Mammalian cDNA and prokaryotic reporter sequences silence adjacent transgenes in
transgenic mice
Mammalian cDNA and prokaryotic reporter sequences silence adjacent transgenes in transgenic mice
A. J.
Clark*
,
G.
Harold
and
F. E.
Yull
Roslin Institute,
Roslin
, Midlothian EH25 9PS,
UK
Received November 19, 1996;
Revised and Accepted January 12, 1997
ABSTRACT
The ovine
[beta]
-lactoglobulin gene is expressed efficiently and at high levels in the
mammary gland of transgenic mice. In contrast, when this gene is linked to a
second gene construct comprising a mammalian cDNA or a
CAT
reporter sequence it fails to be expressed in the majority of transgenic lines
generated. We suggest that mammalian cDNAs and prokaryotic reporter sequences
can serve as active foci for gene silencing in the mammalian genome.
INTRODUCTION
The use of cDNA constructs in transgenic mice often results in poor expression,
even when the same constructs are expressed efficiently in cell culture (
1
,
2
). Similar problems are encountered with prokaryotic reporter sequences such as [beta]-galactosidase (
lacZ
) or chloramphenicol acetyltransferase (
CAT
) (
3
) even if these sequences are linked to regulatory elements such as the globin
locus control region (LCR), which can mediate position-independent expression of other genes (
4
). We have constantly encountered these problems in our attempts to target the
expression of foreign genes to the mammary gland of transgenic animals (
5
,
6
). In these studies we have used regulatory elements derived from the ovine [beta]-lactoglobulin (
BLG
) gene. The unmodified gene is expressed very efficiently in the mammary gland
of transgenic mice (
7
) and only 400 bp of the proximal promoter are required to mediate position-independent expression (
8
). In contrast, hybrid gene constructs comprising the
BLG
promoter linked to human [alpha]-1-antitrypsin ([alpha]
1AT
) or factor IX (
fIX
) cDNAs (
5
) or
CAT
reporter sequences (
6
) are expressed very inefficiently and these transgenes are silent in the
majority of the lines generated.
To overcome the problem of poor expression of cDNA constructs we developed a
strategy involving co-integration with the unmodified
BLG
gene and showed that this approach can be used to rescue the expression of
poorly expressed gene constructs. For example, a
BLG
-
fIX
construct, FIXD, was completely silent in transgenic mice but was expressed in
most of the transgenic lines generated when it was co-integrated with
BLG
(
9
). Transgenic mice expressing relatively high levels of human fIX in the milk
have been made using this strategy (
10
). Notwithstanding the success of the approach, the level of expression of the
rescued gene was usually less than the expression of the co-integrated
BLG
gene and the degree of rescue was highly variable when different transgenic
lines were compared. Other workers have also noted this when using this
approach to enhance transgene expression (
11
).
Co-injection generally results in the co-integration of the two transgenes, but their mutual arrangement is
highly variable when different lines are compared (
9
). We reasoned that this variability in the structure of the transgene arrays
contributes to the unpredictability of the rescue effect. To address this
problem we constructed plasmids in which
BLG
-
fIX
or
BLG
-
CAT
sequences were linked to the
BLG
gene prior to injection so that their relative arrangement in the transgene
arrays would be more consistent. The major effect seen was not so much
improvement of expression of the hybrid genes but, surprisingly, complete
silencing of the adjacent
BLG
gene in the majority of the transgenic lines generated.
MATERIALS AND METHODS
Constructs
The genomic
BLG
gene (BLG[Delta]Dp),
BLG
-
CAT
construct ([Delta]DpCAT) and
BLG
-
fIX
construct (FIXD) have been described previously (
5
,
6
,
8
). For CAFI the BLG[Delta]Dp gene was linked at the
Xba
I site in the 3' flanking region to the 5'-end of the 400 bp
BLG
promoter contained in a modified FIXD[Delta]3' construct lacking the 3'
Bam
HI-
Xba
I flanking region. The
fIX
cDNA sequences have a small deletion in the 3'-UTR that removes a cryptic acceptor splice site, preventing
aberrant splicing(
10
). In BRAT, the 5'-end of the 400 bp
BLG
promoter in [Delta]DpCAT was linked to the 3'
Xba
I site and in TARB the two constructs were joined at their respective 3'
Xba
I sites. All three constructs were generated in pBluescript SK2 (Stratagene) and
were excised from the vector sequences by
Sfi
I digestion. All constructs are depicted in Figure
1
.
Generation and characterization of transgenic mouse lines
Injection fragments were purified by agarose gel electrophoresis and injected
into pronuclear eggs from C57/Bl6 * CBA mice as previously described (
7
). Tail biopsies were analysed by PCR and Southern blotting using
BLG
primers and probes and transgenic lines were established by breeding from
positive founder mice. Southern blotting of
Eco
RI and
Hin
dIII digested tail and liver DNA was used to confirm transgenic status and to
ensure the integrity of the transgenic loci using a
BLG
-specific
Bam
HI-
Nru
I probe spanning exon IV,
fIX
cDNA sequences isolated from the plasmid p5'G 3'cVI (a gift from G.Brownlee) and
CAT
reporter sequences (Fig.
1
). Transgene copy numbers were estimated by reference to dilutions of the
constructs run in parallel on the same gel.
Analysis of expression
Expression analysis was normally carried out in generation 1 (G
1
) or generation 2 (G
2
) mice, although in some cases expression was assessed in the founder (G
0
) animals. Positive females were selected, mated and killed at day 12 of
lactation. Mammary tissue and milk was taken. fIX protein was measured in milk
samples by ELISA as previously described (
10
). CAT assays were carried out on protein extracts prepared by homogenizing and
then centrifuging small amounts of frozen mammary tissue. The protein
concentration in each sample was measured using the BioRad protein assay kit.
CAT activity was assayed in a final volume of 200 [mu]l using 10 [mu]g protein extract in 0.25 M Tris, pH 8.8, with [
14
C]chloramphenicol. The reaction was started with 5 [mu]l 50 mM acetyl-CoA, incubated at 37oC for 3 h and stopped with an equal volume of ethyl acetate.
Samples were freeze dried, resuspended in 30 [mu]l ethyl acetate, spotted on to TLC plates and run in chloroform: methanol
95:5. BLG expression was determined by Northern blotting analysis on 1% agarose-MOPS gels. After electrophoresis the gels were blotted onto Hybond N and
probed with BLG and GAPDH or a mouse casein probe to control for RNA loading. Hybridization signals were imaged with a PhosphorImager and the levels of BLG mRNA in the expressing lines estimated relative to
a standard BLG-containing RNA sample included on the blots using ImageQuant software.
RESULTS
Human
fIX
sequences silence an adjacent
BLG
transgene
We have reported previously that the
BLG
-
fIX
transgene FIXD, comprising human
fIX
cDNA sequences linked to the
BLG
promoter, was not expressed in transgenic mice (
5
). The expression of this construct and its derivative FIXD[Delta]3' (Fig.
1
) were, however, rescued by co-injection with the unmodified
BLG
gene (
9
,
10
). Nevertheless, the levels of expression of these
BLG
-
fIX
transgenes were still unpredictable and, for example, the fIX levels in the
milk of different transgenic lines carrying FIXD[Delta]3' varied from <100 ng/ml to ~60 [mu]g/ml and expression levels bore no relationship to copy
number (
10
). Co-injection of two genes generates highly variable transgene arrays when
different lines are compared and we hypothesized that this contributed to these
unpredictable levels of expression. Therefore, we elected to link the
BLG
and
BLG
-
fIX
constructs prior to injection so as to generate a more consistent arrangement
of the two constructs. CAFI comprises the BLG[Delta]Dp transgene linked to a modified FIXD[Delta]3' transgene (Fig.
1
). Eleven transgenic lines were generated which carried this transgene.
Restriction blotting analysis with
BLG
- and
fIX
-specific probes indicated that the transgene was not rearranged
significantly in any of these lines and they all yielded the expected
Eco
RI fragments. The mice were analysed in the G
0
and G
1
generations for fIX and BLG expression in the mammary gland. Only two of the 11
CAFI lines showed detectable fIX expression either in the milk (Fig.
2
) or as mRNA transcripts (not shown). The expression in these lines was low,
with the highest line exhibiting only 0.38 [mu]g/ml fIX in the milk. This level of fIX was nearly two orders of magnitude
lower than was obtained when FIXD[Delta]3' was co-injected with the
BLG
gene and fIX concentrations in the milk >100 [mu]g/ml were achieved (
10
).
CAT
reporter sequences also silence the
BLG
gene
The 0.4 kb
BLG
promoter is sufficient to target expression of a heterologous reporter gene to
the mammary gland in transgenic mice. The construct [Delta]DpCAT (Fig.
1
) containing the 0.4 kb
BLG
promoter linked to
CAT
, however, was expressed with low efficiency and only one of eight lines
expressed this transgene (Fig.
3
;
6
). In an attempt to rescue expression of this construct it was linked to BLG[Delta]Dp in a head-to-tail fashion to generate a double gene construct named BRAT
(Fig.
1
). Eleven lines of mice were generated with this construct, characterized by
Southern blotting after
Hin
dIII digestion and investigated for CAT and BLG expression. Linking the [Delta]DpCAT construct to BLG[Delta]Dp improved its expression marginally, with three out of 11 lines
showing measurable levels of CAT activity (Fig.
3
). Northern blotting experiments showed that nine of the 11 BRAT lines analysed
failed to express the linked
BLG
gene and, as for the CAFI construct described above, the predominant effect was
silencing of the
BLG
gene.
Variable silencing within transgenic lines
For most of the transgenic lines in this study expression was analysed in two G
1
females and the line was scored negative if both mice failed to express either
transgene. For two of the BRAT and two of the TARB lines additional expressing
females within the line were also analysed for expression.
In some of these lines we observed differing levels of CAT and/or BLG expression
on comparison of individuals within the line. In some cases this variability
was extreme. For example, in line BRAT 62 CAT was only seen in one of the four
individuals analysed (Fig.
4
) and, likewise, BLG was only detectable in this individual. In contrast, mice
from line BRAT 22 showed a more stable pattern of CAT expression, with all the
individuals exhibiting high levels of CAT expression and low levels of BLG
mRNA. All the mice analysed in line TARB 1 showed a significant level of CAT
activity and BLG mRNA. In contrast, individuals from line TARB 8 all showed
some CAT activity but only one exhibited detectable BLG expression (Fig.
4
). Southern blotting analyses indicated that the differences in expression
within the variably expressing lines were not due to segregation of two or more
transgenic loci or instability of the DNA sequences at the locus (data not
shown).
DISCUSSION
Silencing of endogenous genes has been reported for a variety organisms,
including mice,
Drosophila
and yeast (
13
-
15
). The silenced state is stable and clonally heritable and is thought to be due
to the packaging of the affected DNA into stable heterochromatic structures (
16
). There is increasing evidence that transgenes are subject to similar effects
and, for example, in
Drosophila
it has been proposed that transgene arrays act as foci for the formation of
heterochromatin leading to variegated patterns of expression (
17
).
The silencing of BLG transgenes observed in this present study contrasts with
our previous reports (
9
,
10
). Thus, when
BLG
was co-injected with FIXD or FIXD[Delta]3' (Fig.
1
) the main effect was activation of expression of the
fIX
construct; silencing of
BLG
was not observed in any of the lines (Table
1
). There are two possibilities for these differences. Firstly, co-integration was achieved by co-injection in our previous experiments, rather than by linking the
two constructs together prior to injection. The ratio of
BLG
to the co-injected second
BLG
-
fIX
transgene was 3:1 and this was broadly reflected in the average composition of
the arrays. These were multicopy and complex comprising
BLG
in tandem, FIXD in tandem as well as the two transgenes adjacent to one another
(
9
). We suggest that
BLG
transgenes in a region of the array containing few copies of FIXD escape
silencing and are expressed. Presumably, these actively expressed
BLG
transgenes suppress the silencing of one or more of the FIXD transgenes and
this accounts for the rescue effect. There appears to be a balance between gene
activation and gene silencing and in the transgene arrays resulting from co-injection of
BLG
and FIXD the former prevails. The efficient silencing of
BLG
observed in the present experiments is presumably due to the fact that the
BLG
gene is invariably adjacent to the
fIX
cDNA or
CAT
sequences because they are physically linked. Although this changes the spacing
between [Delta]DpBLG transgenes in the arrays, this is almost certainly not the cause of
silencing, because transgenes comprising [Delta]DpBLG plus various lengths of 5' or 3'
BLG
flanking sequences are expressed at high levels in all transgenic lines (
8
). In the present experiments the
BLG
gene was not silenced in seven of the 31 lines analysed and this was generally
coincident with expression of the second transgene. We do not know whether this
reflects some positive influences at the site of integration or the structure
of the arrays in these lines. We have not attempted to elucidate the precise
structure of the arrays, since previous experience has taught us that when
there are multiple copies at a locus it can be very difficult, if not
impossible, to deduce the map (
9
).
The failure of
fIX
cDNA sequences to silence
BLG
in co-injection experiments (
9
) could also reflect the fact that greater lengths of 5' and 3' flanking
BLG
sequence were incorporated in the co-injected constructs than were included in the CAFI, BRAT or TARB
constructs. Therefore, the
CAT
and
fIX
cDNA sequences will be positioned closer to the promoter of the
BLG
gene in these tandem arrays than was the case for the co-injected transgenes. It is, therefore, interesting to note that we have
observed silencing of
BLG
when
BLG
and a
BLG
-[alpha]
1AT
construct (AATD) comprising 4.2 kb of 5' and 2.2 kb of 3'
BLG
flanking sequences were co-injected. In these experiments
BLG
and AATD (analogous to FIXD, but comprising [alpha]
1AT
cDNA sequences) were co-injected in a 1:1 ratio. Expression of the AATD construct was rescued (7/9
lines expressed, Table
1
) but the two lines failing to express [alpha]1AT did not express BLG either (
9
). Thus, even though the frequency of gene silencing observed was lower than in
the present experiments, it suggests that these silencing effects can be
transmitted over longer stretches of 5' and 3'
BLG
sequence and that they may occur in arrays generated by co-injection.
Analysis of additional animals in some of the expressing BRAT and TARB lines
showed variable penetrance of
BLG
silencing. It is possible, therefore, that some of the negative lines in this
study would eventually yield an expressing mouse if more individuals had been
analysed. The key point, however, is the comparison with expression of the BLG[Delta]Dp transgene alone, which was expressed at high levels in every mouse in
all 10 of the lines analysed (
8
). Nevertheless, unmodified
BLG
transgenes do appear to be subject to a less extreme form of gene silencing.
This manifests itself as variable levels of BLG expression within some but not
all lines, corresponding to a heterocellular pattern of BLG expression in the
mammary gland (
18
). However, even the lowest expressors exhibited high levels of BLG compared
with most of the expressing lines described in this present study. The variable
silencing seen with some of the BRAT and TARB lines may be a more extreme
version of that observed for
BLG
transgenes alone and the complete silencing observed for the majority of lines the end point.
Repressive effects are thought to be mediated by chromatin at the transgene
insertion site and different insertion sites vary in their ability to repress
expression. The same transgene may be expressed at one integration site but not
at another, giving rise to so-called `position effects'. Predictable expression of reporter sequences
such as
CAT
or cDNA constructs in transgenic mice has always been elusive (
3
) and these types of construct seem particularly susceptible to these effects.
In the light of our data this may be something of an understatement, since the
cDNA and
CAT
reporter constructs appeared to silence the expression of an adjacent
BLG
transgene in a high proportion of the transgenic lines generated. Perhaps the
problem encountered with expressing prokaryotic reporters or cDNAs is not that
they lack the appropriate regulatory elements or that they are prone to
repressive position effects but, rather, that they themselves serve as active
foci for gene silencing. Indeed, these effects are reminiscent of those
described some years ago for plasmid and [lambda] vector sequences, which were shown to silence linked transgenes when
retained in the constructs (
19
). Similarly, a review article (
1
) referred to observations that v-
src
sequences severely inhibited the expression of an
MT
-
hGH
transgene in the liver of transgenic mice when the two sequences were
juxtaposed. Our present study with
BLG
transgenes illustrates how potent this type of silencing can be, although the
degree of the effect will depend on a number of factors, including the
transgene sequences involved, the structure and size of the array and the site
of integration.
One of the best characterized systems of gene silencing is the repression that
occurs at the telomere in yeast. The telomere exerts its suppressive effect
through the collaboration of telomere-specific factors and histones; terminal DNA sequences are recognized by
binding of RAP1 and this initiates polymerization of SIR proteins which
interact with the N-termini of histones H3 and H4, repressing the adjacent domain. Alternative
models for gene silencing in other systems postulate loop domains wherein
factors bound to DNA at specific sites interact to isolate domains
topologically and inactivate the genes therein (
20
). Possibly, mammalian cDNAs and prokaryotic reporter and vector sequences bind
ubiquitous factors that catalyse polymerization or inactive loop formation and
these effects spread to adjacent transgenes and silence them. Alternatively,
they could also serve as foci for methylation or histone H1 deposition. An
obvious feature of both types of sequence is that they lack introns and the corollary to this is that
introns serve to ameliorate these silencing effects. In direct comparisons of
intron-containing genes with their intronless counterparts in transgenic mice,
introns dramatically enhance the efficiency of expression (
2
,
5
,
21
). Recently introns have been shown to stimulate nucleosome alignment in the rat
growth hormone gene, showing that their presence can directly influence
chromatin structure (
22
). We suggest that the chromatin state formed in the absence of introns may be
highly repressive, spread to adjacent genes and probably resembles
heterochromatin.
The frequencies of expression of BLG and BLG-derived transgenes after co-integration by co-injection or ligation prior to injection are compared.
ACKNOWLEDGEMENTS
We thank Roberta Wallace and Francis Thomson for the production and maintenance
of transgenic mice, Claire Neil for technical assistance, Ian Cottingham and
Judith Percy for carrying out the fIX ELISA assays, Anthea Springbett for
statistical advice and Bruce Whitelaw for critical discussions. This work was
supported by the BBSRC and PPL Therapeutics Ltd.
