
disequilibrium (r^2 >0.8).
One region containing two SNPs on chro-
mosome 24 met genome-wide significance.
Nine genes were identified in this region
with diverse physiological effects on cellu-
lar and tissue homeostasis. These included
bactericidal/permeability-increased protein
(BPI), lipopolysaccharide-binding protein
(LBP), Ral GTPase activation protein beta sub-
unit (RALGAPB), adipogenin (ADIG), solute
carrier family 32, member 1 (SLC32A1), ARP5
actin-related protein 5 (ACTR5), protein phos-
phatase 1, regulatory subunit 16B (PPP1R16B),
family with sequence similarity 83, member
D(FAM83D), and DEAH (Asp-Glu-Ala-His)
box polypeptide 35 (DHX35). An additional 98
regions were identified at the candidate level.
All regions were combined for pathway analy-
sis using DAVID (Huanget al. 2009). Pathway
analysis revealed an association with a cluster
of 11 genes encoding proteins with antimicro-
bial activity (correctedP=0.02) and a cluster of
24 genes encoding carbohydrate-binding pro-
teins (correctedP=1.21E-04). Notably, the clus-
ter of antimicrobial genes includedLBPandBPI,
which were also present in the genome-wide