Deficiency of Mbd2 suppresses intestinal tumorigenesis Owen J Sansom1, Jennifer Berger2, Stefan M Bishop1, Brian Hendrich2, Adrian Bird2 & Alan R Clarke1 Gene silencing through de novo methylation of CpG island promoters contributes to cancer. We find that Mbd2, which recruits co-repressor complexes to methylated DNA, is essential for efficient tumorigenesis in the mouse intestine. As Mbd2-deficient mice are viable and fertile, their resistance to intestinal cancer may be of therapeutic relevance. Given the known dependence of mouse intestinal tumorigenesis on DNA methylation1,2, we asked whether deficiency of the methyl–CpG binding repressor Mbd2 (refs. 3,4) could inhibit development of intestinal adenomas in the tumor-prone ApcMin/+ (or Min) mouse5. Mice lacking Mbd2 are viable and fertile, although they are defective in maternal behavior6 and in silencing of certain developmentally regulated genes4. To test whether a reduced ability to interpret the DNA methylation signal might modulate tumorigenesis, we crossed an Mbd2-null mutation6 onto an ApcMin/+ background5 that was also homozygous with respect to the C57Bl/6 Pla2g2a (also called Mom-1) allele7 (Supplementary Note online). We aged cohorts of mice that were homozygous (ApcMin/+ Mbd2–/–), heterozygous (ApcMin/+ Mbd2+/–) or wild-type (ApcMin/+ Mbd2+/+) with respect to Mbd2 (at least 15 mice of each genotype) until they had symptoms of intestinal neoplasia. Mbd2–/– mice survived significantly longer than control Min mice, whereas Mbd2+/– mice survived for an intermediate length of time (Fig. 1). At death, Mbd2–/– mice had ∼10 times fewer adenomas than Mbd2+/+ controls (Fig. 2a). These findings were confirmed when we aged new cohorts (10 mice of each genotype) to a fixed 180-d time point (Fig. 2a). The data indicate a direct proportionality between adenoma burden and dosage of the Mbd2+ allele. The study was carried out using outbred mice (75% C57Bl/6J). We cannot, therefore, rigorously exclude the involvement of a hitherto unknown linked genetic modifier of Min. The few adenomas that arose in ApcMin/+ Mbd2–/– mice were restricted to specific locations, with five mice developing a single small intestinal adenoma at the Brunners gland and all other adenomas occurring within the distal 2 cm of the large intestine. Histological analysis showed no additional aberrant crypts or microadenomas in the intestines of the ApcMin/+ Mbd2–/– mice. The lower total adenoma burden in Mbd2+/– heterozygotes reflects specific repression of tumorigenesis in the small intestine. In Mbd2-null mice, however, both large and small intestinal adenomas were markedly reduced in number (Fig. 2b,c). Adenomas at the 180-d time
point were significantly smaller in Mbd2-deficient mice (Fig. 2d). Furthermore, despite their doubled median lifespan (354 versus 183 days), adenomas from ApcMin/+ Mbd2–/– mice were no larger at death than those arising in ApcMin/+ Mbd2+/+ mice. Our data are therefore consistent with a slower growth rate, as well as a lower incidence, of Mbd2–/– adenomas. Independent of its role as a transcriptional repressor, Mbd2 has been proposed to be a DNA 5-methylcytosine demethylase8, although this claim is controversial3,6,9. As reduced DNA methylation inhibits Min tumorigenesis1,2, loss of a DNA demethylase might be predicted to stimulate tumor formation. Against the Mbd2-demethylase hypothesis, we found that tumorigenesis was greatly suppressed in the Mbd2-null mouse (Figs. 1 and 2). Analysis of the few remaining Mbd2-null tumors, however, showed an atypical frequency of CpG island methylation. Real-time PCR analysis of the genes Itga4 (encoding integrin-α4) and Timp3 (encoding tissue inhibitor of metalloproteinase 3; Supplementary Fig. 1 online), both of which are methylated to varying degrees in adenomatous polyps of Min mice2, showed more frequent methylation in adenomas of mice lacking Mbd2. The CpG island of Cdkn2a (also called p16) was not methylated in wild-type or mutant tumors. Interpretation of the increase in CpG island methylation is complicated by the rarity of Mbd2–/– tumors. For example, it is possible that tumors of this kind arise with a similar frequency in Mbd2+/+ and Mbd2–/– Min mice but make up a negligible fraction of all adenomas in the former. The alternative
1.0 0.9 0.8
ApcMin Mbd2–/–
0.7
Probability
© 2003 Nature Publishing Group http://www.nature.com/naturegenetics
B R I E F C O M M U N I C AT I O N S
0.6 0.5 0.4 0.3
ApcMin Mbd2+/–
0.2 0.1
ApcMin Mbd2+/+
0.0 0
50
100
150
200
250
300
350
d Figure 1 Deficiency of Mbd2 increases lifespan of ApcMin/+ mice in a dosedependent manner. Kaplan–Meier survival plot of ApcMin/+ Mbd2+/+ mice, black line; ApcMin/+ Mbd2+/– mice, red line; ApcMin/+ Mbd2–/– mice, blue line. ApcMin/+ Mbd2–/– mice live significantly longer (median = 354 d) than ApcMin/+ Mbd2+/+ (median = 183 d, P < 0.0001, Log Rank) and ApcMin/+ Mbd2+/– mice (median = 238 d, P = 0.002, Log Rank). ApcMin/+ Mbd2+/– mice also live significantly longer than ApcMin/+ Mbd2+/+ mice (P = 0.002, Log rank).
1Cardiff
School of Biosciences, Cardiff University, Cardiff, Wales, UK. 2Wellcome Trust Centre for Cell Biology, Kings Buildings, Edinburgh University, Edinburgh EH9 3JR, UK. Correspondence should be addressed to A.B. (
[email protected]).
NATURE GENETICS VOLUME 34 | NUMBER 2 | JUNE 2003
145
hypothesis that CpG island methylation is favored in the absence of Mbd2, perhaps due to lower demethylase activity, seems unlikely for several reasons. First, as discussed, enhanced de novo methylation of CpG islands might be expected to increase tumorigenesis, which is not observed. Second, the level of methylation at Itga4 and Timp3 in normal intestine seemed unchanged in Mbd2–/– mice (Supplementary Fig. 1 online). Third, epigenetic silencing of the wild-type Apc allele by promoter hypermethylation, as observed in human colon cancer10, is not detected in residual Mbd2–/– adenomas (Supplementary Fig. 1 online). The Apc+ allele in adenomas of Min mice is normally inactivated by deletion (loss of heterozygosity, LOH; ref. 11), but analysis of large intestinal tumors confirmed a high frequency of LOH in control ApcMin/+ Mbd2+/+ mice (94%, n = 33), ApcMin/+ Mbd2+/– (98%, n = 43) and ApcMin/+ Mbd2–/– mice (86%, n = 21). These frequencies did not differ significantly (P > 0.05), indicating genetic inactivation rather than epigenetic silencing of Apc+.
a
45
*
80
40
70
Adenoma number
Adenoma number
The dual dependencies of Min tumorigenesis on an enzyme that methylates DNA (Dnmt1; refs. 1,2) and also on a protein that interprets the DNA methylation signal to repress transcription (Mbd2) argue that DNA methylation–mediated gene silencing is integral to the Min tumorigenic pathway. Whether the pathway is specific to the Min model or has broader importance (for example, repression of methylated CpG islands generally) is not yet known. Relevance beyond the Min model is suggested by evidence that MBD2 is bound in vivo to the methylated CpG islands of repressed genes for CDKN2A (also called p16Ink4A and p14ARF) in human colorectal carcinoma cells12. MBD2 is also implicated in silencing of the gene encoding glutathione S-transferase in breast cancer cells13. Finally, MBD2 overexpression can occur in breast cancer14, and MBD2 downregulation reportedly inhibits the growth of cultured transformed cells15. Although the effects on tumorigenesis of deficiencies in Mbd2 and Dnmt1 are closely parallel, the consequences for the host
b
90
60 50 40 30
35 30 25
*
20 15
*
10
20 10
Mbd2: +/+
+/–
–/–
*
5
*
0
0
+/+
+/–
Mbd2: +/+
–/–
+/–
–/–
+/+
Small intestine
180 d
+/–
–/–
Large intestine
Death
c
d 70 60
50
50
Adenoma weight (mg)
Adenoma number
© 2003 Nature Publishing Group http://www.nature.com/naturegenetics
B R I E F C O M M U N I C AT I O N S
*
40
** **
30 20
*
10 0
*
40
*
30 20
*
10 0
Mbd2: +/+
+/–
Small intestine
–/–
+/+
+/–
–/–
Large intestine
Mbd2: +/+
–/– 180 d
+/+
–/– Death
Figure 2 Tumors of Mbd2–/– mice have lower incidence, are smaller in size and have an altered intestinal distribution. (a) Box plots show numbers of adenomas per mouse at 180 d and at death. The horizontal boxed line represents the median. ApcMin/+ Mbd2–/– mice have significantly fewer tumors at 180 d and at death than do ApcMin/+ Mbd2+/+ and ApcMin/+ Mbd2+/– mice (P < 0.0002). ApcMin/+ Mbd2+/– mice have significantly fewer tumors than do ApcMin/+ Mbd2+/– mice at 180 d (P = 0.019) but not at death (P = 0.15). (b) ApcMin/+ Mbd2–/– mice have fewer tumors in both the small and large intestine at 180 d than do ApcMin/+ Mbd2+/– mice and ApcMin/+ Mbd2+/+ mice (P < 0.01). ApcMin/+ Mbd2+/– mice have significantly fewer tumors in the small intestine (P = 0.0007), but not in the large intestine (P = 0.22), than do ApcMin/+ Mbd2+/+ mice at 180 d. (c) ApcMin/+ Mbd2–/– mice have significantly fewer tumors in both the small and large intestine at death than do ApcMin/+ Mbd2+/+ mice and ApcMin/+ Mbd2+/– mice (P < 0.001). No differences were observed between ApcMin/+ Mbd2+/+ mice and ApcMin/+ Mbd2+/– mice (P > 0.1) at death. (d) Wet weight of tumors at 180 d and at death. ApcMin/+ Mbd2–/– tumors (n = 8) weighed less at 180 d than did tumors from ApcMin/+ Mbd2+/+ (n = 14) mice (P = 0.004) but were not significantly smaller at death (P = 0.4; n = 14 for ApcMin/+ Mbd2+/+; n = 9 for ApcMin/+ Mbd2–/–). Significance was estimated in all cases using the Mann–Whitney test.
146
VOLUME 34 | NUMBER 2 | JUNE 2003 NATURE GENETICS
B R I E F C O M M U N I C AT I O N S organism are very different. Absence of Dnmt1 causes embryonic lethality1, whereas Mbd2-null mice appear healthy6. If MBD2 were ‘dispensable’ in humans also, then it would become an attractive potential target for therapeutic intervention in cancer. Note: Supplementary information is available on the Nature Genetics website.
© 2003 Nature Publishing Group http://www.nature.com/naturegenetics
ACKNOWLEDGMENTS We thank K. Wilson, N. Hill and D. Scarborough for technical assistance. This work was funded by the Wellcome Trust and Cancer Research UK. J.B. is supported by the European Union. A.R.C. is a Royal Society Research Fellow. COMPETING INTERESTS STATEMENT The authors declare that they have no competing financial interests. Received 14 February; accepted 28 March 2003 Published online 5 May 2003; doi:10.1038/ng1155
NATURE GENETICS VOLUME 34 | NUMBER 2 | JUNE 2003
1. 2. 3. 4. 5. 6.
Laird, P.W. et al. Cell 81, 197–205 (1995). Eads, C.A., Nickel, A.E. & Laird, P.W. Cancer Res. 62, 1296–1299 (2002). Ng, H.-H. et al. Nat. Genet. 23, 58–61 (1999). Hutchins, A. et al. Mol. Cell 10, 81–91 (2002). Su, L.K. et al. Science 256, 668–670 (1992). Hendrich, B., Guy, J., Ramsahoye, B., Wilson, V.A. & Bird, A. Genes Dev. 15, 710–723 (2001). 7. Dietrich, W.F. et al. Cell 75, 631–639 (1993). 8. Bhattacharya, S.K., Ramchandani, S., Cervoni, N. & Szyf, M. Nature 397, 579–583 (1999). 9. Santos, F., Hendrich, B., Reik, W. & Dean, W. Dev. Biol. 241, 172–182 (2002). 10. Esteller, M. et al. Cancer Res. 60, 4366–4371 (2000). 11. Luongo, C., Moser, A.R., Gledhill, S. & Dove, W.F. Cancer Res. 54, 5947–5952 (1994). 12. Magdinier, F. & Wolffe, A.P. Proc. Natl. Acad. Sci. USA 98, 4990–4995 (2001). 13. Lin, X. & Nelson, W.G. Cancer Res. 63, 498–504 (2003). 14. Billard, L.M., Magdinier, F., Lenoir, G.M., Frappart, L. & Dante, R. Oncogene 21, 2704–2712 (2002). 15. Slack, A. et al. J. Gene Med. 4, 381–389 (2002).
147
