To understand the molecular mechanism underlying JAK/STAT-mediated tumour formation, we have previously investigated the role of JAK in a Drosophila
leukaemia model, in which a hyperactive mutant form of JAK (Tum-1) causes leukaemia-like overproliferation of blood cells5,8
. We have demonstrated that oncogenic JAK disrupts heterochromatin formation globally, allowing transcriptional activation of genes that are not necessarily direct targets of STAT7,9
. The molecular mechanism underlying the effects of Hopscotch (Hop, Drosophila
JAK) on heterochromatin remains unclear. It may be mediated by phosphorylation of STAT92E, as in the canonical JAK/STAT pathway10,11
. Alternatively, Hop may activate cellular targets other than STAT92E9
To investigate whether disruption of heterochromatin induced by Hop-activation7
is mediated by STAT92E, we examined the effects of reducing stat92E+
dosage on heterochromatic gene silencing, which can be measured by position-effect variegation (PEV)12
. As shown previously7
, Hop gain- and loss-of-function mutations suppress and enhance PEV, respectively (). We found that reducing stat92E+
dosage had the opposite effect of reducing Hop and strongly suppressed PEV, causing marked derepression of white+
in two independent variegated lines DX1
, resulting in a substantial increase in red eye-pigmentation (). Heterozygosity of stat92E
did not affect expression in a control P[white+]
) element, which does not induce heterochromatin formation () or the white+
gene in its original locus (data not shown; also see ref. 7
), suggesting that P[white+]
is not simply a transcriptional target of STAT92E. Thus, in contrast to Hop, which counteracts heterochromatic gene silencing7
, STAT92E seems to promote this silencing process.
Figure 1 STAT92E is required for heterochromatic gene silencing. (a–c) Effects of mutations in hop and stat92E on white+ variegation induced by transgene repeats, DX1 (a) or a control transgene (b) are shown as changes in red eye-pigmentation in representative (more ...)
We next examined the epistatic relationship between HP1 and STAT92E (). HP1, encoded by Su(var)205
, is a constitutive component of heterochromatin and is essential for heterochromatic gene silencing12,13
. Consistent with the results described above, we found that increasing stat92E+
dosage by a chromosomal duplication (referred to as 3 × stat92E+
, ) or a stat92E+
transgene (Supplementary Information, Fig. S1
) enhanced heterochromatic gene silencing, resulting in complete silencing of the variegated white+
gene. This effect was antagonized, however, by reducing HP1+
gene dosage (). It was also evident that the effect of halving HP1+
gene dosage on derepression of the silenced white+
gene was strongly opposed by the presence of 3 × stat92E+
(compare ). Thus, increasing STAT92E levels promotes gene silencing even when HP1 levels are reduced. Consistent with previous reports14
, increasing HP1 levels by expression of an hsp70–HP1+
transgene promoted heterochromatin formation, as shown by increased variegation (). We found that halving stat92E+
gene dosage counterbalanced the effects of increased HP1, allowing increased white+
expression (). However, expression levels of white+
double-heterozygous flies were much lower that those in stat92E+/−
flies (compare ), suggesting that increasing HP1 levels also antagonizes the effects of reducing stat92E+
dosage. Thus, STAT92E and HP1 are interdependent and both required for heterochromatic gene silencing.
We next investigated the requirement for STAT92E in heterochromatin formation by genetic mosaic analyses (see Methods) using HP1 and dimethylated histone H3mK9 as markers in larval salivary glands7,13
. We found that clonal overexpression of stat92E+
resulted in higher levels of heterochromatin (H3mK9 and HP1 accumulation), compared with neighbouring wild-type cells (). An increase in the size of the HP1 foci was also found in cells overexpressing a STAT92E–GFP fusion protein (see below). Conversely, in stat92E−/−
cells, heterochromatin levels, as characterized by HP1 foci, were markedly reduced, compared with neighbouring wild-type cells (). Notably, the effects of STAT92E overexpression were opposite to those of JAK overactivation on heterochromatin shown previously7
, consistent with the opposite effects we observed on PEV. Thus, STAT92E levels seem to be important for heterochromatin formation or stability.
Figure 2 STAT92E levels control heterochromatin abundance and HP1 localization. (a, b) Third-instar salivary glands overexpressing stat92E+ in random cells marked with GFP (green) and stained with anti-H3mK9 (magenta in a) or anti-HP1 (magenta in b) are shown (more ...)
To further confirm that STAT92E is required for HP1 localization on heterochromatin, we performed chromatin immunoprecipitation (ChIP) using an anti-HP1 antibody before and after stat92E
knockdown by RNAi in Drosophila
S2 cells. We chose the transposable element 1360
as a heterochromatin marker to detect the co-immunoprecipitated heterochromatin. 1360
is a repetitive sequence found most abundantly in constitutive heterochromatin regions of all Drosophila
chromosomes and is believed to be essential for initiating heterochromatin formation15,16
. Enrichment of HP1 binding to 1360
sequences has been detected previously by ChIP17
. We found that association of HP1 with 1360
was significantly reduced following stat92E
RNAi knockdown (). Thus, STAT92E is essential for the association of HP1 with heterochromatin.
To investigate whether changing STAT92E levels alters HP1 localization in general, we examined additional developmental contexts. First, HP1 localization in heterochromatin can be observed in cellularization-stage (nuclear cycle 14) embryos as foci enriched at the apical region of elongated wild-type nuclei13,18,19
(). This pattern of HP1 distribution, however, was absent in mutant embryos lacking maternal stat92E+
(referred to as stat92Emat−
embryos; see Methods; ). In these stat92Emat−
embryos, HP1 protein assumed a punctate and clustered distribution that was no longer restricted to the apical heterochromatin, whereas there were no detectable changes in the total levels of HP1 (compare ). This is consistent with the notion that STAT92E is required for HP1 localization in centromeric heterochromatin. Second, we examined stat92E
mutant clones in imaginal discs. The small size of imaginal disc cells does not allow examination of HP1 sub-nuclear distribution, but does allow assessment of total HP1 levels. We found no detectable changes in HP1 levels in stat92E−/−
imaginal disc cells, compared with neighbouring wild-type cells (Supplementary Information, Fig. S2
), suggesting that loss of STAT92E does not immediately cause a substantial reduction in HP1 protein levels. However, long-term loss of stat92E+
indeed reduced HP1 levels (Supplementary Information, Fig. S3
), suggesting that HP1 transcription may be regulated directly or indirectly by STAT92E. Finally, we examined heterochromatin by performing blots of total protein from embryo extracts. We found that H3mK9 levels were increased significantly in embryos that overexpressed STAT92E and decreased in stat92Emat−
embryos, whereas HP1 levels were not changed and moderately reduced, respectively (). These results indicate that, although the canonical JAK/STAT pathway may regulate HP1 transcription, both loss of STAT92E and overactivation of JAK cause HP1 delocalization and overall loss of H3mK9.
The observation that loss of stat92E+
is equivalent to JAK overactivation and that increasing STAT92E levels has similar effects to loss of hop+
on heterochromatin formation seemed paradoxical, as JAK activates STAT in the canonical JAK/STAT pathway3,4
. To understand how STAT92E levels affect heterochromatin formation, we examined STAT92E protein localization by immunostaining with an anti-STAT92E antibody (see Supplementary Information, Figs S4
for specificity). We found that STAT92E was localized mainly in the nucleus in discrete patterns, with the highest levels colocalizing with the HP1 foci (). Interestingly, it has been shown that unphosphorylated mammalian STAT3 is localized predominantly in the nucleus20
, despite the conventional view that it should be in the cytoplasm. STAT92E colocalization with HP1 in heterochromatin regions was also detected in larval fat bodies or using a different anti-STAT92E antibody (data not shown), or in glands overexpressing a STAT92E–GFP transgene, or by ChIP with S2 cells (see below), indicating that STAT92E is indeed localized in heterochromatin. Nuclear localization of STAT92E–GFP is also evident in unstimulated S2 cells21
. Moreover, staining of squashed polytene chromosomes from wild-type salivary glands indicated that STAT92E and HP1 colocalize in several regions of heterochromatin, including the chromocentre and telomere (). These results suggest that a significant amount of STAT92E is normally localized on heterochromatin.
Figure 3 STAT92E colocalizes and physically associates with HP1. (a–c) Third-instar wild-type salivary glands were stained with anti-HP1 (red) and anti-STAT92E (green) antibodies in whole-mount tissues or squashed polytene chromosomes. Representative giant (more ...)
As STAT92E colocalizes with HP1, we performed co-immunoprecipitation studies to explore the possibility that STAT92E may interact physically with HP1. Indeed, we found that bacterially expressed GST–HP1 can pull down the endogenous STAT92E from embryo extracts (), and that HP1 and STAT92E co-immunoprecipitate when transfected in S2 cells (, lane 4). STAT92E contains a perfect and an imperfect HP1-binding sequence motif PVL22
(). We found that mutations of only one or the other site had minimal effects on STAT92E–HP1 binding (, lanes 5, 6). However, mutations of both sites abolished the STAT92E–HP1 interaction (, lane 7), suggesting that either of the two sites may be sufficient to mediate STAT92E binding to HP1. Moreover, we found that endogenous HP1 can co-immunoprecipitate with STAT92E (, left). Despite the presence of higher levels of both HP1 and STAT92E proteins in hop-Tum–l/+
embryos, less HP1 was co-immunoprecipitated with STAT92E from these embryos (). This suggests that unphosphorylated STAT92E binds to HP1, as the ratio of unphosphorylated-to-phosphorylated STAT92E levels was lower in hopTum–l/+
than in wild-type embryos (caused by increased STAT92E phosphorylation). Finally, increasing STAT92E phosphorylation, by co-transfecting Hop, reduced the amount of HP1 co-immunoprecipitated with transfected STAT92E (). These results suggest that unphosphorylated STAT92E can associate with HP1.
We next determined whether the heterochromatin-associated STAT92E is ‘active’ (that is, tyrosine-phosphorylated) by staining with an anti-phospho-STAT92E antibody23
(see Supplementary Information, Figs S4
for specificity). Compared with total STAT92E, phosphorylated STAT92E seemed to be distributed more uniformly in the nucleus and did not colocalize with HP1 (). Moreover, we examined squashed polytene chromosomes from larvae expressing STAT92E–GFP by immunostaining and found that STAT92E–GFP was colocalized either with pSTAT92E or HP1 signals, but rarely with both (). These results indicate that it is the unphosphorylated form of STAT92E that colocalizes with HP1 on heterochromatin.
Figure 4 Only unphosphorylated STAT92E is localized on heterochromatin. (a) Wild-type larval salivary gland stained with anti-HP1 and anti-pSTAT92E antibodies. Note that the signals do not overlap. (b) In polytene chromosomes from sgs–Gal4<UAS–stat92E–GFP (more ...)
To investigate how STAT92E activation by phosphorylation leads to heterochromatin destabilization, we examined STAT92E–GFP and HP1 localization in ex vivo
cultured salivary glands at different times after stimulating STAT92E phosphorylation with pervanadate (H2
/vanadate), which potently and rapidly increases STAT92E phosphorylation (Supplementary Information, Fig. S6a
. Before stimulation, STAT92E–GFP was localized in both cytoplasm and nucleus, with the nuclear portion mostly colocalized with HP1 (). The higher levels of STAT92E–GFP than endogenous STAT92E in the cytoplasm (compare with ) may be caused by transgene overexpression. At 20 min post-stimulation, the nuclear STAT92E–GFP was enriched, due to entry of the cytoplasmic protein, and was not colocalized with HP1 (). This pattern of STAT92E–GFP resembles that of phosphorylated STAT92E (). We suggest that the change in STAT92E–GFP distribution is caused by nuclear STAT92E–GFP moving away from heterochromatin and cytoplasmic STAT92E–GFP moving into the nucleus following phosphorylation. At 60 min post-stimulation, STAT92E–GFP was mostly bound to euchromatic chromosomes in distinct bands and HP1 had diffused from heterochromatin (). When an unphosphorylatable mutant stat92EY704F
–GFP was overexpressed, however, pervanadate treatment did not cause complete HP1 dispersal and STAT92EY704F
–GFP remained colocalized with HP1 (). Thus, STAT92E dispersal precedes, and is required for, HP1 displacement from heterochromatin. This conclusion is consistent with observations of HP1 and STAT92E distribution in hopTum–l
mutant larval salivary glands (Supplementary Information, Fig. S5
). It thus seems that unphosphorylated STAT92E is associated with heterochromatin and phosphorylation causes STAT92E dispersal and translocation to euchromatic regions, presumably binding to cognate promoters.
To confirm this observation, we carried out ChIP experiments in Drosophila S2 cells to determine the effects of pervanadate treatment on the abundance of 1360 (heterochromatin) and stat92E (target gene) promoter sequences in ChIP, using an anti-STAT92E antibody. Indeed, we found enrichment of 1360 sequences in STAT92E–chromatin complexes before, but not after, treatment with pervanadate (). The same treatment caused enrichment of phosphorylated STAT92E in the promoter of stat92E itself ().
To investigate whether STAT92E activation could affect heterochromatin indirectly by transcriptionally inducing other genes, we blocked protein synthesis by treating salivary gland with cycloheximide before pervanadate stimulation. We found that cycloheximide treatment did not prevent pervanadate-induced HP1 dispersal (), but rather, accelerated heterochromatin loss (). This suggests that heterochromatin destabilization induced by STAT92E activation does not require synthesis of new protein and therefore, cannot be mediated by induction of STAT92E transcription targets.
Figure 5 Protein-synthesis-independent heterochromatin disruption following STAT92E phosphorylation. (a) Wild-type salivary glands were cultured with or without cycloheximide (CHX) for 1 h before addition of pervanadate for 1 h in vitro. Represented nuclei are (more ...)
To further explore the consequences of cycloheximide and pervanadate treatment on STAT92E activation and heterochromatin stability, we examined total protein blots of S2 cell extracts at different times after drug treatment. STAT92E phosphorylation (activation) was evident 5 min after pervanadate treatment and reached a maximum at 1 h in the absence of cycloheximide treatment, but continued to increase in the presence of cycloheximide (; Supplementary Information, Fig. S6a
). The increase in phosphorylated STAT92E levels occurred at the expense of unphosphorylated STAT92E, which was most evident when synthesis of new protein was blocked (, second panel, lower band), such that after 4 h of drug treatment, unphosphorylated STAT92E levels were significantly reduced (, second panel, lower band, lane 9). Drug treatment did not significantly affect HP1 levels () and cycloheximide treatment alone had no effects on HP1 localization (), H3mK9 or STAT92E levels or its phosphorylation status (Supplementary Information, Fig. S6b
). However, a decrease in the heterochromatin marker H3mK9 was detected 2 h after combined pervanadate and cycloheximide treatment ().
These results suggest that, in the presence of cycloheximide, pervanadate causes a greater increase in phosphorylated STAT92E levels and a greater decrease in unphosphorylated STAT92E levels, presumably caused by a lack of both positive and negative feedback mechanisms. The negative feedback on STAT activation is mediated by induction of inhibitors such as SOCS36E and PTP61F, which reduce pSTAT92E levels, and the positive feedback by STAT92E autoregulation, which increases STAT92E levels10
. Both feedback mechanisms require new protein synthesis, which was blocked by cycloheximide. The resultant loss of unphosphorylated STAT92E after combined drug treatment would thus lead to accelerated heterochromatin destabilization. Taken together, these results suggest that activation of STAT92E by phosphorylation causes STAT92E dispersal, leading to HP1 displacement and heterochromatin destruction in a transcription-independent manner ().
The results of these genetic and biochemical studies suggest a model in which at least a portion of unphosphorylated STAT is normally localized in heterochromatin, where STAT activation causes its diffusion from heterochromatin and relocalization to euchromatin, leading to HP1 dispersal and heterochromatin destabilization (). Our results, however, do not rule out the possibility that activation of JAK reduces methylation of histone H3mK9 through a STAT-independent mechanism. We envision that under physiological conditions, a low-intensity signal may be sufficient for activating specific STAT target genes without affecting heterochromatin, whereas high intensity or persistent activation of JAK/STAT may bring about widespread heterochromatin destruction. It remains to be determined whether JAK enters the nucleus to phosphorylate STAT or whether the redistribution of unphosphorylated nuclear STAT is a result of an altered equilibrium between nuclear and cytoplasmic or phosphorylated and unphosphorylated STAT in response to JAK activation.