Full text of DOI:10.1093/gerona/56.6.B268

Journal of Gerontology: BIOLOGICAL SCIENCES
Copyright 2001 by The Gerontological Society of America
2001, Vol. 56A, No. 6, B268–B276
Unexpected Effects of a Heterozygous Dnmt1 Null
Mutation on Age-Dependent DNA Hypomethylation
and Autoimmunity
Raymond Yung,1 Donna Ray,1 Julie K. Eisenbraun,1 Chun Deng,1 John Attwood,1
Michael D. Eisenbraun,2 Kent Johnson,3 Richard A. Miller,3 Samir Hanash,4 and Bruce Richardson1,5
Departments of 1Internal Medicine, 2Cell and Molecular Biology, 3Pathology, and 4Pediatrics and Communicable Diseases,
University of Michigan, Ann Arbor.
5Ann Arbor Veterans Affairs Hospital, Michigan.
DNA methylation modifies gene expression. Methylation patterns are established during ontog-
eny, but they change with aging, usually with a net decrease in methylation. The significance of
this change in T cells is unknown, but it could contribute to autoimmunity, senescence, or both.
We examined the effects of a null mutation in DNA methyltransferase 1 (Dnmt1), a gene main-
taining DNA methylation patterns, on immune aging. Whereas aged control mice developed hy-
pomethylated DNA, autoimmunity, and signs of immune senescence as predicted, the knockout
mice surprisingly increased DNA methylation and developed signs of autoimmunity and senes-
cence more slowly. To identify potential mechanisms, we compared transcripts of DNA methyl-
transferase and methylcytosine binding protein family members in control and knockout mice.
MeCP2, a methylcytosine binding protein involved in gene suppression and chromatin inactiva-
tion, was the only transcript differentially expressed between old knockout mice and controls,
and thus it is a candidate for a gene product mediating these effects.
NA methylation modifies gene expression by targeting
methylcytosine binding proteins to specific promoter
tion may also contribute to the development of autoimmu-
nity in aged mice.
D
sequences, which suppresses gene expression directly, or
indirectly by promoting chromatin inactivation through
transcriptional repression domains (1). DNA methylation
patterns are established during development in a tissue-spe-
cific fashion (2), and they serve to suppress the expression
of genes not essential or potentially detrimental to cellular
function (3,4). Our group has examined the role DNA meth-
ylation plays in regulating T-lymphocyte function and gene
expression. We reported that inhibiting T-cell DNA meth-
ylation with DNA methyltransferase (DNA MTase) inhibi-
tors such as 5-azacytidine modifies gene expression and in-
duces autoreactivity, and that T cells made autoreactive by
this mechanism induce an autoimmune disease in vivo. Ab-
normalities resulting from the autoreactive cells include
anti-DNA antibodies, immune complex renal disease, lym-
phocytic interstitial pneumonitis, and an autoimmune liver
disease resembling primary biliary cirrhosis (5,6). These
observations suggest that T-cell DNA hypomethylation can
contribute to autoimmunity. In other studies, our group and
others have reported that T-cell DNA demethylates with age
(7,8), similar to many but not all tissues (9). Significantly,
aging mice also develop anti-DNA antibodies and T-cell-
dependent autoimmune lesions in several organs, including
the lung and liver (10,11), which resemble those induced by
experimentally demethylated T cells. This raises the possi-
bility that age-dependent changes in T-cell DNA methyla-
T-lymphocyte function also declines with age. These
changes include decreased interleukin-2 (IL-2) production
and the accumulation of “memory” T cells (12). Because
DNA methylation patterns change with aging, altered DNA
methylation is an attractive explanation for age-dependent
changes in cellular function. However, despite evidence for
age-dependent changes in DNA methylation, there is little
information available regarding the functional significance
of the changes, or their relevance to immune senescence.
The inactivation of methylation genes in mouse models
permits a correlative assessment of questions regarding the
relationship between DNA hypomethylation and autoimmu-
nity, as well as the functional significance of age-dependent
changes in DNA methylation. A heterozygous DNA meth-
yltransferase 1 (Dnmt1) “knockout” mouse has been described
(13). Dnmt1 is an enzyme responsible for maintaining DNA
methylation patterns through mitosis (14). Homozygosity
for the null allele is lethal to the embryo, whereas the het-
erozygous knockout mouse is phenotypically normal but
has hypomethylated DNA (13). We hypothesized that the
loss of one Dnmt1 allele might increase the rate of the age-
dependent decrease in T-cell deoxymethylcytosine (dmC)
content, resulting in an increase in the rate of onset of age-
dependent autoimmunity. We also hypothesized that age-
dependent changes in T-cell function and gene expression
would be increased in the Dnmt1-deficient mice, as a result
B268

EFFECTS OF A Dnmt1 NULL MUTATION ON AGING
B269
of an effect of the Dnmt1 mutation on methylation changes
with aging.
washed three to four times with phosphate-buffered saline
(PBS), and 105 cells were added to each well in 200 ꢂl of
Roswell Park Memorial Institute 1640 (RPMI) supplemented
with 10% fetal calf serum (FCS) and 50 ꢂM of 2-mercapto-
ethanol. Where indicated, anti-CD28 (Pharmingen, 2.25 ꢂg/
ml) was also added; 150 ꢂl of media was removed at 48
hours, saved for enzyme-linked immunosorbent assays
(ELISAs), and replaced with an equal volume of fresh me-
dia. At 54 hours, 1 ꢂCi of tritiated thymidine (3H-TdR) was
added, and the cells were harvested 2 hours later and prolif-
To test these hypotheses, we compared the development
of histologic and serologic signs of autoimmunity in young,
middle aged, and old heterozygous Dnmt1 knockout mice
with their wild-type littermates. Selected parameters of
T-cell function and gene expression known to change with
aging were also compared. Additional studies included a
comparison of total genomic dmC content and expression of
DNA MTase and methylcytosine binding protein transcripts
between groups. The results support the association be-
tween age-dependent, progressive DNA hypomethylation
and the development of autoimmunity. Surprisingly, how-
ever, the Dnmt1 knockout mice increased genomic dmC
content with aging, and they showed a decline in the devel-
opment of several age-dependent immune abnormalities. A
search for possible secondary effects of the knockout muta-
tion showed that these mice maintained MeCP2 expression
with aging, which may maintain patterns of DNA methyla-
tion and gene expression with aging. (MeCP2 is a methylcy-
tosine binding protein involved in gene suppression and
chromatin inactivation.)
3
eration measured by H-TdR incorporation. IL-2, IL-4, IL-
10 and interferon gamma (IFN-ꢄ) in culture supernatants
were detected by two-site ELISA, using matched pair anti-
bodies (Pharmingen) according to the manufacturer’s in-
structions. Levels were determined graphically by using a
standard curve generated with recombinant murine cyto-
kines (Genzyme, Cambridge, MA).
Flow Cytometric Analysis
Phenotypic characterization was performed by using anti-
CD4-fluorescein isothiocyanate (FITC), anti CD8-FITC,
anti-CD11a-FITC, and anti-CD44-phycoerythrin (PE) (all
from Pharmingen) as described by Quddus and colleagues
(5). Controls included MSIg-PE and mouse serum immuno-
globulin (MSIg)-FITC. All samples were analyzed by using
a Coulter ELITE flow cytometer.
MATERIALS AND METHODS
Mice
Heterozygous female Dnmt1 knockout mice, bred onto a
C57BL/6 background, and their wild-type littermates were
obtained from Jackson Labs (Bar Harbor, ME) and housed
in a specific pathogen-free environment in the University of
Michigan Unit for Laboratory Animal Medicine. Sentinel
mice from this colony were tested every 3 months for anti-
bodies to murine viral pathogens, and all tests were negative
during the experimental period. At the time of the study, the
groups were aged approximately 6, 11, and 18 months, with
a variability of ꢀ1 month.
Autoantibodies
Sera were diluted 1:200 with PBS, and then they were
added to Immulon 4 flat-bottomed microtiter wells (Dynex
Technologies, VA) coated with ssDNA or dsDNA as de-
scribed (5,6). The plates were then washed and bound im-
munoglobulin (Ig) detected with horse radish peroxidase
(HPO)-conjugated polyvalent goat antimouse IgM or IgG
(Sigma, St. Louis, MO) as described (5,6). All determina-
tions were performed in quadruplicate and repeated at least
once. A reference standard consisting of pooled sera from
mice with lupus (17) was included on each plate, and results
are expressed relative to this standard. Total IgG and IgM
were also measured by ELISA, using previously published
protocols (5,6).
Histologic Analysis
Tissues were fixed, embedded in paraffin, sectioned, and
stained with hematoxylin and eosin, using previously pub-
lished protocols (5,6). Slides were read by a pathologist
who was blinded to the groups being studied, and they were
scored on a 0–4ꢁ scale. Activity of glomerulonephritis was
scored according to published criteria (15).
Quantitation of Genomic Deoxymethylcytosine Content
The dmC content was compared between groups by using
SssI methylase assays (18). Briefly, 500 ng of genomic
DNA from CD4-enriched T cells was incubated 37ꢃC for 4
hours with 4 U of Sss1 CpG methylase (New England Bio-
Labs, Beverly, MA), 1.5 ꢂM of S-adenosyl-L-[methy-3H]
methionine (3H-SAM) (77 Ci/mmol; Amersham, Arlington
Heights, IL), and 1.5 ꢂM of nonradioactive S-adenosylme-
thionine (New England BioLabs) in buffer provided by the
manufacturer. Five ꢂl of 2.5-mM nonradioactive SAM was
then added and the reaction mixture was spotted on What-
man GF/C 2.4-cm2 filter disks, which were air dried for 15
minutes. The filters were washed with 6 ml of 5% TCA fol-
lowed by 6 ml of 70% ethanol, and they were air dried over-
Isolation of Lymphocytes and Measurement of
Lymphocyte Function
Spleens were removed from mice and splenocytes recov-
ered as described (5). Where indicated, CD4ꢁ T cells were
enriched by two cycles of treatment with anti-CD8 plus
anti-Iab (Pharmingen, San Diego, CA) and rabbit comple-
ment (Pel-Freeze, Brown Deer, WI) and previously pub-
lished protocols (5). This typically resulted in a twofold to
threefold increase in the percentage of CD4ꢁ cells. Prolif-
erative and IL-2 responses were measured by using flat-bot-
tomed microtiter wells (Costar, from Corning; Corning,
NY) and protocols described by others (16). Briefly, 50 ꢂl
of anti-CD3 (10 ꢂg/ml in phosphate-buffered saline; 2C11,
from Pharmingen) were added to the microtiter wells, and
allowed to bind overnight at 4ꢃC. The wells were then
3
night; H incorporation was measured by using a scintilla-
tion spectrometer. All determinations were performed in
triplicate, and results were normalized to the young control

B270
YUNG ET AL.
group, arbitrarily defined as 100%. Similar studies were
performed on brain tissue from old mice.
MeCP2: 5ꢅ-CAA CCT TCA GCC CAC CAT TC;
5ꢅ-TCT CCA GGA CCC TTT TCA CC.
Statistical Analysis
The difference between means was tested by using the
unpaired Student’s t test with pooled variance, calculated
with Systat and Sigmastat software (SPSS Science, Chi-
cago, IL). Linear regression and analysis of variance was
similarly calculated with Systat and Sigmastat software.
Differences between grouped assays were tested by using
Fisher’s Method of Combining Independent Tests (19).
Fisher’s Exact Test was used to test the significance of his-
tologic data.
RT-PCR Amplification of DNA Methyltransferase and
Methylcytosine Binding Protein Families
Tissue stored in liquid nitrogen was rapidly thawed,
placed in Trizol (GIBCO-BRL, Gaithersburg, MD) and ho-
mogenized. Total RNA was purified according to the manu-
facturer’s protocol. Real-time reverse transcription poly-
merase chain reaction (RT-PCR) was performed with a
LightCycler (Roche, Indianapolis, IN). With the use of a
One-Step RNA Amplification Kit with SYBR Green I
(Roche), 200 ng of total RNA was converted to cDNA and
then amplified in one step. Conditions for the reactions were
as follows: reverse transcription at 55ꢃC for 10 minutes; de-
naturation at 95ꢃC for 30 seconds; amplification at 95ꢃC for
1 second, 56ꢃC for 10 seconds, and 72ꢃC for a period deter-
mined by the product length (1 second per 25 nucleotides)
for a total of 45 cycles. At the completion of amplification,
melting characteristics of the product were determined. In
each experiment, water was included as a negative control
to rule out primer dimer formation. A series of five dilutions
of one RNA sample were also included to generate a standard
curve, and this was used to obtain relative concentrations of
the transcript of interest in each of the RNA samples. A
standard curve was generated during each experiment by
using one of the primer pairs. Relative concentrations were
then determined by the second derivative method with the
LightCycler computer software. Amplification of GAPDH
was performed to confirm that equal amounts of total RNA
were added for each sample and that the RNA was intact
and equally amplifiable among all samples.
RESULTS
Development of Autoimmunity
The effect of the Dnmt1 null mutation on the long-term
health of mice is unknown, so weights and overall health
were compared among young, middle aged, and old knock-
out mice and controls. No differences in weights, general
health, or mortality were observed among groups during the
duration of this study. Complete autopsies were performed
on 6 mice from each of the 6 groups (young, middle aged,
and old; knockout and control), and no evidence for malig-
nancy was identified. No splenomegaly was identified in
any of the groups, and total numbers of splenocytes were
similar between the control and old knockout mice (6.92 ꢀ
0.48 ꢆ 107 vs 6.54 ꢀ 0.98 ꢆ 107, mean ꢀ SEM of six mice/
group, control vs knockout). This suggests that the Dnmt1
mutation has no major effect on the health of the mice at
these ages.
The development of autoimmunity in the Dnmt1 knock-
out mice and their wild-type littermates was compared. At
later ages, C57BL/6 mice develop lymphocytic infiltrates of
the salivary glands and liver, as well as of other organs (10),
so histologic evidence for these lesions was sought in
young, middle aged, and old Dnmt1 knockout and control
mice. The most consistent abnormalities detected were
mononuclear infiltrates of the liver and salivary glands. Fig-
ures 1A and 1B show representative histology of these or-
gans from affected mice. By 18 months of age, a majority of
mice from both groups had evidence for autoimmune sali-
vary gland lesions. These results compare favorably with
those reported by others (10). However, in the 11-month-
old age group, significantly (p ꢇ .02) more of the control
mice had moderate to severe lesions compared with the
knockout mice (Table 1). These lesions are typically ꢀ86%
T cells and 7% B cells in aged C57BL/6 mice (10), although
the composition of the lesions was not determined in the
Dnmt1 knockout mice. The liver lesions were observed in 5
of 6 old control mice, whereas only 1 of 6 old knockout
mice had the mononuclear cell infiltrates. Four of the six
aged control mice had evidence for glomerulonephritis and
interstitial inflammation (average activity score 3.8 ꢀ 1.5,
mean ꢀ SEM), whereas 2 of the 6 aged knockout mice had
similar lesions (activity 1.7 ꢀ 1.1). Three of six of the old
control mice had mild to moderate lymphocytic infiltration
of the lungs; none of six old knockout mice did. It is note-
worthy that in all organs, the controls had greater evidence
Primer sequences were as follows, listed as forward; re-
verse:
GAPDH: 5ꢅ-TTG AAG GGT GGA GCC AAA CG;
5ꢅ-TGG GAG TTG CTG TTG AAG TCG
Dnmt1: 5ꢅ-AGG ATG AGA GGG AGG AGA AGA
GAC; 5ꢅ-GGT GGA GTC GTA GAT GGA CAG
TTT C
Dnmt3a: 5ꢅ-CAC CAG CCA AGA AAC CCA GAA
AG; 5ꢅ-TCC CAT CAA AGA GAG ACA GCA CG
Dnmt3b: 5ꢅ-CAA ACC CAA CAA GAA GCA ACG
AG; 5ꢅ-CCA GAC ACT CCA CAC AGA AGC ATC
Mbd1: 5ꢅ-CGT TTG GAC GCT CAG ACA TC;
5ꢅ-TCC TCT TTC ACC AGG CAA GC
Mbd2: 5ꢅ-CAC GAA CCA CCC GAG CAA TAA G;
5ꢅ-TCA GTG CCT CCT CCA GTT TCT TG
Mbd3: 5ꢅ-ACC CCA GCA ACA AGG TCA AG;
5ꢅ-TTT GGA AAG GAG GTG GAC G
Mbd4: 5ꢅ-ACT CGG CTT GCT TGC TAT TG;
5ꢅ-TGA AGG ATT TGC GTC TTG GG

EFFECTS OF A Dnmt1 NULL MUTATION ON AGING
B271
knockout and control mice, and they were tested for IgM
and IgG anti-ssDNA and anti-dsDNA antibodies (Figure 2).
No difference in titer was observed in the young and middle
aged mice. However, in all cases, sera from the aged control
mice had greater titers of anti-ssDNA (p ꢈ .05) and anti-
dsDNA (p ꢈ .025) antibodies relative to their Dnmt1-defi-
cient littermates (p ꢈ .005 overall). Total IgM and IgG
were also measured in the aged groups, to exclude the possi-
bility that an increase in autoantibodies reflected a poly-
clonal increase in immunoglobulin. ELISA assays demon-
strated no significant differences in total IgM (OD 0.377 ꢀ
0.007 vs 0.363 ꢀ 0.004, old knockouts vs old controls,
mean ꢀ SEM, using a 1:100 serum dilution) or IgG (OD
0.456 ꢀ 0.005 vs 0.447 ꢀ 0.009). No significant differ-
ences in total IgG or IgM levels were observed in the young
or middle aged groups as well (not shown). Together, these
data indicate that the wild-type mice develop anti-DNA an-
tibodies and tissue evidence for autoimmunity more rapidly
than the Dnmt1 knockout mice.
Immune Characterization
The observation that autoimmunity developed more
slowly in the Dnmt1-deficient mice raised the possibility
that the knockout mice had impaired T-cell function, thus
preventing the development of autoimmunity. It was also
possible that the Dnmt1 null mutation modified the devel-
opment of immune senescence. As a way to test these possi-
bilities, T-cell subsets, proliferative responses, and cytokine
secretion, all reported to change with aging (12), were com-
Figure 1. Histologic characterization of autoimmunity. The livers
and salivary glands were removed from Dnmt1-deficient or wild-type
littermates at approximately 6, 11, and 18 months of age. Tissues
were fixed and sectioned, and then stained with hematoxylin and
eosin. A shows a representative section from the liver, and B shows
one from the salivary glands of 18-month-old control mice (400ꢆ).
for autoimmunity compared with their knockout littermates
(p ꢈ .01 overall by Fisher’s method for combining indepen-
dent tests). Overall the incidence of these lesions is some-
what higher than those reported previously (10). However,
in the present study, tissues were counted as positive if lym-
phocyte aggregates were observed in the sections, whereas
in the previous report (10), tissues were only scored as posi-
tive if more than 50 mononuclear cells were observed in the
center of the tissue at 100ꢆ magnification.
Serologic evidence for autoimmunity was also sought.
Sera were obtained from young, middle aged, and old
Figure 2. Anti-DNA antibodies in Dnmt1-deficient and control
mice. Sera were obtained from Dnmt1-deficient (knockout, or KO;
ꢀ) or wild-type littermates (control; ꢁ) at the indicated ages, and they
were tested for the following autoantibodies: A, IgM anti-ssDNA; B,
IgM anti-dsDNA; C, IgG anti-ssDNA; and D, IgG anti-dsDNA. Re-
sults represent the mean of quadruplicate determinations performed
on 10 mice/group and are expressed relative to a standard consisting
of sera from mice with lupus induced with a DNA methylation inhib-
itor. Standard errors averaged 15.6% of the means.
Table 1. Salivary Gland Inflammation: Middle-Aged Mice
Inflammation*
Mouse Strain
Negative
Mild
Moderate–Severe
Control
Knockout
1
4
1
3
4
0
*p ꢇ .02, negative ꢁ mild vs moderate ꢁ severe.

B272
YUNG ET AL.
pared between groups. Flow cytometric analysis was used
to compare the percentage of CD4ꢁ and CD8ꢁ T cells in
the spleen in the young, middle aged, and old knockouts and
controls. No significant differences were observed at any
age (not shown). Others have reported that “memory”
CD4ꢁ T cells, expressing high levels of CD44, accumulate
with age (12), so CD44 expression was compared between
groups (Figure 3). Overall, the number of CD4ꢁ memory
cells increased with age in both groups (p ꢇ .001 by analy-
sis of variance, or ANOVA). However, T cells from the
knockout mice tended to have fewer CD4ꢁ, CD44 high T
cells than controls (p ꢇ .031 overall by ANOVA), which
reached statistical significance in the old group (p ꢇ .003
by t test), but not at the younger ages, suggesting a more
pronounced difference in the older mice.
Proliferative responses were tested by stimulating unfrac-
tionated splenocytes and CD4-enriched splenocytes from
young and old Dnmt1 knockout mice and controls with im-
mobilized anti-CD3 with and without anti-CD28 (Figure 4).
No differences were seen between the young groups with
either form of stimulation. The control group showed essen-
tially no change in response to either stimulus with aging.
Although murine T-cell proliferative responses usually de-
cline with age by 18 months of age (12), the decline appears
to be delayed when immobilized anti-CD3 is used as the
stimulus (16), perhaps reflecting a relatively stronger stimulus
from immobilized anti-CD3. In contrast, the CD4-enriched
population from the knockout mice demonstrated an in-
creased proliferative response with age to anti-CD3 with
and without anti-CD28 (p ꢈ .05, young vs old). These re-
sponses were significantly greater than those in the age-
matched controls (p ꢈ .005 overall).
Figure 4. T-cell proliferative responses in Dnmt1-deficient and
control mice. A, Unfractionated splenocytes (UnFx) or CD4-enriched
splenocytes (CD4ꢁ) from young (Y) and old (O) knockout (KO,
open bars) and control (hatched bars) mice were stimulated with im-
mobilized anti-CD3. B, Similar preparations were stimulated with im-
mobilized anti-CD3 and soluble anti-CD28. In both panels prolifera-
tion was measured 54 hours later by H-TdR incorporation. Results
represent the mean ꢀ SEM of 5–7 young and 5–7 old mice. CPM ꢇ
counts per minute.
3
and supernatants harvested 48 hours later. T cells from the
control mice showed an age-dependent decline in IL-2 pro-
duction (Figure 5) as described by others (12; p ꢈ .005
young vs old), whereas no significant decrease was ob-
served in the knockouts. The difference in the slopes (young
to old) was significant (ꢉ0.480 ꢀ 0.099 vs ꢉ0.029 ꢀ
Cytokine responses were also compared by using CD4-
enriched populations from the young, middle aged, and old
knockouts and controls. The cells were stimulated with im-
mobilized anti-CD3 and soluble anti-CD28 as in Figure 4,
Figure 5. IL-2 secretion by CD4-enriched T cells from Dnmt1-defi-
cient and control mice. T cells from knockout (KO; ꢀ) or control
(control; ꢁ) mice were stimulated with anti-CD3 ꢁ anti-CD28 as in
Figure 3, and 48 hours later supernatants were harvested and tested
for IL-2 by ennzyme-linked immunosorbent assay. Results represent
the mean ꢀ SEM of 6–7 young, 3–4 middle aged, and 7 old mice per
group.
Figure 3. Development of the CD4ꢁ memory subset in Dnmt1-
deficient and control mice. Splenocytes from Dnmt1-deficient
(knockout, or KO; ꢀ) and control (control; ꢁ) mice were stained
with anti-CD4-FITC and anti-CD44-PE and then analyzed by flow
cytometry. Results represent the mean ꢀ SEM of 8–12 mice/point.

EFFECTS OF A Dnmt1 NULL MUTATION ON AGING
B273
0.129, control vs knockout, p ꢈ .02). The amount of IL-2
secreted by T cells from control and Dnmt1 knockout mice
was also significantly different in the old group (p ꢇ .041).
Although there was a trend for the knockout mice to secrete
less IL-2 in the younger age groups, this did not achieve sta-
tistical significance. No significant differences were ob-
served in IL-4, IL-10, or IFN-ꢄ production between T cells
from both young and old groups (not shown). These results
argue against impaired T-cell responsiveness as a mecha-
nism for the protection of autoimmunity developing in the
knockout mice. In addition, the observation that the knock-
out mice develop signs of autoimmunity more slowly, de-
velop fewer “memory” cells, and do not lose IL-2 secretory
capacity by 18 months of age suggests that some aspects of
the immune system may age more slowly in the knockout
mice.
published reports (13,22). Unexpectedly, however, total
dmC content increased with age in the knockout mice, so
that by 18 months the dmC content was significantly greater
(p ꢈ .01) than in age-matched littermates, and approxi-
mated that of the young wild-type mice. Similar results
were observed in the brains of the aged mice (p ꢈ .02, Fig-
ure 6B), suggesting the existence of compensatory mecha-
nisms for Dnmt1 deficiency.
Expression of DNA Methyltransferase and
Methylcytosine Binding Protein Familes
Because brain tissue from the aged knockout mice dem-
onstrated greater amounts of dmC than did age-matched
controls, similar to lymphocytes, levels of DNA MTase and
methylcytosine binding protein transcripts were compared
by real-time RT-PCR in brains from young and old knock-
out and control mice, using 3 mice/group. There was a small
and statistically insignificant decrease in Dnmt1 mRNA in
brains from young knockout mice relative to controls (96 ꢀ
5 vs 132 ꢀ 16, mean ꢀ SEM, knockout vs control), and
similarly no significant difference in the brains from old
mice (106 ꢀ 25 vs 112 ꢀ 27). Similar results were found
for Dnmt3b, whereas Dnmt3a mRNA was barely detectable
in all groups, requiring ꢀ30 cycles of amplification for de-
tection. Methylcytosine binding proteins have been impli-
cated in modifying chromatin structure and gene expression
(1,23), so expression of MBD1–4 and MeCP2 transcripts
were compared. No significant differences in MBD1,
MBD2, or MBD4 mRNAs were found between groups as
well. MBD3 mRNA demonstrated a significant decrease
with aging in the control (p ꢇ .03) and knockout (p ꢇ .008)
groups (Figure 7A). However, because the change was
identical in both groups, it is unlikely that this gene product
contributes to the phenotypic differences between the con-
trols and knockout mice. In contrast, although MeCP2 de-
creased with age in the control mice (p ꢇ .001), no decrease
was observed in the knockout mice, and the difference in
MeCP2 levels between old control and knockout mice was
significant (Figure 7B; p ꢇ .03). GAPDH was run as a con-
trol and showed no significant differences between groups.
DNA Methylation Analysis
The effect of age on genomic dmC content was compared
in the knockout mice and controls. DNA from the CD4-
enriched splenocytes was incubated with 3H-SAM and SssI
3
methylase, and then it was precipitated and washed; H in-
corporation was compared between groups. SssI will meth-
3
ylate all unmethylated CG pairs, so H incorporation is
3
inversely proportional to dmC content, with greater H in-
corporation reflecting a lower dmC content. Figure 6A
shows that lymphocyte DNA from control mice demeth-
ylates in an age-dependent fashion as described by others
(8,20,21). In addition, DNA from young Dnmt1 knockout
mice is relatively hypomethylated compared with that of
age-matched littermates (p ꢈ .01), which is consistent with
DISCUSSION
These experiments were designed to determine the ef-
fects of heterozygous Dnmt1 deficiency on lymphocyte
DNA methylation with aging, and to correlate the results
with the development of autoimmunity and immune senes-
cence. Four significant observations were made. First,
whereas lymphocyte DNA from the young Dnmt1 knockout
mice was hypomethylated relative to age-matched controls,
total dmC content increased with age in the knockouts. This
is in contrast to the wild-type littermates, which demon-
strated the expected age-dependent decrease. Second, the
onset of age-dependent autoimmunity was delayed in the
knockout mice compared with controls. Third, characteris-
tics of T-cell aging, in particular development of the mem-
ory subset and the rate of decrease in IL-2 secretion, showed
altered developmental patterns in the knockout mice.
Fourth, levels of MBD3 mRNA decreased with age in both
control and knockout mice, whereas MeCP2 mRNA de-
creased only in the control mice.
Figure 6. Genomic DNA methylation in CD4-enriched T cells
from Dnmt1-deficient and control mice. A, DNA from lymphocytes
was isolated from knockout mice (—) or controls (- - -), and it was la-
3
beled with H-SAM, using SssI methylase. The DNA was then pre-
cipitated and washed, and 3H was incorporation measured. All deter-
minations were performed in triplicate. The results represent the
mean ꢀ SEM of determinations performed on 4–5 mice/data point,
normalized to young control mice. B, Similar studies were performed
on DNA from brains of old knockout mice and controls. Results rep-
resent the mean ꢀ SEM of [3H]-CH3 incorporation into the DNA
from 3 mice/group.

B274
YUNG ET AL.
with pharmacologic inhibitors such as 5-azacytidine or
procainamide results in a compensatory increase in Dnmt1
mRNA expression and increased DNA MTase enzyme ac-
tivity (18,25). This is supported by our observation that
brain Dnmt1 mRNA levels were not significantly different
between groups. It is thus possible that similar regulatory
mechanisms lead to overexpression of the remaining Dnmt1
allele. Two additional DNA MTases with de novo methyla-
tion capabilities, Dnmt3a and Dnmt3b, have recently been
described (26), but minimal Dnmt3a mRNA was detectable
in the brains of the mice, and no significant differences in
Dnmt 3b mRNA were observed. However, as gene expres-
sion at the protein level was not measured, it is possible that
differential expression in the activity of one or more of
these enzymes might be responsible. It is also possible that a
relative increase in a demethylase in the control mice could
account for progressive hypomethylation in this group. A
recent report suggests that methylcytosine binding domain 2
(MBD2) might have demethylase activity, although others
have been unable to confirm this result (27–29). It is thus
relevant to note that there were no differences in MBD2
transcripts between groups. Others have described a ribonu-
cleoprotein complex with glycosidase activity in chickens,
which results in DNA demethylation (30). The differential
expression of a similar enzyme in mammals might account
for the decrease in dmC content observed in the control
mice. Future experiments will have to compare expression
of these enzymes at the protein level in various tissues from
young, middle aged, and old Dnmt1 knockout mice and
controls under both quiescent and mitogen-stimulated con-
ditions, to test these possibilities.
Comparison of transcripts of the methylcytosine binding
protein family demonstrated that MBD3 decreased with age
in both controls and knockouts, and thus is unlikely to con-
tribute to the differences in age-dependent DNA methyla-
tion between the groups. However, MeCP2 transcripts de-
creased in the controls but not the knockouts. This protein
binds methylcytosine and contains a transcriptional repres-
sion domain that interacts with a transcriptional corepressor
complex containing histone deacetylases (31). MeCP2 can
also suppress gene expression through mechanisms inde-
pendent of histone deacetylases (32). A decrease in this pro-
tein could potentially lead to abnormal expression of genes
repressed by both mechanisms, contributing to the aging
process. It is also possible that persistance of its expression
may contribute to an increase in dmC with age in the knock-
out mice by promoting chromatin inactivation. This hypoth-
esis is indirectly supported by the observation that the
knockout mice are phenotypically unchanged with aging
despite increasing dmC levels, consistent with progressive
methylation and inactivation of transcriptionally silent re-
gions. However, as noted above, a number of other mecha-
nisms could also account for the difference in DNA meth-
ylation between the two groups. Future studies will be
required to clarify the significance of these changes in these
transcripts. Such studies would include confirmation of dif-
ferences at the protein level, effects of overexpression and
underexpression of the relevant proteins on DNA methyla-
tion, and studies on how the Dnmt1 mutation leads to pres-
ervation of its expression.
Figure 7. Effect of age on MBD3 and MeCP2 expression in brains
from Dnmt1-deficient and control mice. A, MBD3 mRNA was quan-
titated as described in Materials and Methods from brains of young
(solid bars) and old (hatched bars) control or knockout mice. Results
represent the mean ꢀ SEM of 3 mice/group. B, MeCP2 mRNA was
quantitated from the same groups as described in A.
The relative hypomethylation of lymphocytes from the
young knockout mice was not unexpected, because others
have reported similar findings (22). However, the observa-
tion that levels of Dnmt1 transcripts in the knockouts were
identical to controls was surprising. Because Dnmt1 levels
are decreased in the Dnmt1 knockout embryonal stem cells
(13), it is possible that a decrease in Dnmt1 exists during de-
velopment, resulting in hypomethylation that is maintained
at least through 6 months of age despite higher levels of
Dnmt1 transcripts. This is consistent with the observation
that the increase in methylation is delayed and not observed
until 11 months of age.
The changes in genomic dmC content observed in both
groups could potentially affect gene expression. However,
phenotypic changes occurred largely in the control group.
Therefore, it seems reasonable to propose that the changes
in the control group occur in transcriptionally relevant re-
gions of the genome. Other work from our group demon-
strates that in mice, age-dependent methylation changes oc-
cur outside of CpG islands (manuscript in preparation),
suggesting the changes in methylation occur in promoters
lacking islands. Slightly less than half of murine genes lack
CpG islands (24), so the changes could potentially affect a
large number of genes. In contrast, the paucity of functional
changes in the knockout group suggest that the increases in
dmC may occur largely outside of regulatory regions.
The age-dependent increase in DNA methylation ob-
served in the knockout mice could be viewed as paradoxi-
cal, because expression of only one Dnmt1 allele would be
expected to lead to a more pronounced decrease in DNA
methylation with age. However, the increase in DNA meth-
ylation observed in the knockout mice may represent a com-
pensatory mechanism, for which there is precedent. Our
group has reported that inhibiting T-cell DNA methylation

EFFECTS OF A Dnmt1 NULL MUTATION ON AGING
B275
The observation that DNA hypomethylation occurs after
11 months of age in the control mice and correlates with the
development of autoimmunity is consistent with the previ-
ously noted association between T-cell DNA hypomethyla-
tion and autoimmunity. In this case hypomethylation was
caused by aging of the control mice, rather than pharmaco-
logic treatment of T lymphocytes. This association is sup-
ported by the delayed appearance of autoimmunity and the
progressive increase in DNA methylation in the knockout
mice. It is also of interest that the relatively hypomethylated
young knockout mice did not develop autoimmunity. This
suggests either that autoimmunity develops slowly and was
prevented by the compensatory methylation, or that the
methylation status of genes crucial to the development of
autoimmune status is tightly regulated, and is methylated
normally despite a total decrease in genomic dmC content.
It is possible that DNA hypomethylation contributes to
autoimmunity in normal aging and our drug-induced model,
because similar features are seen in both models, including
the development of anti-DNA antibodies and an autoim-
mune liver disease with a chronic, focal inflammatory infil-
trate. Renal and lung lesions were also more common in the
control mice, and they are seen in the drug-induced DNA
hypomethylation model (6). The features of autoimmunity
in C57BL/6 mice are more pronounced when they are at 24
months of age (10), suggesting that examination of older
mice might give further evidence of autoimmunity. The
aged C57BL/6 mice also developed sialadenitis, which is in
contrast to the drug-induced model. It is possible that the
sialadenitis is a strain-specific feature, and C57BL/6 mice
have not been studied in the drug-induced lupus model. Al-
ternatively, other changes in methylation patterns associated
with aging but not produced in the drug-induced model may
contribute, or methylation change in the salivary glands
themselves might participate.
There were also several observations indicating that the
development of age-dependent changes in the immune sys-
tem are different between the Dnmt1 knockout mice and
controls. The memory subset, which increases with age, de-
veloped more slowly in the knockout mice. Similarly, IL-2
responses decreased more slowly with age in the knockout
mice. The mechanisms causing the appearance of the mem-
ory subset and the decrease in IL-2 production with aging
are unknown. A decrease in T-cell proliferative capacity in
the knockout mice could explain both observations, but pro-
liferative capacity was actually greater in the knockouts
than in controls at 18 months, arguing against this interpre-
tation. The observation that the decrease in IL-2 production
and development of the memory subset is delayed in the
knockout mice suggests that changes in DNA methylation
may be responsible for their appearance in the control mice,
and that preventing DNA hypomethylation slows their de-
velopment, although other mechanisms are possible, includ-
ing increased genomic instability caused by DNA hypo-
methylation (33). If the alterations caused by the mutation
prevent or delay the development of immune senescence, an
important test of this hypothesis will be to determine if mice
with heterozygous Dnmt1 deficiency live longer and are
more disease resistant than their wild-type littermates.
These studies are in progress.
The identity of the lymphocyte genes affected by age-
dependent changes in DNA methylation is of interest and
highly relevant to these studies. Future studies will have
to address this question. However, to approximate the num-
ber of genes affected, our group has treated human CD4ꢁ
T-lymphocyte lines with the DNA methylation inhibitor
2-deoxy-5-azacytidine, and we have compared gene expres-
sion in treated and untreated cells by using Affymetrix oli-
gonucleotide arrays. Approximately 619 out of 7000 genes
tested demonstrated a twofold or greater change in their
level of expression (unpublished results). Furthermore, our
group has reported that ꢀ15% of human T-lymphocyte
genes with CpG islands are variably methylated on a clonal
basis by middle age (34). These observations suggest that a
relatively large number of genes could be modified by age-
dependent changes in DNA methylation.
An alternative interpretation of these results is that effects
of the Dnmt1 mutation on nonlymphoid tissues are respon-
sible for the immunologic effects, preventing the develop-
ment of autoimmunity and immune senescence indirectly.
This seems unlikely, because adoptive transfer studies dem-
onstrate that lymphocytes from young mice will restore im-
mune responses in old recipients (35), and that lymphocytes
from old mice will induce autoimmunity in young recipients
(36). This suggests that these age-dependent phenomena are
unique to the lymphocytes and are not affected by the host.
Exclusion of this possibility will require adoptive transfer
experiments, in which lymphocytes from the control group
are transferred into the knockout mice, and retention of phe-
notype assessed.
In summary, in this study, wild-type control mice demon-
strated an age-dependent decrease in T-cell genomic dmC
content and developed autoimmunity similar to that re-
ported by our group to be induced by hypomethylated T
cells, whereas the Dnmt1 mice increased lymphocyte dmC
content with age and demonstrated delayed development of
autoimmunity. This supports the proposed association be-
tween DNA hypomethylation and autoimmunity (5,6). The
observation that dmC content increased with aging in the
knockout mice also suggests that counteracting DNA hy-
pomethylation may prevent immune changes with aging.
Persistance of high expression of MeCP2 may also help
maintain methylation levels with aging. Immune senescence
is a complex area, with many potential abnormalities noted,
and multiple mechanisms possible. The data presented in
this report suggest that one mechanism might be changes in
DNA methylation patterns, with subsequent effects on gene
expression and chromatin structure.
Acknowledgments
This work was supported by the University of Michigan Nathan Shock
Center (AG13282), PHS Grants AG014783, AR42525, AI42753, and AR/
AI01977, and a Merit grant from the Department of Veterans Affairs and
the Geriatrics Research, Education, and Clinical Center of the Ann Arbor
VA Hospital.
We thank Ms. Janet Stevens for her expert secretarial assistance. We
also acknowledge the advice, support, and encouragement freely given by
the late Dr. Monte Hobbs.
Address correspondence to Bruce Richardson, MD, PhD, 5310 Cancer
Center and Geriatrics Center Bldg., Ann Arbor, MI 48109-0940. E-mail:
Brichard@umich.edu

B276
YUNG ET AL.
References
20. Ono T, Uehara Y, Kurishita A, Tawa R, Sakurai H. Biological signifi-
cance of DNA methylation in the ageing process. Age Ageing. 1993;
22:S34–S43.
21. Ono T, Shinya K, Uehara Y, Okada S. Endogenous virus genomes be-
come hypomethylated tissue—specifically during aging process of
C57BL mice. Mech Ageing Devel. 1989;50:27–36.
22. Laird PW, Jackson-Grusby L, Fazeli A, et al. Suppression of intestinal
neoplasia by DNA hypomethylation. Cell. 1995;81:197–205.
23. Ng H-H, Bird A. DNA methylation and chromatin modification. Curr
Opin Genet Dev. 1999;9:158–163.
24. Antequera F, Bird A. Number of CpG islands and genes in human and
mouse. Genetics. 1993;90:11995–11999.
25. Yang J, Deng C, Hemati N, Hanash SM, Richardson BC. Effect of mi-
togenic stimulation and DNA methylation on human T-cell DNA
methyltransferase expression and activity. J Immunol. 1997;159:
1303–1309.
26. Okano M, Xie S, Li E. Cloning and characterization of a family of
novel mammalian DNA (cytosine-5) methyltransferases. Nat Genet.
1998;19:219–220.
27. Bhattacharya SK, Ramchandani S, Cervoni N, Szyf M. A mammalian
protein with specific demethylase activity for mCpG DNA. Nature.
1999;397:579–583.
28. Ramchandani S, Bhattacharya SK, Cervoni N, Szyf M. DNA methyla-
tion is a reversible biological signal. Proc Natl Acad Sci USA. 1999;
96:6107–6112.
29. Ng H-H, Zhang Y, Hendrich B, et al. MBD2 is a transcriptional re-
pressor belonging to the MeCP1 histone deacetylase complex. Nat
Genet. 1999;23:58–61.
30. Weiss A, Keshet I, Razin A, Cedar H. DNA demethylation in vitro: in-
volvement of RNA. Cell. 1996;86:709–718.
31. Kaludov N, Wolffe A. MeCP2 driven transcriptional repression in
vitro: selectivity for methylated DNA, action at a distance and contacts
with the basal transcription machinery. Nucl Acids Res. 2000;28:
1921–1928.
32. Yu F, Thiesen J, Strätling W. Histone deacetylase-independent tran-
scriptional repression by methyl-CpG-binding protein 2. Nucl Acids
Res. 2000;28:2201–2206.
33. Chen RZ, Pettersson U, Beard C, Jackson-Grusby L, Jaenisch R. DNA
hypomethylation leads to elevated mutation rates. Nature. 1998;395:
89–93.
34. Zhu-XX, Deng C, Kuick R, et al. Analysis of human peripheral blood
T cells and single-cell derived T cell clones uncovers extensive clonal
CpG island methylation heterogeneity throughout the genome. Proc
Natl Acad Sci USA. 1999;96:8058–8063.
35. Orme IM. Aging and immunity to tuberculosis: increased susceptibil-
ity of old mice reflects a decreased capacity to generate mediator T
lymphocytes. J Immunol. 1987;138:4414–4418.
36. Kanda K, Onji O, Ohta Y. Spontaneous occurrence of autoimmune
cholangitis in senescent mice. J Gastroenterol Hepatol. 1993;8:7–14.
1. Bird A, Wolffe A. Methylation-induced repression—belts, braces, and
chromatin. Cell. 1999;99:451–454.
2. Razin A, Cedar H. DNA methylation and embryogenesis. In: Jost JP,
Saluz HP, eds. DNA Methylation: Molecular Biology and Biological
Significance. Basel, Switzerland: Birkhäuser Verlag; 1993:343–357.
3. Mostoslavsky R, Bergman Y. DNA methylation: regulation of gene
expression and role in the immune system. Biochim Biophys Acta.
1997;1333:F20–F40.
4. Turker M, Bestor T. Formation of methylation patterns in the mamma-
lian genome. Mutat Res. 1997;386:119–130.
5. Quddus J, Johnson KJ, Gavalchin J, et al. Treating activated CD4ꢁ
T cells with either of two distinct DNA methyltransferase inhibitors,
5-azacytidine or procainamide, is sufficient to induce a lupus-like dis-
ease in syngeneic mice. J Clin Invest. 1993;92:38–53.
6. Yung RL, Quddus J, Chrisp CE, Johnson KJ, Richardson BC. Mecha-
nisms of drug induced lupus. I. Cloned Th2 cells modified with DNA
methylation inhibitors in vitro cause autoimmunity in vivo. J Immunol.
1995;154:3025–3035.
7. Golbus J, Palella TD, Richardson BC. Quantitative changes in T cell
DNA methylation occur during differentiation and ageing. Eur J Im-
munol. 1990;20:1869–1872.
8. Wilson VL, Smith RA, Ma S, Cutler RG. Genomic 5-methyldeoxycy-
tidine decreases with age. J Biol Chem. 1987;262:9948–9951.
9. Vanyushin BF, Nemirovsky LE, Klimenko VV, Vasiliev VK, Belozer-
sky AN. The 5-methylcytosine in DNA of rats. Gerontologia. 1973;
19:138–152.
10. Hayashi Y, Utsuyama M, Kurashima C, Hirokawa K. Spontaneous de-
velopment of organ-specific autoimmune lesions in aged C57BL/6
mice. Clin Exp Immunol. 1989;78:120–126.
11. Klinman D. Similarities in B cell repertoire development between au-
toimmune and aging normal mice. J Immunol. 1992;148:1353–1358.
12. Miller RA. Immune system. In: Masoro E, ed. Handbook of Physiol-
ogy: Aging. New York: Oxford University Press; 1995:555–590.
13. Li E, Bestor TH, Jaenisch R. Targeted mutation of the DNA methyl-
transferase gene results in embryonic lethality. Cell. 1992;69:915–926.
14. Bestor T, Laudano A, Mattaliano R, Ingram V. Cloning and sequenc-
ing of a cDNA encoding DNA methyltransferase of mouse cells. J Mol
Biol. 1988;203:971–983.
15. Austin, H III, Muenz L, Joyce K, et al. Prognostic factors in lupus ne-
phritis: contribution of renal histologic data. Am J Med. 1983;75:382–
391.
16. Hobbs MV, Weigle WO, Noonan DJ, et al. Patterns of cytokine gene
expression by CD4ꢁ T cells from young and old mice. J Immunol.
1993;150:3602–3614.
17. Yung RL, Chang S, Hemati N, Johnson K, Richardson BC. Mecha-
nisms of drug-induced lupus. IV. Comparison of procainamide and
hydralazine with analogs in vitro and in vivo. Arthritis Rheum. 1997;
40:1436–1443.
18. Deng C, Yang J, Scott J, Hanash S, Richardson B. Role of the ras-
MAPK signaling pathway in the DNA methyltransferase response to
DNA hypomethylation. Biol Chem. 1998;379:1113–1120.
19. Littell RC, Folks JL. Asymptotic optimality of Fisher’s method of
combining independent tests. J Am Statist Assoc. 1971;66:802–806.
Received October 8, 2000
Accepted January 29, 2001
Decision Editor: Edward Masoro, PhD