ABSTRACT
Variegation of transgene expression, a heterocellular or mosaic pattern of
expression seen in all mice in a given transgenic line, is a frequently
observed but unexplained phenomenon. We have encountered variegation with
globin transgenes; when
lacZ
expression is driven by globin control elements a proportion of erythrocytes
express
[beta]
-galactosidase (
[beta]
-gal), while the remaining erythrocytes express none. The percentage of
expressing cells is constant within each line (at any particular developmental
stage), but varies between lines. Such variation may account for much of the
line-to-line variability which has been reported in the expression of a
transgene construct. We have now extended these observations by studying
expression of several globin/
lacZ
transgenes with increasing age. Expression of
[beta]
-gal is variegated in all lines in adult mice, including those made with a
[beta]
-globin promoter and locus control region driving
lacZ
. The extent of variegation differs widely between lines, but in all lines there
is a marked decline in the number of erythrocytes expressing
[beta]-gal with increasing age. Progression of silencing continues long past the
point at which globin switching is complete, suggesting that it is not related
to this process. We observe that age-dependent silencing is most severe in high copy number animals. Increasing
variegation of transgene expression with ageing of mice is likely to complicate
interpretation of the developmental regulation of transgenes. We speculate that
it reflects a general mechanism of epigenetic regulation.
Transgenic animals have been widely employed in the analysis of DNA sequences
responsible for expression of genes in specific lineages and developmental stages. These mice carry an expression unit as an array containing one to dozens of copies. While transgenic analysis
has proven a powerful method of analysing complex mechanisms of gene
regulation, it is clear that transgenes often do not behave as independent
units, but rather are significantly and variably influenced by factors such as
the site of integration and the number of transgene copies in an array. These
influences sometimes lead to marked variations in expression patterns between
different transgenic lines carrying the same construct.
Regulation of the globin genes has been intensively and productively studied
with transgenic mice, but the problems mentioned above are prominent
confounding factors in these studies. Because regulation of the human globin
genes in mice follows a developmental pattern similar to that seen in humans (
1
), transgenic mice have in recent years been used to study the control of this
process. The expression of human globin genes follows a coordinated temporal
programme in specific erythroid tissues. The embryonic [zeta] and [epsilon] genes, from the [alpha] and [beta] clusters respectively, are expressed in yolk sac
blood islands. Later, when the site of erythropoiesis shifts to the fetal
liver, the embryonic globins are replaced by the [alpha]- and [gamma]-globins. Finally, in the adult erythropoiesis moves to
the bone marrow and spleen, where the [alpha] gene continues to be expressed and [beta] gene expression replaces that of [gamma]. Sequences far upstream of the globin structural genes are
required for erythroid-specific expression of genes from both the [alpha]- and [beta]-globin clusters. The [beta] locus control region ([beta] LCR) spans a region from 6 to 20 kb
upstream of the [epsilon]-globin gene and contains four DNase I hypersensitive sites (HS1-4) (
2
-
4
), while the [alpha] locus has an analogous element ([alpha]HS-40) associated with a hypersensitive site 40 kb upstream of
the [zeta]-globin gene (
5
,
6
). Both the [beta] LCR and the [alpha]HS-40 confer high level erythroid-specific expression on linked promoters. However, the [alpha]HS-40 is apparently not equivalent to the [beta]-globin LCR in its ability to
confer position independence or copy number dependence (
7
-
9
).
We have previously described variegation in transgenic mouse lines which express
[beta]-galactosidase ([beta]-gal) under the control of globin promoters linked to [alpha]HS-40 (
10
). Use of [beta]-gal permitted analysis of transgene expression in individual red
blood cells, while nearly all other globin transgenes have been assayed by
analysis of RNA from pooled cells. We found that expression of
lacZ
within all mouse lines is heterocellular. Within a given mouse individual
erythroid cells either do not express the transgene at all or express it at a
level which is characteristic of each line. The number of [beta]-gal expressing cells varies greatly between different lines of
transgenic mice carrying the same construct, but is consistent within a given
line, suggesting that the degree of heterocellular expression is determined at
least in part by the site of integration. This finding is directly relevant to
the interpretation of transgenic studies. The heterocellular expression we
observe is similar to position-effect variegation (PEV) seen in
Drosophila
, yeast and in mice with X chromosome translocations (
11
). In PEV, genes translocated to new chromosomal regions are inactivated in a
stochastic manner. This inactivation is associated with proximity to
heterochromatic regions, which may spread along a chromosome and exert a
repressive effect on the expression of flanking genes, and is also influenced
by the
cis
-acting elements flanking the gene (
12
) and the presence of multiple copies of the transgene (
13
).
The present study examines the effect of age on transgene expression in mice
harbouring
lacZ
transgenes driven by different globin control elements. We find that with all
constructs expression is variegated at certain developmental stages, although inclusion of the entire [beta]-globin LCR results in pancellular expression at early stages. Most
notably there is a progressive decrease in the proportion of expressing
erythrocytes in all lines and this decrease continues far past the point at
which globin switching is complete, implying that this silencing is not
equivalent to normal globin developmental regulation. The degree to which
silencing progresses after birth is related to the copy number of the
transgene, however, variegation is observed even in low copy number animals.
These observations indicate that multiple factors may contribute to variegation
of transgene expression, including the site of integration and nature of the
transgene array.
All constructs were synthesized by linking portions of the human globin
promoters to a construct containing the reporter gene
lacZ
, which produces [beta]-gal (a gift from Dr J.Rossant, Mount Sinai Research Institute,
Toronto) (see Fig.
1
). The 550[zeta] and 127[zeta] constructs contain from +6 to -550 and from +6 to -127 respectively of the human [zeta]-globin promoter. The 4 kb
Hin
dIII fragment
containing the [alpha]HS-40 element (
6
) was inserted into the 550[zeta]R construct in the opposite orientation relative to the 550[zeta] and 127[zeta] constructs. The [beta] construct contained the same
lacZ
reporter gene linked to the [beta]-globin promoter from +37 to -613 and the [beta]-globin miniLCR (
2
) with HS1-4. Prior to microinjection DNA fragments were excised from vector
sequences with
Kpn
I for the 550[zeta]R construct,
Not
I/
Kpn
I
for the 550[zeta] and 127[zeta] constructs and
Not
I for the [beta] construct and purified by agarose gel electrophoresis.
Transgenic mice were generated by microinjection of linear DNA fragments into
the pronuclei of fertilized eggs from the outbred P.O. mouse strain by standard
procedures (
14
). Transgenic progeny were identified and copy number was determined by Southern
analysis of tail DNA. Hemizygous lines were established by mating transgenic
founders to P.O. mice.
Embryos were obtained 12.5 and 17.5 days post-coitum (d.p.c.) from P.O. females mated to hemizygous transgenic males.
After bleeding whole embryos into phosphate-buffered saline (PBS) erythrocytes were gently spun in a microcentrifuge
and then fixed in 0.25% glutaraldehyde, washed and stained with 5-bromo-4-chloro-3-indolyl [beta]-D-galactopyranoside (X-gal) as described (
10
,
15
) for at least 24 h at 37oC. We have shown previously that all cells containing [beta]-gal activity can be detected by light microscopy following
these staining conditions (
10
). Adult peripheral blood obtained from hemizygous transgenic mice from 1 week
of age onwards was stained with X-gal in a similar way. All the data in this paper has been collected from
established transgenic lines.
Table 1
.
This was done using a similar method to that reported previously (
16
,
17
). Embryos at 12.5 d.p.c. were dissected and bled into PBS to give a cell
suspension at ~1-5 * 10
7
cells/ml. Aliquots of 30 [mu]l of this cell suspension were dispensed into 5 ml polystyrene tubes and
incubated for 5 min at 37oC. An aliquot of 50 [mu]l 4 mM fluorescein digalactoside (FDG; Molecular Probes, Eugene, OR) in
reverse osmosis grade water, pre-warmed to 37oC, was then added to each tube. The tubes were held in a 37oC water bath for 75 s, after which time the tubes were removed
from the water bath and quickly filled with ice cold PBS. The cells were
pelleted by centrifugation at 4oC for 5 min at 300
g
. The cells in each tube were gently resuspended in 0.5 ml ice-cold PBS and held on ice protected from light before analysis in a Becton
Dickinson FACsort. Primitive erythrocytes from non-transgenic embryos were also analysed to determine the level of
autofluorescence in non-expressing cells. The percentage of [beta]-gal-positive cells in transgenic embryos was determined from
the number of cells which showed fluorescence above this background level.
Mouse lines (4-8 for each construct) were generated carrying [zeta] promoter/
lacZ
/[alpha]HS-40 transgenes or [beta] promoter/
lacZ
/[beta] LCR transgenes (see Fig.
1
). When embryos were stained with X-gal to assay for transcription from the
lacZ
reporter transgene [beta]-gal expression was detected in circulating primitive erythrocytes
from the majority of lines (Table
1
), however, in some lines no expression of the transgene was found, even after
careful examination of ~10
6
erythroid cells. In some lines derived from the [zeta] promoter constructs [beta]-gal expressing cells were readily detected in a variety of
other tissues in a pattern unique to each mouse line (Table
1
). Ectopic expression is not unique to mice carrying [zeta] promoter transgenes; we have found ectopic expression of constructs
driven by the [alpha]-globin and TK promoters when linked to [alpha]HS-40 (H.Sutherland and E.Whitelaw, unpublished data). No
lines carrying the [beta] promoter/
lacZ
/[beta] LCR transgene showed ectopic expression, consistent with the idea that
the [beta] LCR, but not the [alpha]HS-40, is able to suppress expression in non-erythroid lineages.
Traditionally analysis of globin transgene expression has involved the
measurement of mRNA levels in cell lysates, which gives an average of
expression in all cells. We developed a technique for staining circulating
erythrocytes obtained from dissected embryos and adults permitting the
detection of [beta]-gal in individual erythrocytes and found that mice carrying a hybrid
[gamma]/[zeta]-globin promoter/
lacZ
construct expressed the transgene in only a portion of erythrocytes. Moreover,
the percentage of positive cells varied markedly between transgenic lines (
10
). We have expanded this analysis with the constructs shown in Figure
1
. These contain the [zeta]-globin promoter truncated to -127 or -550 or the [beta]-globin promoter to truncated to -600. The [zeta]-globin promoter
constructs also contain the [alpha]HS-40 element and the [beta]-globin promoter construct contains the miniLCR
cassette, which consists of the four upstream DNase I hypersensitive sites of
the human [beta]-globin locus. When peripheral blood from 12.5 d.p.c. embryos
carrying the 127[zeta] promoter transgene was stained with X-gal it was found that not all cells expressed [beta]-gal. Moreover, the percentage of positive cells varied
markedly between transgenic lines (carrying the same construct) (Table
2
). Within a given transgenic line, however, the percentage of positive cells did
not vary significantly between individual mice (Table
2
). This can be seen by the relatively small standard errors for each data point. These findings are consistent with our previous
observations with a different transgene construct (
10
). Since the variation from line to line does not correlate with copy number
(compare Tables
0
and
1
), we conclude that these transgenes are being influenced by the site of
integration. Thus the short [zeta] promoter (127[zeta]) construct directs expression in erythroid cells, but expression is
heterocellular and clearly influenced by position effects. In an attempt to
overcome these effects we went on to produce mice with a larger promoter (550[zeta]) and with the [alpha]HS-40 in the opposite orientation (550[zeta]R). However, mice expressing these transgenes also did
so in a heterocellular manner (Table
2
). Finally, we produced mice with a [beta] promoter/[beta] LCR construct and these did show pancellular expression at this
early stage of development. Pancellular expression is consistent with previous
reports that the [beta] LCR confers integration site independence (
3
). In one of the lines carrying the [beta] construct, [beta]1, 56% of erythroid cells are expressing, but the transgene in this
line has integrated into the X chromosome and so the mosaic expression is
presumably a result of random X inactivation. All positive embryos analysed
here were female, since they resulted from the mating of a transgenic male with
a wild-type female.
When we determined the percentage of expressing cells at later stages of
development we found that the proportion of expressing cells decreased in mice
carrying both [zeta] and [beta] transgenes (Table
2
). This decrease in the [zeta] promoter transgenic mice could be interpreted as being consistent with
correct developmental stage-specific regulation at the promoter. However, in this case we would expect
expression to be off in all lines after birth and this is not the case: 550[zeta]1, 550[zeta]3, 127[zeta]2 and 127[zeta]5 still have significant numbers of expressing erythroid
cells in the adult. Furthermore, the erythroid cells from mice carrying the [beta] promoter also show a significant drop in the percentage of `on' cells as
the animals age and this is clearly inconsistent with the behaviour of the
endogenous adult [beta]-globin gene. Most previous studies of the behaviour of [beta]-globin transgenes in mice have not investigated
transgene activity beyond the late fetal stage, because high levels of
transgene expression lead to globin chain imbalance, thalassaemia and fetal
death. By using the
lacZ
reporter gene we can follow expression throughout the life of the mouse.
Photographs of the erythroid cells purified at different stages of development
from two lines are shown in Figure
3
. At 12.5 d.p.c. in line 550[zeta]1 (Fig.
3
A) all cells were stained blue, while in line 550[zeta]R4 (Fig.
3
D) there were clearly unstained primitive erythrocytes. At later stages of
development and after birth mice from both lines showed a lower percentage of
erythroid cells which stained blue. In fact, at 8 weeks in line 550[zeta]R4 there were very few expressing cells (Fig.
3
F), with examination of many similar fields of view needed to find a positive
cell. Thus in both lines the number of expressing cells declines with age.
The rate at which the transgene is switched off during adult life varied from
line to line; 127[zeta]5 dropped only 2-fold from 1 to 8 weeks of age, while another line with the same
construct, 127[zeta]4, dropped 40-fold over the same period. In the [beta] promoter/
lacZ
/[beta] LCR lines there was also a large but highly variable decline in the adult
stage; line [beta]5 dropped 40-fold between 1 and 8 weeks, while line [beta]3 dropped only 4-fold. In Figure
4
the decline in the percentage of [beta]-gal expressing cells between 1 and 8 weeks for the 550[zeta],127[zeta] and [beta] lines is compared with their copy number. The
decline was greatest for lines which have the highest copy number. All the
lines with the 550[zeta]R construct had already declined to a low percentage by birth and so were
not included in this comparison.
We have extended our studies of variegated transgene expression in mice and
confirmed that this phenomenon is prevalent in globin transgenes. All lines in
which expression of the
lacZ
transgene occurred, including those in which expression was driven by the [beta]-globin LCR, exhibited some degree of variegation (heterocellular or
mosaic expression) at some developmental stage. Moreover, we observed a general
tendency for expression of the transgene to decline with age, well past the
point at which definitive erythrocytes have replaced those of earlier stages
and globin switching is complete. The extent of the decline was greatest in
lines with large numbers of copies of the transgene. These findings of
variegation and a correlation with copy number are similar to a large number of
observations made in other systems (
13
) and suggest that the interpretation of transgene activity requires
consideration of these factors.
Variegation of transgene expression has been widely observed in mice, although
it has been commented on very little (
10
). Classical PEV was described in chromosomal translocations that place a marker
gene near constitutive heterochromatin, resulting in stochastic and clonally
heritable inactivation of the marker (
11
). Related phenomena have been described when genes are placed in or near
inactive chromatin at telomeres, the X chromosome and the mating type loci in
yeast and when plasmids are randomly integrated in mammalian cell lines, and
have also been described in plants (
18
-
23
). When this extensive evidence for stochastic silencing of transgene expression
is considered, variegation of globin transgenes is not surprising. Other work
suggests that this phenomenon is not confined to
lacZ
constructs. Heterocellular expression has been observed in mice carrying human
and murine [beta] LCR elements linked to human globin genes (
24
,
25
). In studies of inbred albino and black mice carrying transgenes with a
tyrosinase cDNA a genetically stable pattern of coat colour variegation was
observed (
26
,
27
). Other examples of transgenes that exhibit variegated expression include
myelin basic protein antisense cDNA (
28
) and a hypoxanthine phosphoribosyltransferase transgene on the Y chromosome (
29
). The globin/
lacZ
transgene permits convenient observation and quantification of expression in
individual erythrocytes, as does another recently described system that uses
expression of human CD2 in T cells (
30
). These transgenes may be useful in analysing the factors responsible for
variegation. Transgene silencing may be associated with methylation of the DNA
and we are currently investigating this possibility. The factors responsible
for variegated transgene silencing are likely to be considerably more complex
than the effect of flanking heterochromatin and include the
cis
-acting control elements flanking the transgene (which suppress silencing)
(
12
), the lineage in which the gene is expressed and the presence of genetic
modifiers (
11
). Variegation has also been found to be associated with tandem arrays of an
integrated construct, which may form foci of heterochromatinization (
13
,
31
,
32
), and this is consistent with our observation that age-related transgene silencing is greatest in lines with higher transgene
copy number.
This general decrease in transgene expression during adult life is the most
dramatic observation made in this study. Only a few studies (
7
,
8
,
33
) have looked in detail at globin transgene expression in older adults and in
all of these a decline in expression was observed with increasing age. The
percentage of erythroid cells expressing a [gamma]-globin transgene was found to decrease by 50% between 2 and 6
months after birth (
33
). Since these studies used globin coding sequences and not
lacZ
, this implies the phenomenon is not an artifact of the
lacZ
gene. One trivial explanation for the decrease which we see in the percentage
of expressing cells with age would be that the transgene array is unstable in
somatic cells. However, detailed Southern transfer analysis and sequencing of
PCR-amplified transgene DNA using tail DNA from mice at different ages
suggests that this is not the case (
34
; data not shown). The fact that the rate of silencing during adult life varies
from line to line suggests to us that it may depend upon the site of
integration and/or the nature of the transgene array, however, we see no
correlation between the rate of decline and the initial level of heterocellular
expression (percentage of `on' cells at 12.5 d.p.c.), suggesting that age-dependent silencing is independent of site of integration. We do see a
correlation between the transgene copy number and the rate of silencing; the
higher the copy number the faster the silencing. Lines with very low copy
number show little decline in numbers of expressing cells after birth, but
expression is variegated. This suggests that the integration site may exert a
constant influence on the probability of expression, while large arrays are
progressively silenced. It is worth noting here that we see no correlation
between copy number and level of expression in each `on' cell (
13
), excluding the possibility that silencing is greater in high copy number mice
simply because higher levels of [beta]-gal place the `on' cells at a selective disadvantage.
Silencing of transgene expression as a result of multicopy arrays is a
widespread phenomenon and has recently been reviewed at length by Dorer (
13
). Observations similar to ours have been made in both
Drosophila
(
31
) and plants (
35
). Expression of most mouse transgenes is analysed with methods that produce an
average of expression in a whole tissue or lineage. Silencing associated with
high copy number would in these systems appear as a lower expression per copy
with increasing copy number, and this correlation has been observed (
7
,
8
,
9
,
36
). We suggest that the late silencing of
lacZ
expression in this study is due to progressive heterochromatinization of larger
transgene arrays. In a previous study (
33
) in which [gamma]-globin transgene expression was found to be heterocellular the rate
of transgene silencing was found to be relatively low (2-fold over a 4 month period). Interestingly, these mice contained only
three copies of the transgene.
In summary, we have shown that erythroid cells from mice carrying the human [zeta]- or [beta]-globin promoter attached to the
lacZ
gene display variegated silencing and that the number of expressing cells
declines with age. The extent of silencing varies from line to line, even
between lines with the same transgene copy number, suggesting that it is at
least in part dependent on the site of integration of the transgene. The fact
that variegation occurs in all of the lines we have made suggests that the
cis
-acting control elements in our constructs (even the [beta]-globin miniLCR) are sufficiently weak that they are unable to
overcome repressive chromatin effects in all cells. Presumably, the endogenous
globin genes, like most other genes, lie within a chromosomal region in which
sufficient regulatory elements ensure that transcription will occur. The
expression of globin/
lacZ
transgenes in erythroid cells permits the convenient quantification of
variegation in a mammalian tissue during development. Coupled with the ability
to separate expressing from non-expressing cells by FACS-Gal, this system should enable study of the factors involved in
transgene silencing. It also provides an assay to test the ability of elements
such as `insulators' and `boundary elements' that lie within or flank a
transgene to protect against general transgene silencing in mammals. Variegated
silencing and the influence of multicopy arrays should be taken into account
when analysing transgene expression, even with transgenes that are not amenable
to single cell analysis. An understanding of this phenomenon has far reaching
implications for the stability of novel phenotypes created by genetic
engineering.
This study was supported by a grant from the National Health and Medical
Research Council of Australia (to EW) and by NIH grant 5RO1HL48790 (to DM). DM
is supported by the James S.McDonnell Foundation and is a Scholar of the
Leukemia Society of America. We would like to thank Wu Wenlian, Margot Kearns
and Peter Smith for their contributions and Merlin Crossley for helpful
comments on the manuscript.
Construct
Line
Copy
Erythroid
Ectopic
number
expression
expression
a
127[zeta]
127[zeta]1
50-100
-
n.d.
127[zeta]2
10-15
+
-
127[zeta]3
3-5
-
n.d.
127[zeta]4
15-20
+
-
127[zeta]5
8-10
+
Nose, ventral spinal chord, hindbrain
127[zeta]6
1
-
-
127[zeta]7
1
-
-
127[zeta]8
1
-
-
550[zeta]
550[zeta]1
3-5
+
Limb bud (AER,ZPA), notochord, nose
550[zeta]2
15-20
+
Limb bud (AER)
550[zeta]3
1-2
+
Midbrain, hindbrain, mouth, limbs, somites
550[zeta]4
3-5
-
-
550[zeta]R
550[zeta]R1
10-15
+
-
550[zeta]R2
10-15
+
-
550[zeta]R3
10
+
Ear, nose, mouth
550[zeta]R4
10
+
Ear, nose, mouth, mid/forebrain junction
550[zeta]R5
40-50
+
Midbrain
[beta]
[beta]1
30
+
-
[beta]2
3-5
+
-
[beta]3
5
+
-
[beta]4
15-20
+
-
[beta]5
30
+
-
REFERENCES
Return