Wang, X. et al. TCR-dependent transformation of mature memory phenotype T cells in mice. J. Clin. Invest. 121, 3834-3845
A fundamental goal in cancer research is the identification of the cell types and signaling pathways capable of initiating and sustaining tumor growth, as this has the potential to reveal therapeutic targets. Stem and progenitor cells have been implicated in the genesis of select lymphoid malignancies. However, the identity of the cells in which mature lymphoid neoplasms are initiated remains unclear. Here, we investigate the origin of peripheral T cell lymphomas using mice in which Snf5, a chromatin remodelling-complex subunit with tumor suppressor activity, could be conditionally inactivated in developing T cells. In this model of mature peripheral T cell lymphomas, the cell of origin was a mature CD44hiCD122loCD8⁺ T cell that resembled a subset of memory cells that has capacity for self-renewal and robust expansion, features shared with stem cells. Further analysis showed that Snf5 loss led to activation of a Myc-driven signaling network and stem cell transcriptional program. Finally, lymphomagenesis and lymphoma proliferation depended upon TCR signaling, establishing what we believe to be a new paradigm for lymphoid malignancy growth. These findings suggest that the self-renewal and robust proliferative capacities of memory T cells are associated with vulnerability to oncogenic transformation. Our findings further suggest that agents that impinge upon TCR signaling may represent an effective therapeutic modality for this class of lethal human cancers.
3834 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011
TCR-dependent transformation of mature
memory phenotype T cells in mice
Miriam B.F. Werneck,
Boris G. Wilson,
Michael J. Kluk,
Christopher S. Thom,
Jonathan W. Wischhusen,
Julia A. Evans,
Jonathan L. Jesneck,
Courtney G. Sansam,
and Charles W.M. Roberts
Department of Pediatric Oncology, Dana-Farber Cancer Institute, Boston, Massachusetts, USA.
Division of Hematology/Oncology, Children’s Hospital Boston,
Boston, Massachusetts, USA.
Department of Pediatrics, Harvard Medical School, Boston, Massachusetts, USA.
Department of Cancer Immunology and AIDS,
Dana-Farber Cancer Institute, Boston, Massachusetts, USA.
Department of Pathology, Harvard Medical School, and
Department of Pathology,
Brigham and Women’s Hospital, Harvard Medical School, Boston, Massachusetts, USA.
Broad Institute of Harvard and Massachusetts Institute of Technology, Boston, Massachusetts, USA.
A fundamental goal in cancer research is the identification of the cell types and signaling pathways capable
of initiating and sustaining tumor growth, as this has the potential to reveal therapeutic targets. Stem and
progenitor cells have been implicated in the genesis of select lymphoid malignancies. However, the identity of
the cells in which mature lymphoid neoplasms are initiated remains unclear. Here, we investigate the origin of
peripheral T cell lymphomas using mice in which Snf5, a chromatin remodelling–complex subunit with tumor
suppressor activity, could be conditionally inactivated in developing T cells. In this model of mature periph-
eral T cell lymphomas, the cell of origin was a mature CD44
T cell that resembled a subset of
memory cells that has capacity for self-renewal and robust expansion, features shared with stem cells. Further
analysis showed that Snf5 loss led to activation of a Myc-driven signaling network and stem cell transcriptional
program. Finally, lymphomagenesis and lymphoma proliferation depended upon TCR signaling, establishing
what we believe to be a new paradigm for lymphoid malignancy growth. These findings suggest that the self-
renewal and robust proliferative capacities of memory T cells are associated with vulnerability to oncogenic
transformation. Our findings further suggest that agents that impinge upon TCR signaling may represent an
effective therapeutic modality for this class of lethal human cancers.
Little is known definitively about the cellular origins of cancer,
as initiation occurs long before tumors become apparent. Conse-
quently, the identity of the initiating cell is frequently speculative,
based upon extrapolations from tumor cell phenotypes. However,
since the selective pressure that occurs during oncogenic trans-
formation is intense, nascent cancer cells can undergo substantial
phenotypic evolution, making the validity of such extrapolations
uncertain. One approach to investigate the potential origins of
cancer has been to expose selected cell populations to exogenously
expressed oncogenes. Such studies have shown that long-lived stem
cells and early progenitor cells are capable of giving rise to cancers,
although such data is not derived from spontaneously arising can-
cers in vivo (1, 2). Additional support for a stem cell origin of cancer
has come from findings that stem cell– and self-renewal–associated
programs are enriched in multiple tumor types (1, 3–7).
However, at least some cancers may arise from more differen-
tiated cells. For instance, transduction of Ink4a/Arf
with constitutively active EGFR can induce a high-grade glioma
phenotype (8). Additionally, ectopic expression of MLL-AF9 can
drive transformation of both committed progenitors and cells
expressing mature myeloid lineage-specific antigens (1, 9). Con-
sequently, the intrinsic cellular features that confer the greatest
susceptibility to transformation in vivo and the mechanisms that
underlie the reprogramming are largely unclear.
The SWI/SNF complex, also known as the BRG1-associated fac-
tor (BAF) complex, regulates chromatin structure and plays funda-
mental roles in the epigenetic regulation of gene expression and in
the control of cell fate (10). Its activity has been implicated in the
maintenance of embryonic stem cell pluripotency and in enhancing
the formation of iPS cells (11, 12). Inactivating mutations in SWI/
SNF subunits are increasingly being identified at high frequency
in a variety of human cancer types, including SWI/SNF-related,
matrix-associated, actin-dependent regulator of chromatin, sub-
family b, member 1 (SMARCB1, also known as SNF5) in rhabdoid
tumors, Schwannomatosis, and a familial cancer predisposition
syndrome; AT-rich interactive domain 1A (ARID1A, also known
as BAF250A) mutations in ovarian and endometrioid carcinomas;
SWI/SNF-related, matrix-associated, actin-dependent regulator
of chromatin, subfamily a, member 4 (SMARCA4, also known as
BRG1) in lung cancers; and most recently polybromo 1 (PBRM1,
also known as BAF180) in renal carcinomas (10, 13–18). In addi-
tion to homozygous inactivation, haploinsufficiency for SWI/SNF
subunits has been implicated in a number of cancers as well. The
SWI/SNF complex serves specific roles in lymphoid development
and may also play a role in lymphoid malignances, as 50% of T cell
prolymphocytic leukemias display deletions at 22q11, the location
of SNF5 (19, 20). Also, inactivation of Snf5 in mice leads to rapid
onset of mature peripheral T cell lymphomas (PTCLs) in all mice,
with a median onset of only 11 weeks (21, 22). Consequently, muta-
tion of the Swi/Snf complex is relevant to a variety of lethal human
cancers, making its underlying biology of great interest.
Despite its roles in human cancer, the mechanisms underlying
the tumor suppressor activity of the SWI/SNF complex, its role in
lymphoid development, and the origin of these cancers are poorly
Authorship note: Xi Wang and Miriam B.F. Werneck contributed equally to this
Conflict of interest:
The authors have declared that no conflict of interest exists.
Citation for this article:
J Clin Invest. 2011;121(10):3834–3845. doi:10.1172/JCI37210.
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011 3835
understood. The T cell compartment provides an ideal model with
which to genetically pinpoint the origin of cancer and elucidate
mechanisms in view of its well-characterized stepwise develop-
ment from HSCs to lineages of mature T cells. In addition, unlike
differentiated cells in other tissues, a subset of mature T cells,
termed memory cells, can renew themselves and persist through-
out the lifetime of an individual. This property is associated with
expression of a transcriptional program associated with cellular
self-renewal, similar to the program of long-term HSCs (23), a cell
type that has been implicated in the initiation of cancer (2, 24, 25).
The susceptibility of HSCs to transformation has been linked to
their intrinsic capacities for self-renewal and proliferation, leading
to our hypothesis that memory T cells, in which self-renewal is
uncoupled from the stem/progenitor stage of differentiation, may
be similarly susceptible to malignant transformation.
Lastly, although postulated over 25 years ago (26), the contribu-
tion of TCR signaling to the initiation and expansion of T cell lym-
phomas has been difficult to define and has remained speculative.
Such a role would be of great interest, both because of mechanis-
tic insight and because of the therapeutic implications. The chal-
lenge in evaluating this arises in part because immune restriction
precludes such questions from being directly tested using xeno-
transplantation. Thus primary animal models of mature T cell
lymphoma are needed, which have been lacking.
Here, we have generated a staged series of genetically engineered
in vivo models in order to identify the population of cells capable
of giving rise to spontaneous lymphomas that possess phenotypic
markers of mature cells, to characterize the underlying mecha-
nisms, and to evaluate the role of TCR signaling in this process.
Deletion of Snf5 at a late stage in T cell development results in lymphom-
agenesis. Inactivation of Snf5 in mice using an inducible Mx-Cre
transgene results in the rapid formation of mature CD3
PTCLs and sarcomas with 100% penetrance and a median onset of
11 weeks (22). To initiate investigation of the mechanism of lym-
phomagenesis, we crossed Snf5-conditional mice to a series of del-
eter strains that act at different stages during T cell developmental
progression. The hCD2-Cre transgene is expressed within the earli-
est stem/progenitor stages of T cell development, becoming active
at the common lymphoid progenitor (CLP) stage that gives rise to
the B, T, and NK cell lineages; the Lck-Cre transgene is expressed
during early intrathymic development as nascent T cells arrive in
the thymus; and the CD4-Cre transgene is not expressed until late
in thymic development, at the CD4
stage (Figure 1). Consistent with the early expression of hCD2-
Cre, thymocyte precursors in hCD2-Cre floxed Snf5 mice (hCD2-
mice) already harbored deleted Snf5 genes upon their
arrival in the thymus at the CD4
double-negative 1 (DN1)
stage (Supplemental Figure 1A; supplemental material available
online with this article; doi:10.1172/JCI37210DS1). This resulted
in a 90% reduction in thymocyte numbers, reflecting a block at the
DN3 stage of development (Figure 2A).
Snf5 deletion driven by the Lck-Cre transgene occurred some-
what later and was nearly complete by the DN4 stage (Supplemen-
tal Figure 1A). Similar to the effect seen in hCD2-Cre mice, thymo-
cyte development in Lck-Cre animals was blocked at the DN3 stage
(Figure 2A). However, consistent with the known incomplete pen-
etrance of Lck-Cre expression (27), a small subpopulation (<10%)
of thymocytes did not delete Snf5 until after the DN3 stage. Con-
sequently, some cells proceeded past the DN3 stage and gave rise
to mature Snf5-deficient peripheral T cells in Lck-Cre Snf5
a situation that did not occur in hCD2-Cre mice (Figure 2B).
In contrast to those of hCD2-Cre and Lck-Cre, the developmental
effects caused by the CD4-Cre transgene differed substantially. This
transgene reproducibly led to a loss of Snf5 protein within mature T
lymphocytes but not earlier. Although CD4 is expressed by DP thy-
mocytes, due to the relatively long half-life of the Snf5 protein, loss
of Snf5 was not complete until mature T cells reached the periphery
(Supplemental Figure 1B). As a result, CD4-Cre Snf5
the CD2-Cre and Lck-Cre mice, possessed normal thymocyte num-
bers and developmental thymocyte profiles (Figure 2A).
To determine whether the Snf5-deficient lymphomas were initi-
ated by stem/progenitor cells and later acquired a mature pheno-
type or whether they arose from mature cells, we aged cohorts of
mice from each of the Cre-transgene expressing lines. Mice from the
line aged normally and were not cancer prone. In
contrast, all mice from both the Lck-Cre Snf5
and CD4-Cre Snf5
Schematic of T cell development. CLP cells originate in the bone marrow and migrate to the thymus. Cells at the CD4
DN stage are
divided into 4 sequential subsets (DN1–DN4) according to expression of CD44 and CD25. Expression of the pre-TCR at DN3 leads to progres-
sion to DN4, followed by expression of CD4 and CD8 and progression to the CD4
DP stage. Positive selection then leads to commitment
toward the CD4 or the CD8 T cell lineage. Mature CD4 and CD8 single-positive cells then migrate to the periphery. The hCD2-Cre transgene
is expressed at the CLP stage, leading to Snf5 deletion prior to arrival in the thymus. The Lck-Cre transgene is initially activated during early
intrathymic development. The CD4-Cre transgene results in loss of Snf5 protein only in mature (CD4 and CD8) T cells in the periphery.
3836 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011
lines rapidly developed monoclonal/oligoclonal TdT
, and CD4
mature PTCLs (Figure 2C and Supplemental Fig-
ures 2 and 3) located exclusively in spleen, liver, and occasionally
lymph nodes but never thymus. Consequently, loss of Snf5 from
mature T cells (CD4-Cre) but not early stem/progenitors (hCD2-
Cre) resulted in lymphoma development.
subpopulation of T cells is enriched in the absence of
Snf5 and propagates lymphomas in vivo. In order to gain insight into the
pathways and mechanisms driving oncogenesis, we next character-
ized the effect of Snf5 loss on T cell populations. We used a floxed
GFP reporter transgene to identify individual cells that express Cre
(Supplemental Figure 4). Analysis of GFP
T cells in 3- to 4-
week-old Lck-Cre Snf5
mice and in CD4-Cre Snf5
before overt lymphoma development) revealed a predominant phe-
notype of CD44
(Figure 3A). High expression of CD44
T cells is a marker of memory
phenotype (MP). Based on expression
of the cell surface markers CD44 and
CD122, mature CD8
T cells can be
divided into 3 subsets: CD44
MP cells (28). The
T cells in the Lck-Cre Snf5
mice had a surface
phenotype consistent with that of the
subset of MP CD8
We next examined the Snf5-defi-
cient lymphoma cells. A subpopula-
tion of lymphoma cells (15% of total)
expressed the CD44
a similar subpopulation of CD44
was noted in a cell line that we gener-
ated from a Snf5-deficient lymphoma
(Figure 3, B and C). Since expression
of CD44 is a marker of the self-renew-
ing, tumor-propagating subpopula-
tion in solid tumors of several tissue
types (2, 29), we hypothesized that
MP cells were responsible for
sustaining the lymphoma. To test this
hypothesis, we sorted the Snf5-defi-
cient cell line into CD44
fractions before incubation for 40 days
in culture, followed by repeat analysis
of CD44 expression. The phenotype of
fraction was stable: all cells
after extended time
in culture. In contrast, cells from the
fraction gave rise to both CD44
progeny (Figure 3D), thus
recapitulating the parental lymphoma
phenotype, consistent with the hypoth-
esis that CD44
cells can self-renew and
give rise to CD44
As a more direct test of this hypoth-
esis, we injected graded numbers of
phoma cells into recipient mice. None
of the mice injected with 10
cells developed lymphoma,
while all recipients of either 10
and 75% of mice
injected with as few as 10
cells developed tumors (Table 1
and Figure 3E). The heterogeneous expression patterns of CD44
in the secondary tumors resembled the phenotypic complexity of
the lymphoma cells from which they were derived. These second-
ary tumors contained both CD44
populations (Figure 3F). Together, these in vitro and in vivo obser-
vations indicate that CD8
cells are the tumor-
We next evaluated whether lymphoma formation was dependent
upon intrinsic properties of MP cells. To determine whether addi-
tional surface markers expressed by CD44
cient lymphoma cells were consistent with the surface phenotype
Deletion of Snf5 in T cells results in mature T cell lymphoma. (A) Flow cytometric analysis of thy-
mocytes from WT (CD2-Cre Snf5
), Lck-Cre Snf5
, hCD2-Cre Snf5
, and CD4-Cre Snf5
The numbers above the plots indicate the total numbers of thymocytes averaged from 9 WT mice,
6 Lck-Cre Snf5
mice, 9 hCD2-Cre Snf5
mice, and 6 CD4-Cre Snf5fl/fl mice. The top row shows
CD4 versus CD8 staining for all thymocytes. The bottom row is gated to show only the CD4
DN population, and CD25 versus CD44 staining is used to resolve the DN1, DN2, DN3, and DN4
populations, as labeled in blue. The percentage of cells within each quadrant is labeled in red. (B)
T cells were sorted from the spleens of Lck-Cre Snf5
mice (n = 8) or
mice (n = 12), and deletion of the Snf5 allele was quantified by quantitative real-
time PCR. Due to known variable penetrance of Lck-Cre expression, some T cells in Lck-Cre Snf5
mice do not undergo deletion until later stages of development. Consequently, the Lck-Cre Snf5
mice contain a population of mature peripheral T cells that are Snf5 deficient, a key distinction from
the hCD2-Cre Snf5
mice in which the deleted allele was undetectable (U/D) in peripheral T cells.
(C) Tumor-free survival curve of hCD2-Cre Snf5
mice (n = 24), Lck-Cre Snf5
mice (n = 11), and
(n = 16) mice.
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011 3837
MP cells, we examined expression of CD25,
CD62L, CD69, and CD127. In each case, we found the expres-
sion pattern was consistent with that of the CD44
population (ref. 28 and Supplemental Figure 5A). Moreover, like
MP cells and unlike naive cells and other MP pop-
ulations, proliferation of Snf5-deficient lymphomas in culture was
not enhanced by addition of recombinant IL-15 (Supplemental
Figure 5B), and these tumor cells grew in IL-15–deficient recipi-
ents with the same kinetics as in WT hosts (Supplemental Figure
5C). We then performed array-based gene expression analysis of
the purified progenitor fraction from the Snf5-deficient lym-
phoma cell line. We first compared the genome-wide expression
signature of the Snf5-deficient lymphoma cells to a published
data set containing both naive and memory expression signatures
(23). The analysis revealed that the gene expression program in
subset of lymphoma cells is distinct from that of naive
T cells but highly similar to that of memory T cells (Figure 4A). In
contrast, the gene expression signature of the CD44
tion is distinct from that of both naive and memory cells. Next, we
performed gene set enrichment analysis (GSEA) (30), using a set of
genes specifically expressed in memory cells, and also found signif-
icant enrichment of this set within the CD44
cells are the tumor-propagating cells. (A) CD3
splenocytes from 4-week-old Lck-Cre GFP Snf5 WT (black lines) or Lck-Cre
mice (red lines) were isolated, stained with antibodies for CD122 and CD44, and analyzed by FACS. Representative plots are shown.
(B) Immunoblot of Snf5 expression in WT CD8 T cells and a lymphoma cell line. (C) Cells from the Snf5-deficient lymphoma cell line uniformly
expresses CD3 and CD8, lack CD122, but consist of 2 populations with respect to CD44, CD44
. The percentage of cells within each
gate is indicated. (D) Snf5-deficient lymphoma cells were double sorted into CD44
populations and then maintained in culture. CD44
and CD122 staining is shown from cells 5, 15, and 40 days after sorting. Only the CD44
cells were capable of recapitulating the parent cell line
phenotype by giving rise to both the CD44
and the CD44
populations. The percentage of cells within each quadrant is indicated. (E) Survival
curve of mice injected with specified number of CD44
cells derived from Snf5-deficient lymphomas. Each group contains 4 mice. (F) Flow
cytometry analysis of tumors arising from recipient mice injected with 100 CD44
cells. The percentage of cells within each gate is indicated.
3838 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011
cell population (P < 0.0001) (Figure 4B). Finally, we specifically
evaluated the expression of RANTES, a hallmark gene expressed
in memory cells but not naive cells (31, 32), and found that it was
highly expressed in the CD44
subset (Figure 4C).
To address the hypothesis that lymphomas arose from
MP cells, we asked whether there was a correlation
between enrichment of this population and the tumor-promoting
activity of Snf5 loss. Brg1 is a core ATPase subunit of the Swi/Snf
complex and, like SNF5, is mutated in a subset of human malig-
nant rhabdoid tumors (10). Inactivation of Brg1 results in altera-
tions in T cell development that are nearly indistinguishable from
those of Snf5 loss (33, 34) but never results in lymphoma formation,
regardless of the promoter driving Cre expression (Mx-Cre, Lck-Cre,
or CD4-Cre) (21, 33). We evaluated peripheral T cells in CD4-Cre
mice by immunofluorescence and found that, unlike Snf5
loss, Brg1 loss does not lead to enrichment of CD44
MP cells (Sup-
plemental Figure 6), indicating a strong correlation between enrich-
ment of MP cells in the absence of Snf5 and tumorigenesis.
Myc and stem cell–associated programs are aberrantly activated in Snf5-
deficient lymphomas. These findings suggest that Snf5 loss preferen-
tially promotes survival of MP cells and expression of a MP gene
expression program within lymphoma cells. We next sought to
investigate the mechanism by which Snf5 loss drives transforma-
tion of this population. We recently demonstrated that Snf5 loss
leads to upregulation of stem cell–associated programs (35). It has
been reported that the similarities between gene expression pro-
grams in ES cells and cancers are due in large part to the pervasive
activation of a Myc regulatory network rather than to a conserved
stem cell program per se (36). We therefore evaluated expression of
the Myc regulatory network as well as the newly identified ES cell
core module, which is more specific for ES cell identity. The Myc
module is composed of genes that are common targets of 7 Myc
network factors (Myc, Max, nMyc, Dmap1, E2F1, E2F4, and Zfx) (36),
while the stem cell core module is composed of genes co-occupied
by at least 7 factors among 9 ES core factors (Smad1, Stat3, Klf4,
Oct4, Nanog, Sox2, Nac1, Zfp281, and Dax1) (36). We found that both
the Myc module and the stem cell core module were enriched in
lymphoma cells compared with WT T cells (Figure 5A). Western
blot and GSEA further confirmed the upregulation of Myc itself
and activation of an independently identified larger group of Myc
targets (Figure 5, B–D). Notably, while these signatures were differ-
entially expressed in the bulk Snf5-deficient lymphoma population
compared with those in WT T cells, they were similarly activated
in both CD44
populations (data not shown). Con-
sequently, the signatures reflected the effects of Snf5 loss but did
not underlie the differential capacity of CD44
to propagate or give rise to the bulk of the tumor. To determine
whether these signatures were driven by Snf5 loss or were rather
a general consequence of oncogenic transformation, we evaluated
the signatures after Snf5 inactivation in non-transformed pri-
mary fibroblast cells. Importantly, in contrast to the lymphoma
cells, Snf5 inactivation in primary fibroblasts results in cell cycle
arrest (37). Nonetheless, the Myc and ESC-like modules were both
enriched upon Snf5 loss (Figure 5, E and F), indicating that these
changes were caused by Snf5 loss and not simply due to increased
proliferation or transformation. Lastly, we asked whether these sig-
natures were similarly enriched in human SNF5-deficient malig-
nant rhabdoid tumors and found this to be the case (Figure 5G).
Initiation and maintenance of Snf5-deficient lymphomas depends on
TCR expression. While TCR engagement may be dispensable for the
survival of some MP subsets, the CD8
tion is dependent upon TCR signaling for survival and expansion
(28). Although oncogenic transformation results in acquisition of
unlimited proliferation and diminished dependence upon growth
factors, cancer cells can retain residual dependence upon pathways
that normally control proliferation within the lineage of origin,
such as breast cancer dependence upon estrogen and prostate can-
cer dependence upon androgen. We thus investigated the depen-
dence of the lymphoma cells upon TCR engagement and signaling.
Of note, the CD3 receptor complex is highly expressed on the lym-
phoma cells (Supplemental Figure 7A). Activation of TCR signaling
results in the activation of NFAT1 through calcineurin-mediated
dephosphorylation (38, 39), which can then activate Bclx, rendering
cells resistant to apoptosis. Western blotting revealed that NFAT1
was activated in Snf5-deficient lymphoma cells (Supplemental
Figure 7B), and both microarray and RT-PCR experiments revealed
upregulation of NF-κB and Bclx in these cells (Supplemental Fig-
ure 7, C and D), consistent with active TCR signaling.
We next evaluated whether the failure to provoke lymphomas
when Snf5 is deleted in immature thymocytes by hCD2-Cre could
be due to a lack of TCR signaling and whether it could thus be rem-
edied by provision of a mature TCR transgene. We chose the OT1
αβ TCR to evaluate the role of MHC class I–restricted TCR signal-
ing and introduced it into hCD2-Cre Snf5
mice, which otherwise
have a DN3 block and fail to develop lymphomas. The presence of
the transgene resulted in presence of Snf5
mature T cells in the
periphery (Figure 6A), and all mice developed CD3
lymphomas (Figure 6B), indicating that the addition of TCR trans-
gene was sufficient to drive lymphoma formation in this model.
To ask the reciprocal question, we evaluated whether elimination
of TCR expression would block oncogenesis. We therefore crossed
the Lck-Cre Snf5
mice onto a Rag2
background. The Lck-
mice failed to develop lymphoma (Figure 6C).
Together, these findings indicate that TCR expression is essential
for tumor initiation in Snf5 mutant mice.
That TCR expression is essential for transformation led us to
ask whether TCR-MHC interaction was essential for lymphoma
maintenance and proliferation. We began by evaluating the effects
of CD8 coreceptor blockade upon cultured lymphoma cells. Incu-
bation of lymphoma cells in the presence of APCs led to robust
proliferation (Figure 7A). Blockade of the CD8 TCR coreceptor
by anti-CD8 antibody inhibited the proliferative response of lym-
Tumorigenic lymphoma cells were highly enriched in the CD44
No. of cells Resultant no. of tumors
injected in recipients/total recipients
cells were isolated by flow cytometry, yielding
a purity of 99.8% for CD44
and 100% for CD44
cells. The indicated
number of cells of each phenotype was intravenously injected into
the tails of Rag2
recipient mice. The mice were observed daily for 6
months or until they became sick from lymphomas.
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011 3839
phoma cells (Figure 7B). Since this pathway depends on activation
of calcineurin and can be inhibited by cyclosporin A (CsA) (40), we
asked whether CsA treatment would affect lymphoma cell growth.
CsA inhibited proliferation of Snf5-deficient lymphoma cells (Fig-
ure 7C), and combined addition of CsA and anti-CD8 antibody to
lymphoma cultures exerted a synergistic inhibitory effect (Figure
7D). Importantly, TCR signaling was essential for proliferation of
the Snf5-deficient lymphoma cells and was not simply serving a
T cell survival function, as the reduction in thymidine incorpora-
tion caused by TCR blockade occurred without an increase in cell
death (Supplemental Figure 8).
Lastly, in order to more definitively test the dependence of Snf5-
deficient lymphomas upon TCR signaling, we evaluated the role
of engagement of the TCR by MHC/peptide ligand on lymphoma
growth in vivo. Lymphoma cells from Lck-Cre Snf5
transferred into MHC class I–deficient K
or control C57BL/6 hosts. Although NK lysis by MHC-deficient
host cells would not be expected to diminish lymphoma growth,
recipients were irradiated (4 Gy) to prevent potential NK effects.
We found that lymphomas developed in all control C57BL/6
recipients but in none of the MHC class I–deficient K
(Figure 7E), indicating that TCR engagement by MHC class I mol-
ecules is essential for lymphoma cell proliferation.
Downregulation of SNF5 in human PTCLs. Since Snf5 deletion in
murine T cells leads to tumorigenesis in mature T cells, we asked
whether SNF5 may play a role in the genesis of human PTCLs as
well. Malignancies in T cells can be classified into 2 major cate-
gories: precursor T cell lymphoblastic neoplasms, which express
markers consistent with maturing thymocytes, and PTCLs, which
express markers consistent with mature post-thymic T cells. We
used published gene expression data from human T cell can-
cers (42) to evaluate the levels of SNF5 expression. In normal
subset of Snf5-deficient lymphoma cells has memory-like features. (A) Evaluation of global gene expression using KNN analysis.
The signatures of CD44
subsets from 4 independent Snf5-deficient lymphomas were compared with previously published gene
signatures for naive and memory cell populations. Enrichment for the memory signature is indicated by an upward positive deflection of the bars,
while enrichment for the naive signature is indicated by a downward negative deflection. (B) GSEA of a set of genes highly expressed in memory
cells. Upward deflection of the green line indicates enrichment of the memory signature within the CD44
population (P < 0.0001). FDR, false
discovery rate; NES, normalized enrichment score. (C) Relative expression of RANTES (top) and Actb (bottom) by sorted CD44
lymphoma subpopulations determined by array-based gene expression analysis. Horizontal bars indicate the mean.
3840 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011
tissues, SNF5 is ubiquitously expressed. Using this data set, we
found that SNF5 is expressed both in normal T cells as well as in
immature lymphoblastic T cell leukemias and lymphomas (Fig-
ure 8). However, we found that SNF5 is expressed at reduced to
undetectable levels in 40% of PTCLs (Figure 8). To exclude the
possibility that peripheral T cells intrinsically have substantial
variation with respect to SNF5 expression level, we performed
immunohistochemical staining of SNF5 in normal spleen, lymph
node, and thymus from both humans and mice. We observed
that T cells and thymocytes expressed SNF5 and that the staining
intensity did not vary substantially (Supplemental Figures 9 and
10). Consequently, the reduced levels of SNF5 in PTCLs compared
with those in normal T cells and in lymphoblastic malignant cells
raised the possibility of a causative link between downregulation/
inactivation of SNF5 and the genesis of human PTCLs.
Transformation of mature T cells. Identification of the cells from
which cancers originate has been an area of intense interest,
albeit difficult to study. Although many tumors may well arise
from relatively undifferentiated cells, some lethal cancers
express features of differentiated mature cells. Perhaps the
most suggestive examples are from human PTCLs that express
monoclonal TCR or Ig rearrangements and express surface
markers characteristic of mature T or B cells (43–45). However,
it is difficult to establish the cell of origin of these cancers in
humans, because tumor initiation occurs long before patients
come to clinical attention. Moreover, characterization of the
tumor-propagating cells and evaluation of TCR signaling using
xenotransplantation is difficult as these lymphomas are largely
dependent upon the human microenvironment and do not
Snf5 loss leads to activation of Myc
module. (A) Average gene expression
) of core stem cell modules
and Myc regulatory network are tested
in WT CD8
T cells and Snf5-deficient
lymphoma cells. (B) Myc protein
levels are elevated in Snf5-deficient
lymphomas. Immunoblot analysis of
Myc in WT CD8
T cells and Snf5-
lymphoma cells. (C–G)
GSEA of defined (C, E, and G) Myc
module or (D and F) Myc target genes
in expression data from (C and D) puri-
fied Snf5-deficient CD8
cells compared with WT CD8
(E and F) Snf5-deficient MEFs com-
pared with WT MEFs, and (G) human
SNF5-deficient malignant rhabdoid
tumor (MRT) samples compared with
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011 3841
grow well in mice (46). Here, we have used a series of deleter
strains to generate genetically engineered murine models of
spontaneously arising PTCLs.
Snf5 is recurrently mutated in human cancers and has potent
tumor suppressor activity in mouse models, but the underlying
mechanism has been largely unclear. We and others have identified
epigenetic antagonism between Polycomb and Swi/Snf complexes,
and we recently reported an essential role for imbalance of this epi-
genetic relationship in transformation caused by Snf5 loss (35, 47,
48). In addition, we found an important role for aberrant activa-
tion of the Hedgehog-Gli pathway, a mediator of stem cell identity,
caused by Snf5 loss (49). Here, we build on these findings by show-
ing that Snf5 loss in T cells leads to activation of both Myc signa-
tures and core stem cell–associated programs. Therefore, the effects
of Snf5 inactivation in T cell lymphomas correlate with those in
primary fibroblasts, suggesting a theme of epigenetically mediated
activation of proliferation and stem cell programs after Snf5 loss.
Additionally, a major finding is that mature T cells that display
phenotype consistent with that of a subset
of memory T cells are uniquely responsive to the consequences of
Snf5 loss, thus providing insight into the initiation and mainte-
nance of these lymphomas. Given that (a) the lymphomas arise
exclusively in the periphery, in which they involve the spleen, liver,
and lymph nodes but never thymus; (b) that Snf5 protein is not
fully lost until cells reach the periphery; and (c) that the propagat-
ing cells express high levels of CD44, which is expressed at low
levels in CD8
thymocytes but at high levels in peripheral memory
cells, the tumor-originating T cell is likely mature and peripheral
in nature. However, we cannot totally exclude initiation from a
late thymic T cell. Memory CD8
T cells are a unique population
of differentiated cells that share several properties with stem cells,
including longevity and continuous and slow self-renewal, com-
bined with the capacity for extensive and robust proliferation (12,
50–53). Indeed, bioinformatic analysis has revealed that HSCs and
T cells share a similar transcriptional program of
activation (54) and self renewal (23), perhaps reflecting a predicted
evolutionary co-opting of cooperative gene sets (55). Collectively,
our findings suggest that unlike mouse embryonic fibroblasts
(MEFs) or naive T cells, memory T cells are intrinsically responsive
to activation of these stem cell–associated and Myc programs, thus
rendering them susceptible to transformation.
subset delineated in this report differs
in several ways from canonical memory T cells. This cellular subset
displays a high background rate of proliferation, expresses surface
markers characteristic of recently activated T cells, preferentially
homes to the spleen (the principal location of Snf5-deficient lympho-
mas), and has features of chronically activated T cells secondary to
indolent infection (28). As chronic infections are a well-recognized
risk factor for development of several types of lymphoma, this opens
the mechanistic possibility that chronic stimulation of memory
cells results in enhanced susceptibility to oncogenic transformation.
memory cells, like Snf5-deficient CD8
phomas, depend on TCR engagement and thus survive poorly in
MHC class I–deficient mice (28). Collectively, these properties and the
high basal rate of proliferation may make this memory subset particu-
larly susceptible to transformation. Lastly, while the precise molecular
function of CD44 is poorly understood, this surface protein marks
cancer stem cells in several types of solid tumors and is repressed by
Snf5 (2, 29, 47). Consequently, it is possible that the high levels of
CD44 expressed by CD8 memory cells and lymphoma-initiating cells
described here may also contribute to cellular transformation.
TCR signaling and lymphomagenesis. It has been shown that many
nonlymphoid cancers, such as prostate and breast cancers, continue
to depend on physiologic lineage-specific signals, such as andro-
gen and estrogen receptor signaling (56–58). However, the role of
immune receptor signaling in lymphoid tumors has been unclear.
Intriguingly, recent data suggest that a chronic immune response
to microbial or possibly self antigens may underlie the formation
of noncancerous, age-related expansions of CD8 memory cells, sug-
gesting that TCR signaling can contribute to clonal disorders (59).
With respect to PTCLs, genome-wide expression analyses have also
revealed that PTCL cells are enriched for a gene expression signature
of activated T cells (45, 60). Moreover, it has recently been shown
that the fusion of ITK-SYK caused by the t(5;9) (q33;q22) chromo-
somal translocation present in 17% of PTCLs can mimic a TCR sig-
nal and can drive tumorigenesis in a conditional mouse model (61).
However, any role for TCR engagement in malignancy has remained
speculative. Collectively, our data demonstrate that, despite being
monoclonal and transplantable, TCR signaling is essential for
proliferation of these mature T cell lymphomas: low-affinity TCR-
MHC interactions that promote survival of naive CD8
T cells and
subset of memory CD8
cells are sufficient to
drive lymphoma expansion. While expression of a transgenic TCR
is essential for lymphoma development in CD2-Cre mice, we can-
not exclude the possibility that TCR signaling does not specifically
drive initial transformation but rather facilitates maturation and is
subsequently required to maintain proliferation of the lymphoma
cells after transformation.
Oncogenesis caused by Snf5 loss is TCR signaling dependent. (A) CD3
T cells were sorted from the spleens of hCD2-Cre Snf5
(n = 6) or hCD2-Cre OT-1 Snf5
mice (n = 4), and deletion of the Snf5 allele was quantified by quantitative real-time PCR. The deleted allele
was undetectable in CD3
T cells from the hCD2-Cre Snf5
mice. (B) Tumor-free survival curve of hCD2-Cre Snf5
mice (n = 24) and
(n = 8) mice. (C) Tumor-free survival curve of Lck-Cre Snf5
mice (n = 11) and Lck-Cre Snf5
(n = 5) mice.
3842 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011
The relevance of this model to human lymphomas is enhanced
by our finding that 40% of human PTCLs express SNF5 at low
or undetectable levels. It is noteworthy that while inactivating
mutations of SNF5 drive cancer formation, hypofunction is
also associated with cancer. For instance, inactivating muta-
tions occur in rhabdoid tumors, while less severe mutations are
found in familial Schwannomatosis (13, 15, 62, 63). Similarly,
reduced expression of SNF5 is associated with high-risk dis-
ease in acute lymphoblastic leukemia (64). Additionally, 50% of
T cell prolymphocytic leukemias carry deletions at 22q11, the
region in which SNF5 is encoded (19), and a possible role for
haploinsufficiency of SNF5 in the genesis of these mature T cell
cancers has been raised (20). Similarly, Brg1-haploinsufficient
mice are predisposed to breast cancers in which the remaining
WT allele is always retained. Most recently, the ARID1A subunit
of the SWI/SNF complex has been found specifically mutated
in 50% of ovarian clear cell carcinomas (65, 66). Notably, while
many of these cancers have biallelic inactivating mutations and
complete loss of the ARID1A protein, a substantial percentage
harbor only heterozygous mutations and express some ARID1A
protein, suggesting a haploinsufficient mechanism (66). Col-
lectively, it is clear that a reduction in the levels of expression
of SWI/SNF subunits can drive cancer formation, and, conse-
quently, the low levels of Snf5 that we’ve identified in human
PTCLs suggest that SNF5 may well act as an important tumor
suppressor in for these cancers.
Blockade of TCR signaling leads to growth arrest of Snf5-deficient lymphoma cells. (A) CD8-enriched cells from Lck-Cre Snf5
as Snf5 WT CD8 T) or tumor-bearing Lck-Cre Snf5
mice (indicated as Snf5
lymphoma) were stimulated for 60 hours in vitro in increasing
numbers in the presence of syngeneic APCs. Proliferation was measured by 3H-Thymidine incorporation. (B) The Snf5-deficient lymphoma cell
line was cultured for 2 days or 5 days in vitro in the presence of 1 μg/ml anti-CD8α or isotype control. Proliferation was measured by 3H-Thymine
incorporation. (C) The Snf5-deficient lymphoma cell line or a control cell line was cultured for 5 days in vitro in the presence or absence of 5 μg/ml
CsA. Proliferation was measured by 3H-Thymine incorporation. (D) The Snf5
lymphoma cell line was cultured for 5 days in vitro in the presence
of 1 μg/ml anti-CD8α and 5 μg/ml CsA. Proliferation was measured by 3H-Thymine incorporation. (E) Tumor-free survival of sublethally irradiated
B6 or MHC class I–deficient K
mice after intravenous transfer of 106 CD8-enriched cells from tumor-bearing Lck-Cre Snf5
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011 3843
Blockade of TCR signaling may be therapeutically useful in peripheral lym-
phomas. Mature PTCLs are diverse morphologically but similar in
that the majority express TCR and carry an extremely poor prog-
nosis (44, 45, 67). Treatment options for these cancers are limited
and tend to be ineffective. The finding that murine Snf5-deficient
lymphomas depend on TCR-MHC interactions led us to ask wheth-
er suppression of TCR signaling may be therapeutically beneficial,
akin to estrogen or androgen blockade in breast and prostate can-
cers. Our results open the possibility that a similar approach may
be effective for mature lymphomas. We have shown here that block-
ade of TCR-MHC interaction effectively reduces proliferation of
established Snf5-deficient lymphoma cells. One of the downstream
events after TCR engagement is activation of calcineurin, which
leads to NFAT nuclear translocation/activation and upregulation of
apoptosis resistance genes (38, 39). Inhibition of the NFAT pathway
by CsA can induce apoptosis, tumor clearance, and substantially
prolonged mouse survival (40). Since the effects of CsA on Snf5-
deficient lymphomas are enhanced in a synergistic fashion by treat-
ment with anti-CD8, administration of CsA in combination with
upstream blockade of the TCR pathway may halt propagation of at
least some PTCLs and is worthy of further investigation.
Snf5 excision PCR
Total genomic DNA extracted from sorted T cells was analyzed for
snf5 excision by PCR. Primers used for the amplification of both
the WT and nonrecombined f loxed allele are Snf5-1, CACCAT-
GCCCCCACCTCCCCTACA, and Snf5-2, CAGGAAAATGGATGCAAC-
TAAGAT. Primers used for the amplification of recombined floxed Snf5
allele are Snf5-3, AGACTGGCCTGATTTGTTTAATATG, and Snf5-4,
TACACGATGAGATCTTGTCTCAAAA. PCR products were detected by
either ethidium bromide staining of agarose gels or quantitative real-
time PCR using iQ SYBR Green Supermix (Bio-Rad).
Mouse models and cell lines
Genotyping was done by PCR analysis of tail DNA as previously reported
(22, 27, 68). The Snf5-deficient lymphoma cell line was generated by 3
sequential adoptive transfers of CD8-enriched splenocytes originally from
tumor-bearing mice into irradiated C57BL/6 host
animals, followed by culture of the cells in DMEM supplemented with 10%
FCS. JAK3AV lymphoma cells were provided by the laboratory of Gary Gil-
liland (Harvard Medical School). Lck-Cre and CD4-Cre mice were obtained
from Christopher Wilson (University of Washington, Seattle, Washington,
USA). Mice were housed under pathogen-free conditions. Animal handling
and experimental procedures were approved by and were in accordance
with institutional requirements for animal care and use of The Harvard
Center for Comparative Medicine (HCCM), Boston, Massachusetts, USA.
Cell preparations and flow cytometry
Data were collected using a FACSCalibur cytometer (Becton Dickinson)
and analyzed with FlowJo software (Tree Star) or isolated by fluorescence-
activated cell sorting on a BD FACSAria sorter (Becton Dickinson). Fluoro-
chrome-conjugated mAbs against CD3ε (145-2C11), CD4 (RM4-5), CD8α
(53-6.7), CD25 (7D4), CD122 (TMβ1), and CD44 (IM7) were purchased
from BD Pharmingen. The Mouse Vβ TCR Screening Panel was purchased
from BD Pharmingen.
Lymphoma cells were purified by negative selection using the mouse CD8
T lymphocyte enrichment set from BD Pharmingen according to manu-
facturer’s protocol (catalog no. 558471) and then were transferred intrave-
nously into sublethally irradiated (4 Gy) B6 (Charles River Laboratories),
(C57BL/6-H-2Kbtm1H-2Dbtm1, NIAID Exchange Program,
NIH:4215; ref. 41), or Il15
In vitro proliferation assay
Enriched CD8 T cells from tumor-bearing mice or littermate controls were
cultured at varying concentrations with 2.5 × 10
irradiated T cell–depleted
SNF5 is expressed at low levels in human PTCLs. Heat map analysis
of SNF5 in 33 PTCLs, 8 immature lymphoblastic T cell leukemia/lym-
phomas, and normal T cells isolated from either peripheral blood (C1
and C2) or reactive lymph nodes. The average expression of SNF5
across all the PTCLs is significantly lower than that in either normal
cells (P < 0.003) or immature lymphoblastic T cell leukemia/lympho-
mas (P < 0.007). Horizontal bars indicate the mean.
3844 The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011
splenocytes and either 1 μg/ml purified anti-CD8α (clone 2.43, National
Cell Culture Center, Minneapolis, Minnesota) or 1 μg/ml purified isotype
control (BD Pharmingen) or at the concentration of 4 × 10
cells per well
in the presence of increasing concentrations of recombinant IL-15 (Pepro-
Tech). The Snf5-deficient lymphoma cell line was cultured in the presence
of 1 μg/ml purified anti-CD8α or 1 μg/ml purified isotype control, with
or without 5 mg/ml CsA (Sigma-Aldrich). Proliferation was measured by
H-Thymidine incorporation (2 μCi/well).
Antibodies for Western blots
Anti-Snf5 antibody (catalog no. 612110) was purchased from BD Biosci-
ences, and anti-actin antibody (mAbcam 8226) was purchased from Abcam.
NFAT1 antibody is a gift from Anjana Rao (Harvard Medical School). Myc
antibody was purchased from Cell Signaling Technology.
RNA purification and RT-PCR
Total RNA was extracted using TRIzol Reagent (Invitrogen) and reverse
transcribed by the Reverse Transcription System (Promega). Gene expres-
sion was normalized to RPS8. Error bars in RNA analysis represent stan-
dard deviations of mean expression or fold changes based on at least 3 inde-
pendent RNA isolations or are indicated in figure legends. Primers used for
real-time PCR reactions are BclxL-F, CCTTCAGGCCTCTCTCTCCT, and
For immunohistochemistry, mouse and human tissues (thymus, spleen, and
lymph nodes) were fixed in 10% formaldehyde, processed, and paraffin embed-
ded; 4-μm sections underwent antigen retrieval (EDTA, pH 8, Invitrogen, cat-
alog no. 00-5500) in a steam pressure cooker (Decloaking Chamber, BioCare
Medical) and were stained with anti-CD3 (rabbit polyclonal antibody, Dako,
catalog no. A0452; 1:250 [for human tissues]; 1:1,000 [for mouse tissues]) or
with anti-SNF5 (BAF47) (mouse monoclonal antibody, BD Transduction
Laboratories, catalog no. 612111; lot no. 04991 [250 μg/ml]; 1:100 [for
human tissues]; 1:200 [for mouse tissues]). In addition, for mouse tissues,
a Mouse-on-Mouse blocking reagent was used (Vector Laboratories, catalog
no. MKB-2213). Diaminobenzidine signals were achieved according to man-
ufacturer’s protocol (EnVision Kit, Dako, catalog no. K4007 and K4011). For
double staining on human tissues, anti-BAF47 antibody was used at 1:100
(diaminobenzidine signal), and anti-CD3 was used at 1:250 with the Vulcan
Fast Red Chromogen Kit 2 (Biocare Medical, catalog no. FR805H). All human
samples used were deidentified, discarded tissues.
Microarray hybridization and data processing
RNA purification. Total RNA was extracted using TRIzol Reagent
(Invitrogen) and reverse transcribed by the Reverse Transcription System
(Promega). The Raw Microarray data were deposited into the GEO data
repository (accession no. GSE23659 [ref. 35] and GSE29732; http://www.
Preprocessing. We processed the raw gene expression values with the robust
multi-array analysis algorithm (69) using GenePattern software (70).
Combining and clustering samples across microarray platforms. As data were
derived from different Affymetrix platforms (Mouse Genome 430A 2.0
and U74v2; see http://www.affymetrix.com for details), we translated the
expression measures from microarray-specific probe sets to platform-
independent representative measures of gene expression. We mapped the
Affymetrix probe set tags to UniGene (71) IDs and discarded UniGene
IDs that were not common to both microarray platforms. With a list of
common UniGene IDs, we translated each data set’s expression values to
scaled expression values that were consistent across the data sets, thereby
making the distributions of gene expression values as similar as possible.
This scaling lessened the experimental and computational artifacts among
data sets. We did this translation with quantile scaling using the following
steps. We calculated the mean expression value of each UniGene across all
samples and stored these mean values in a database and sorted their ranks
in a table. Then, for each sample, for each UniGene, we calculated each
UniGene’s rank, matched that sample-specific rank to the same rank in
the database, and replaced that UniGene’s expression value in the current
sample with the corresponding expression value from the database.
Clustering and classifying. To calculate the similarity among the samples,
we performed k-nearest (KNN) classifier (70), with k = 3 to classify the
samples by using a previously published database of
memory and naive cells as reference samples (23).
GSEA. We created gene sets of the most differentially expressed genes
(with a P value of less than 0.01 and fold change greater than 2) in
either direction between the memory and the naive cells (23). Then we
performed GSEA (30) to measure enrichment of these gene sets in the
samples. For all GSEA runs, we used the signal-to-
noise ratio to rank order the genes based on their ability to distinguish
the reference classes, and we used 1,000 permutations for permutation
testing for statistical significance.
A 2-tailed t test was used to determine significance of difference through-
out this study. Statistics were computed by GraphPad Prism 4. Error bars
in all the figures represent SEM. A value of P < 0.05 was considered statisti-
The authors thank Christopher Wilson for supplying the Lck-Cre
and CD4-Cre strains, Anjana Rao for supplying the NFAT1 anti-
body, and Diana Alvarez Arias and Jonghwan Kim for technical
assistance. The authors are also grateful to John Luckey, Nicho-
las Haining, and members of Cantor and Roberts laboratories for
helpful discussion. C.W.M. Roberts was supported in part by the
Garrett B. Smith Foundation, PHS awards R01CA113794 and
U01-1156106, and a Stand Up to Cancer Innovative Research
Grant from the American Association of Cancer Research. H. Can-
tor was supported in part by a gift from the Gherrin-Gelli Trust.
C.G. Sansam was supported by PHS awards F32CA123776 and
Hope Street Kids foundation. B.G. Wilson was supported by the
Ruth L. Kirschstein National Research Service Award Fellowship 1
F32 CA130312-01A1 from the National Cancer Institute
Received for publication March 28, 2011, and accepted in revised
form August 3, 2011.
Address correspondence to: Charles W.M. Roberts, Dana-Farber
Cancer Institute, Mayer 657, 44 Binney Street, Boston, Massachu-
setts 02115, USA. Phone: 617.632.6497; Fax: 617.582.8096; E-mail:
1. Krivtsov AV, et al. Transformation from commit-
ted progenitor to leukaemia stem cell initiated by
MLL-AF9. Nature. 2006;442(7104):818–822.
2. Park CY, Tseng D, Weissman IL. Cancer stem cell-
directed therapies: recent data from the laboratory
and clinic. Mol Ther. 2009;17(2):219–230.
3. Wong DJ, Liu H, Ridky TW, Cassarino D, Segal E,
Chang HY. Module map of stem cell genes guides
creation of epithelial cancer stem cells. Cell Stem
4. Ben-Porath I, et al. An embryonic stem cell-like
gene expression signature in poorly differenti-
ated aggressive human tumors. Nat Genet. 2008;
5. Hope KJ, Jin L, Dick JE. Acute myeloid leukemia
The Journal of Clinical Investigation http://www.jci.org Volume 121 Number 10 October 2011 3845
originates from a hierarchy of leukemic stem cell
classes that differ in self-renewal capacity. Nat
6. O’Brien CA, Pollett A, Gallinger S, Dick JE. A
human colon cancer cell capable of initiating
tumour growth in immunodeficient mice. Nature.
7. Singh SK, et al. Identification of human brain tumour
initiating cells. Nature. 2004;432(7015):396–401.
8. Bachoo RM, et al. Epidermal growth factor receptor
and Ink4a/Arf: convergent mechanisms governing
terminal differentiation and transformation along
the neural stem cell to astrocyte axis. Cancer Cell.
9. Somervaille TC, Cleary ML. Identification and
characterization of leukemia stem cells in murine
MLL-AF9 acute myeloid leukemia. Cancer Cell.
10. Roberts CW, Orkin SH. The SWI/SNF complex-
chromatin and cancer. Nature Rev Cancer. 2004;
11. Singhal N, et al. Chromatin-remodeling compo-
nents of the BAF complex facilitate reprogram-
ming. Cell. 2010;141(6):943–955.
12. Ho L, et al. An embryonic stem cell chromatin
remodeling complex, esBAF, is essential for embry-
onic stem cell self-renewal and pluripotency. Proc
Natl Acad Sci U S A. 2009;106(13):5181–5186.
13. Biegel JA, Zhou JY, Rorke LB, Stenstrom C, Wain-
wright LM, Fogelgren B. Germ-line and acquired
mutations of INI1 in atypical teratoid and rhab-
doid tumors. Cancer Res. 1999;59(1):74–79.
14. Taylor MD, Gokgoz N, Andrulis IL, Mainprize
TG, Drake JM, Rutka JT. Familial posterior fossa
brain tumors of infancy secondary to germline
mutation of the hSNF5 gene. Am J Hum Genet.
15. Versteege I, et al. Truncating mutations of hSNF5/
INI1 in aggressive paediatric cancer. Nature. 1998;
16. Kreiger PA, et al. Loss of INI1 expression defines
a unique subset of pediatric undifferentiated soft
tissue sarcomas. Mod Pathol. 2009;22(1):142–150.
17. Xia W, et al. BAF180 is a critical regulator of p21
induction and a tumor suppressor mutated in
breast cancer. Cancer Res. 2008;68(6):1667–1674.
18. Varela I, et al. Exome sequencing identifies frequent
mutation of the SWI/SNF complex gene PBRM1 in
renal carcinoma. Nature. 2011;469(7331):539–542.
19. Soulier J, et al. A complex pattern of recurrent
chromosomal losses and gains in T-cell prolym-
phocytic leukemia. Genes Chromosomes Cancer. 2001;
20. Bug S, et al. Recurrent loss, but lack of mutations, of
the SMARCB1 tumor suppressor gene in T-cell pro-
lymphocytic leukemia with TCL1A-TCRAD juxta-
position. Cancer Genet Cytogenet. 2009;192(1):44–47.
21. Wang X, et al. Oncogenesis caused by loss of the SNF5
tumor suppressor is dependent on activity of BRG1,
the ATPase of the SWI/SNF chromatin remodeling
complex. Cancer Res. 2009;69(20):8094–8101.
22. Roberts CW, Leroux MM, Fleming MD, Orkin SH.
Highly penetrant, rapid tumorigenesis through
conditional inversion of the tumor suppressor
gene Snf5. Cancer Cell. 2002;2(5):415–425.
23. Luckey CJ, Bhattacharya D, Goldrath AW, Weiss-
man IL, Benoist C, Mathis D. Memory T and mem-
ory B cells share a transcriptional program of self-
renewal with long-term hematopoietic stem cells.
Proc Natl Acad Sci USA. 2006;103(9):3304–3309.
24. Bonnet D, Dick JE. Human acute myeloid leuke-
mia is organized as a hierarchy that originates
from a primitive hematopoietic cell. Nat Med. 1997;
25. Dick JE. Stem cell concepts renew cancer research.
26. Weissman IL, McGrath MS. Retrovirus lymphoma-
genesis: relationship of normal immune receptors
to malignant cell proliferation. Curr Top Microbiol
27. Lee PP, et al. A critical role for Dnmt1 and DNA
methylation in T cell development, function, and
survival. Immunity. 2001;15(5):763–774.
28. Boyman O, Cho JH, Tan JT, Surh CD, Sprent J. A
major histocompatibility complex class I-depen-
dent subset of memory phenotype CD8+ cells.
J Exp Med. 2006;203(7):1817–1825.
29. Boman BM, Wicha MS. Cancer stem cells: a step
toward the cure. J Clin Oncol. 2008;26(17):2795–2799.
30. Subramanian A, et al. Gene set enrichment analy-
sis: a knowledge-based approach for interpreting
genome-wide expression profiles. Proc Natl Acad Sci
U S A. 2005;102(43):15545–15550.
31. Swanson BJ, Murakami M, Mitchell TC, Kappler
J, Marrack P. RANTES production by memory
phenotype T cells is controlled by a posttranscrip-
tional, TCR-dependent process. Immunity. 2002;
32. Walzer T, Marcais A, Saltel F, Bella C, Jurdic P, Mar-
vel J. Cutting edge: immediate RANTES secretion
by resting memory CD8 T cells following antigenic
stimulation. J Immunol. 2003;170(4):1615–1619.
33. Chi TH, et al. Sequential roles of Brg, the ATPase
subunit of BAF chromatin remodeling complex-
es, in thymocyte development. Immunity. 2003;
34. Chi TH, et al. Reciprocal regulation of CD4/CD8
expression by SWI/SNF-like BAF complexes.
35. Wilson BG, et al. Epigenetic antagonism between
polycomb and SWI/SNF complexes during oncogen-
ic transformation. Cancer Cell. 2010;18(4):316–328.
36. Kim J, et al. A Myc network accounts for similarities
between embryonic stem and cancer cell transcrip-
tion programs. Cell. 2010;143(2):313–324.
37. Isakoff MS, et al. Inactivation of the Snf5 tumor sup
pressor stimulates cell cycle progression and cooper-
ates with p53 loss in oncogenic transformation. Proc
Natl Acad Sci USA. 2005;102(49):17745–17750.
38. Macian F. NFAT proteins: key regulators of
T-cell development and function. Nat Rev Immunol.
39. Rao A, Luo C, Hogan PG. Transcription factors of
the NFAT family: regulation and function. Annu
Rev Immunol. 1997;15:707–747.
40. Medyouf H, et al. Targeting calcineurin activation
as a therapeutic strategy for T-cell acute lympho-
blastic leukemia. Nat Med. 2007;13(6):736–741.
41. Perarnau B, et al. Single H2Kb, H2Db and double
H2KbDb knockout mice: peripheral CD8+ T cell
repertoire and anti-lymphocytic choriomeningi-
tis virus cytolytic responses. Eur J Immunol. 1999;
42. Martinez-Delgado B, et al. Expression profiling of
T-cell lymphomas differentiates peripheral and lym-
phoblastic lymphomas and defines survival related
genes. Clin Cancer Res. 2004;10(15):4971–4982.
43. Kuppers R, Klein U, Hansmann ML, Rajewsky K.
Cellular origin of human B-cell lymphomas. N Engl
J Med. 1999;341(20):1520–1529.
44. Rudiger T, Geissinger E, Muller-Hermelink HK.
‘Normal counterparts’ of nodal peripheral T-cell
lymphoma. Hematol Oncol. 2006;24(4):175–180.
45. Savage KJ. Peripheral T-cell lymphomas. Blood Rev.
46. Waller EK, et al. Growth of primary T-cell non-
Hodgkin’s lymphomata in SCID-hu mice: require-
ment for a human lymphoid microenvironment.
47. Oruetxebarria I, et al. P16INK4a is required for
hSNF5 chromatin remodeler-induced cellular
senescence in malignant rhabdoid tumor cells.
J Biol Chem. 2004;279(5):3807–3816.
48. Kia SK, Gorski MM, Giannakopoulos S, Verrijzer
CP. SWI/SNF mediates polycomb eviction and epi-
genetic reprogramming of the INK4b-ARF-INK4a
locus. Mol Cell Biol. 2008;28(10):3457–3464.
49. Jagani Z, et al. Loss of the tumor suppressor Snf5
leads to aberrant activation of the Hedgehog-Gli
pathway. Nat Med. 2010;16(12):1429–1433.
50. Masopust D, Vezys V, Wherry EJ, Ahmed R. A
brief history of CD8 T cells. Eur J Immunol. 2007;
37 suppl 1:S103–S110.
51. Prlic M, Williams MA, Bevan MJ. Requirements for
CD8 T-cell priming, memory generation and main-
tenance. Curr Opin Immunol. 2007;19(3):315–319.
52. Lanzavecchia A, Sallusto F. Understanding the
generation and function of memory T cell subsets.
Curr Opin Immunol. 2005;17(3):326–332.
53. Surh CD, Sprent J. Homeostasis of naive and mem
ory T cells. Immunity. 2008;29(6):848–862.
54. Chambers SM, et al. Hematopoietic fingerprints:
an expression database of stem cells and their prog-
eny. Cell Stem Cell. 2007;1(5):578–591.
55. Rothenberg EV, Pant R. Origins of lymphocyte
developmental programs: transcription factor evi-
dence. Semin Immunol. 2004;16(4):227–238.
56. Biganzoli L, Licitra S, Claudino W, Pestrin M, Leo
AD. Clinical decision making in breast cancer:
TAM and aromatase inhibitors for older patients
-- a jungle? Eur J Cancer. 2007;43(15):2270–2278.
57. Boccon-Gibod L, Hammerer P, Madersbacher S,
Mottet N, Prayer-Galetti T, Tunn U. The role of
intermittent androgen deprivation in prostate
cancer. BJU Int. 2007;100(4):738–743.
58. Utsumi T, Kobayashi N, Hanada H. Recent perspec-
tives of endocrine therapy for breast cancer. Breast
59. Clambey ET, White J, Kappler JW, Marrack P.
Identification of two major types of age-associ-
ated CD8 clonal expansions with highly diver-
gent properties. Proc Natl Acad Sci U S A. 2008;
60. Piccaluga PP, et al. Gene expression analysis of
peripheral T cell lymphoma, unspecified, reveals
distinct profiles and new potential therapeutic tar-
gets. J Clin Invest. 2007;117(3):823–834.
61. Pechloff K, et al. The fusion kinase ITK-SYK mim
ics a T cell receptor signal and drives oncogenesis
in conditional mouse models of peripheral T cell
lymphoma. J Exp Med. 2010;207(5):1031–1044.
62. Hulsebos TJ, Plomp AS, Wolterman RA, Robanus-
Maandag EC, Baas F, Wesseling P. Germline muta-
tion of INI1/SMARCB1 in familial schwannomato-
sis. Am J Hum Genet. 2007;80(4):805–810.
63. Boyd C, Smith MJ, Kluwe L, Balogh A, Maccollin
M, Plotkin SR. Alterations in the SMARCB1 (INI1)
tumor suppressor gene in familial schwannomato-
sis. Clin Genet. 2008;74(4):358–366.
64. Holleman A, et al. Gene-expression patterns in
drug-resistant acute lymphoblastic leukemia cells
and response to treatment. N Engl J Med. 2004;
65. Jones S, et al. Frequent mutations of chromatin
remodeling gene ARID1A in ovarian clear cell car-
cinoma. Science. 2010;330(6001):228–231.
66. Wiegand KC, et al. ARID1A mutations in endome-
triosis-associated ovarian carcinomas. N Engl J Med
67. de Leval L, Bisig B, Thielen C, Boniver J, Gaulard P.
Molecular classification of T-cell lymphomas. Crit
Rev Oncol Hematol. 2009;72(2):125–143.
68. de Boer J, et al. Transgenic mice with hematopoi-
etic and lymphoid specific expression of Cre. Eur J
69. Irizarry RA, et al. Exploration, normalization, and
summaries of high density oligonucleotide array
probe level data. Biostatistics. 2003;4(2):249–264.
70. Reich M, Liefeld T, Gould J, Lerner J, Tamayo
P, Mesirov JP. GenePattern 2.0. Nat Genet.
71. Wheeler DL, et al. Database resources of the
National Center for Biotechnology. Nucleic Acids