permitting the possible recognition of a newly inherited ERV,
even before the retrovirus has colonized the host genome.
The aim of this paper is to study how the zinc-finger proteins
that recruit KAP1 to the genome evolve to recognize and
target new inherited retroviruses. We first show evidence that
KAP1 is recruited by zinc-finger proteins to sequences on
human ERVs by analyzing a ChIP-seq data set for wild-type
KAP1 and a mutant KAP1 in which the domain that binds to
the zinc-finger proteins was deleted. Second, we study the
evolution of the DNA-binding specificity of the associated
zinc-finger genes in the evolutionary lineage of humans.
We show how many of the zinc-finger genes sit in clusters
associated with copy number variant (CNV) formation hot-
spots on human chromosome 19. In addition, we use a
computational technique to predict the DNA binding sites of
every human zinc-finger gene, to show how the zinc-finger
genes that target sequences contained in ERV DNA are
preferentially located on human chromosome 19.
Finally, in the discussion, we review other mechanisms of
host recognition of transposable elements (TEs) that have
been proposed in the literature. We argue in favor of a new
model in which zinc-finger genes found in a continuous array
on chromosome 19 undergo recombination generating CNVs
and new genes whose proteins recognize novel DNA
sequences, some of which are found in retrotransposons.
Because those retrotransposons that are unchecked by
recognition of a repressor zinc-finger can go on to kill the
host, we expect to observe in present day organisms a good
correspondence between zinc-finger genes that recognize
and bind to those retrotransposons that have recently entered
the human lineage (and survived). We observe that corres-
pondence in the lineage of primates analyzed here. Only in a
scenario in which the host population is producing a large
reservoir of repressors with different DNA binding affinities do
the offspring that inherit a new ERV have a significant chance
to somatically silence it and reproduce at a reasonable rate.
Results
Human KZNF transcription factors recruit KAP1 to
binding sites located on endogenous retroviral DNA
and other TEs. To quantify the frequency with which human
KAP1 is recruited to endogenous retroviral DNA, we
analyzed a recently published ChIP-seq data set for
KAP1.
8
The authors determined the genomic location of
KAP1 by means of chromatin immunoprecipitation of KAP1
followed by next-generation sequencing experiments per-
formed on three different cell lines (human embryonic kidney
293 cells (HEK293), U2OS and K562 cells, see Materials and
Methods). The authors found a total of 18 760 autosomal
peaks spanning 8 900 411 base pairs. We annotated every
repetitive DNA sequence contained in the peaks using
RepeatMasker. Of the 8.9 Mbp spanning 18 760 autosomal
peaks, 3.7 Mbp were annotated as repetitive DNA. In
particular, 3.6 Mbp were annotated as TEs, which include
long terminal repeat (LTR) elements, DNA transposons, long
interspersed nuclear element retrotransposons (LINEs) and
short interspersed nuclear element retrotransposon (SINEs).
LTR elements alone, which include ERVs, spanned 1.64 Mbp
of the 18 760 autosomal peaks. These 1.64 Mbp that are
annotated as LTR elements span 18% of the binding
sequences of KAP1. As these proportions depend on the
particular chromatin states existing in the cell types, we also
compared the relative abundances of TE-derived DNA in the
regions of accessible chromatin of HEK293 and K562 cells
with the relative abundances of TE-derived DNA in the
binding peak sequences (see Table 1 and Materials and
Methods). We observed that the binding of KAP1-associated
TFs on LTR elements and LINEs is between fourfold and
eightfold more frequent than expected in a null model of
random binding.
The next question we explored was what factors that
interact with KAP1 recognize this parasitic DNA? The
available evidence demonstrates that the recruitment of
KAP1 to endogenous retroviral DNA is mediated by the
interaction of its RBCC domain with the Krueppel-Associated
Box (KRAB) domain present in many TFs.
9
To test this
hypothesis, we compared the binding sites of a mutant KAP1
with no RBCC domain (mt KAP1) versus the binding sites of
wild-type KAP1 (wt KAP1) on HEK293 cells
8
(see Materials
and Methods section). We inferred the binding peaks using
the MACS algorithm.
10
We applied a P-value cutoff of 10
10
and identified a total of 20 139 autosomal peaks for wt KAP1
and 732 autosomal peaks for mt KAP1. To reduce the fraction
of misidentified peaks, we considered only the subset of
peaks that had also been inferred previously. We observed a
very large depletion of binding sites for the mutant KAP1-
DRBCC (mt KAP1). In particular, mt KAP1 was found in only
B4% of the binding sites on TE-coding DNA compared with
experiments where wild-type KAP1 was used (wt KAP1) (see
Figure 2). In the case of LTRs and LINEs, only B3.5% of the
wt binding sites on LTR elements were present in the
experiment with mt KAP1. Hence, this supports the hypo-
thesis that KAP1 is recruited to endogenous retroviral DNA by
RBCC
KAP1
KRAB
Zinc Finger
C2H2
HP1
H3K9me3
Target DNA locus
TFBS
Figure 1 KAP1 has an N-terminal tripartite motif (TRIM) containing an RBCC
domain (ring finger, two B-box zinc-fingers and a coiled coil), a central HP1
(heterochromatin protein 1) domain and a C-terminal combination plant homeo-
domain (PHD) and bromodomain (B). These three domains have been shown to
mediate nuclear localization, interaction with TFs, oligomerization and regulation of
transcription.
21
The RBCC domain interacts with the KRAB module present in the
KRAB-C2H2 zinc-finger proteins (KZNF). In addition to the RBCC domain, every
other subdomain of KAP1 contributes to the remodeling of chromatin on genomic
loci targeted by the KRAB-containing TFs.
9
For instance, the HP1-binding domain
(PxVxL) interacts with HP1 family members, whereas the KAP1–HP1 complex has
a role in silencing euchromatic and pericentric heterochromatic regions.
21
The PHD
and bromodomain interact with two chromatin-modifying enzymes: Mi2a and
SETDB1, of which SETDB1 encodes a histone methyltransferase involved in
histone methylation, gene silencing and transcriptional repression
8
Diversity of zinc-finger genes: a host defense against ERVs
S Lukic et al
382
Cell Death and Differentiation