Heterozygosity for Roquin^san leads to angioimmunoblastic T-cell lymphoma-like tumors in mice

Angioimmunoblastic T-cell lymphoma (AITL) is a distinct subtype of peripheral T-cell lymphoma, displaying autoimmune features and poor prognosis.^1–3  Unlike other tumors in which morbidity and mortality occur as a consequence of complications of tumor growth, AITL-associated mortality is thought to be a consequence of profound immune dysregulation.^4,5  AITL disease outcome is poor, with an overall 7-year survival rate of 30%,³  and patients usually do not respond well to cytotoxic chemotherapeutic treatment.

AITL is more common in elderly patients, with a median age of 64 years.⁶  Patients commonly present with lymphadenopathy, hepatosplenomegaly, skin rash, fever, hemolytic anemia, and hypergammaglobulinemia.^7,8  Histologic features of AITL typically include effacement of the lymph node architecture, prominent arborization of endothelial venules, and an expanded extrafollicular meshwork of follicular dendritic cells (FDCs).

The neoplastic T cells account for only a small fraction of the lymphoid infiltrate and are admixed with a large number of reactive immune cell types, including small lymphocytes, eosinophils, histiocytes, plasma cells, and abnormal B cells often infected with EBV.³  Indeed, close to 90% of the gene expression signature is contributed by nonneoplastic cells⁹  and characterized by expression of many genes associated with the tumor microenvironment, including genes expressed by B cells, FDCs, chemokines, the extracellular matrix, and the vascular system.⁹  In addition, AITL has a gene expression profile analogous to that of activated CD4⁺ helper T cells but not of CD8⁺ T cells.¹⁰  It is enriched for expression of follicular helper T (T_FH) cell signature genes, including expression of CXCL13, BCL6, and PD1^9,10  (17284527; 16625095). These findings, together with the identification of T_FH markers in tumor tissues sections by immunohistochemistry,^11–14  have suggested that T_FH cells are the neoplastic cells in AITL.¹⁰  However, the molecular mechanisms that drive T_FH neoplastic transformation remain largely unknown.

To date, there are no mouse models available of this human disease, which hampers both elucidation of pathogenic pathways and the possibility of undertaking preclinical trials using compounds targeting T_FH cells. We have previously reported the characterization of the sanroque mouse strain that bears a homozygous point mutation in the Roquin/Rc3h1 gene,¹⁵  and the resulting allele is designated Roquin^san. ROQUIN has previously been shown to bind Icos mRNA through its ROQ domain and regulate inducible T-cell costimulator (ICOS) mRNA stability.^16–18  Mutant ROQUIN encoded by Roquin^san binds Icos mRNA with higher affinity, leading to slower mRNA decay. In Roquin^san/san mice, aberrant expression of ICOS on CD4⁺ T cells contributes to an excessive accumulation of T_FH cells,^15,17,19  which in turn causes lupus-like disease characterized by lymphadenopathy, splenomegaly, hypergammaglobulinemia, and the development of glomerulonephritis.¹⁵  Interestingly, mice heterozygous for the Roquin “san” allele (Roquin^san/+) also have dysregulated T_FH cells but do not develop the full-blown lupus-like autoimmunity. Instead, a proportion of mice develop asymmetric lymphadenopathy.

Here we report that enlarged lymph nodes from Roquin^san/+ mice display T-cell oligoclonality and many histopathologic features consistent with human AITL. Furthermore, tumor development was found to be dependent on T_FH cells, suggesting that, as in human AITL, T_FH cells may be the neoplastic driver of disease. These findings suggest that Roquin^san/+ mice may provide a good model for human AITL and thereby promote further understanding of disease mechanisms and help in the discovery of novel therapeutic targets.

Roquin^san/+, Roquin^san/+Sap^−/−, Roquin^san/+Cd28^−/−, and Roquin^san/+Icos^−/− mice were bred and maintained under pathogen-free conditions at the Animal Bioscience Services facility. All experimental procedures were carried out in compliance with protocols approved by the Australian National University Animal Ethics and Experimentation Committee. Lymph nodes examined for tumor development were cervical nodes, axillary nodes, brachial nodes, and inguinal nodes. Size of lymph nodes was visually inspected and classified as tumorous when they were at least 5 times the size of normal lymph nodes from wild-type mice. Mice were characterized as tumorous if there was at least 1 enlarged lymph node among the aforementioned inspected nodes.

All antibodies and streptavidin conjugates used for flow cytometry were from BD Biosciences unless otherwise indicated: anti–mouse B220 PECy7, B220 PerCP, CD4 APC Cy7, CD4 APC, CD8 FITC, CXCR5-biotin, GL-7 FITC, FAS PE, and PD-1 PE (eBioscience), mouse anti-Bcl6 A647, streptavidin PerCP Cy5.5, and streptavidin PE Cy7. For immunohistochemistry, the primary antibodies used were goat anti–mouse CD3ϵ (Santa Cruz Biotechnology), goat anti–mouse Pax-5 (Santa Cruz Biotechnology), and rat anti–mouse F4/80 (BMA Biomedicals).

ELISA was performed to analyze total IgG levels in mouse serum as previously described.²⁰ 

Hematoxylin and eosin-stained sections from a series of 19 Roquin^san/+ and 4 Roquin^san/san formalin-fixed, paraffin-embedded lymph nodes were morphologically reviewed. Five lymph nodes belonging to Roquin^+/+ littermates were examined as controls. Immunohistochemistry for mouse CD3ϵ, Pax5, and F4/80 was performed on formalin-fixed, paraffin-embedded sections as follows: tissue sections were deparaffinized, rehydrated, and treated for antigen retrieval. After quenching of endogenous peroxidase and blocking in normal serum, tissues were incubated with primary antibody overnight at 4°C, followed by incubation with biotinylated secondary antibody. Specific interactions were visualized using the Envision System (Dako North America) following the manufacturer's instructions. Slides were counterstained with hematoxylin, dehydrated, and mounted. Slides were visualized on a Leica DMD108 microsystem at 25°C and acquired with LAS-DMD Version 1.3.1 software (Leica).

Single-cell suspensions of lymph nodes were prepared in FACS buffer (PBS/2% BSA/0.05% NaN₃) by sieving and gently pipetting through 70-μm nylon mesh filters. After red blood cell lysis, 3 × 10⁶ cells were incubated with each antibody or conjugate layer for 30 to 60 minutes on ice. Samples were run on a FACSCalibur (BD Biosciences). Analysis was performed using FlowJo Version 7.2.5 (TreeStar).

PCR was performed to analyze the clonal expansion of T and B cells.²¹  DNA was extracted from paraffin sections, and T-cell and B-cell clonal expansion was detected by analysis of TCR-β, TCR-γ, and IgH gene rearrangement. TCR-β gene clonality was assayed at V-DJ and D-J rearrangements in 4 different PCR reactions (Vβ-Jβ1, Vβ-Jβ2, dβ1-Jβ1, Dβ2-Jβ2) as previously described.²²  The TCR-γ gene clonality was assayed at V-J rearrangements in 1 PCR (Vγ4-Jγ1) using Vγ4 and Jγ1 consensus primers previously described by Kawamoto et al.²³  The IgH gene clonality at D-J rearrangements in 1 PCR (DSF-JH4), using DSF and JH4 consensus primers previously described by Kawamoto et al.²³  PCR primers (Jβ1, Jβ2, Jγ1, and JH4), were labeled at the 5′end with 6-Fam for GeneScanning of the PCR products. The fluorochrome-labeled single-strand (denatured) PCR products were analyzed by capillary electrophoresis using the ABI PRISM 3700 Genetic Analyser and Gene Scan Version 1.2 software (Applied Biosystems).²⁴  Appropriate positive and negative controls were included in all experiments.

For repertoire analyses of purified T cells, T_FH (CD4⁺CXCR5⁺PD-1^hi) and non-T_FH (CD4⁺CXCR5⁻PD-1^lo) cell subsets were purified by FACS sorting on a FACSDIVA (BD Biosciences), and RNA was isolated using an RNAqueous-Micro Kit (Ambion) according to the manufacturer's instructions. TCRV-β repertoires were measured using real-time PCR on the Biorad iCycler. One PCR for each Vβ family (22 reactions) was performed in triplicate for each sample to cover the repertoire. Threshold cycles were converted to a percentage mRNA signal presuming a 100% efficiency reaction. For clonal analysis within each Vβ family, the PCR product was amplified with standard PCR using FAM-labeled Vβ constant region reverse primer for 6 to 10 cycles and run on a high-resolution sequencing gel to delineate CDR3 spectratypes.

Using flow cytometry, 1, 10, or 100 T_FH and naive T cells were sorted directly into the wells of a 96-well plate and digested with proteinase K (QIAGEN) to isolate DNA. A primary PCR was performed for the region of Roquin approximately 200 bp either side of the san mutation. A secondary PCR was performed using fluorescently labeled primers to detect the Roquin^san and Roquin⁺ alleles, as regularly used to genotype the sanroque mouse strain. Primer sequences are available on request.

Statistical analysis was performed in Prism Version 5 software (GraphPad Software) with a Mann-Whitney test unless otherwise stated.

We have previously reported that Roquin^san/san mice develop generalized lymphadenopathy, splenomegaly, and T_FH dysregulation.¹⁵  In these mice, accumulation of T_FH cells causes a lupus-like pathology.²⁰  Initial inspection of Roquin^san/san lymph node histopathology showed large polymorphic infiltrates and the presence of atypical T cells characteristic of human AITL (Figure 1). Given that reactive nodes commonly accompany systemic autoimmunity and this may confound the diagnosis of AITL, we investigated heterozygous Roquin^san/+ mice instead: initial observations suggested that a proportion Roquin^san/+ mice developed lymphadenopthy in the absence of overt autoimmunity.

To assess lymph node pathology in Roquin^san/+ mice, groups of mice were killed at monthly intervals from 4 months of age. All the major lymph nodes, including cervical, axillary, brachial, and inguinal, were examined. None of the Roquin^san/+ mice developed the generalized symmetric lymph node enlargement that is observed in 100% of Roquin^san/san mice by 8 weeks of age. Instead, 53% of mice (100 of 188 mice studied) developed 1 to 4 enlarged lymph nodes, whereas other lymph nodes remained normal (Figure 2A; Table 1). This lymphadenopathy was not lethal, with affected mice able to live to more than 1 year (data not shown). From here on, these enlarged lymph nodes are referred to as tumor lymph nodes.

Cellularity was increased by 50- to 150-fold in tumor lymph nodes. Generally, a tumor lymph node consisted of 5.0 × 10⁷ to 1.5 × 10⁸ cells, whereas the normal lymph node in the same mouse or in Roquin^san/+ mice without tumors consisted of approximately 1.0 × 10⁶ cells (data not shown). The prevalence of tumors was 1.6 times higher in female mice (65%) than in male mice (41%; P = .001) regardless of age (Table 1). The sex distribution ratio for human AITL cases has been reported to be 1:1.^25,26 

Although the age of onset could not be accurately determined because of the tumors only being externally palpable once they had reached a considerable size, 4 of 11 mice (36%) investigated at 4 to 5 months of age had already developed tumors, suggesting some tumors are likely to develop before 4 months of age.

In addition to lymphadenopathy, many AITL patients present with splenomegaly.^7,8 Roquin^san/+ mice displayed a slight but significant (P < .05) increase in spleen weight compared with control Roquin^+/+ mice (Figure 2B-C). No significant difference in spleen size was observed between Roquin^san/+ mice with or without tumors, suggesting that splenomegaly does not correlate with, and may precede, lymphadenopathy. This is consistent with observations that nontumor lymph nodes from Roquin^san/+ mice are often larger than lymph nodes from Roquin^+/+ control mice (data not shown).

Hypergammaglobulinemia is another frequent finding in AITL, with 50% of patients displaying increased IgG in the serum.^3,6  Interestingly, although 4-month-old Roquin^san/+ mice do not develop overt lupus-like autoimmunity, Roquin^san/+ mice, both with and without tumors, developed hypergammaglobulinemia, showing that total serum IgG levels increased 2- to 3-fold above Roquin^+/+ mice (Figure 2D). A moderate (1.5-fold) but statistically significant (P < .05) increase in total serum IgG was observed in Roquin^san/+ mice that had developed tumors compared with mice of the same genetic background without tumors. Investigation of serum IgG levels over time suggested that, even at 6 weeks of age, Roquin^san/+ mice exhibit increased serum IgG titers compared with age-matched controls; however, this did not reach significance until 15 weeks of age (supplemental Figure 1, available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

Diagnosis of AITL is based on histologic analysis. The 5 main diagnostic criteria consist of: (1) effaced lymph node architecture, (2) prominent arborization of epithelioid venules, (3) extrafollicular meshwork of FDCs, (4) presence of atypical T cells with T_FH phenotype, and (5) scattered large CD20⁺ B cells.³  Histopathologic examination of hematoxylin and eosin-stained sections of enlarged lymph nodes from Roquin^san/+ mice showed features reminiscent of AITL, including effacement of the nodal architecture, prominent vascularization, atypical T cells, and large B cells (Figure 3).

In addition to these features, there was a striking increase in the number of macrophages within the paracortex as shown by F4/80 staining with occasional foci of myeloid metaplasia (Figure 3B). Dense aggregates of small PAX5-positive lymphocytes were visible all over the parenchyma forming follicles without germinal centers (GCs), although a few large pale lymphoid cells were always found among them (Figure 3C-D). The same reactive B-blasts were also found in the interfollicular areas of the lymph node. Small clusters of mature plasma cells were also seen. A polymorphic infiltrate composed of small- to medium-sized proliferating CD3-positive lymphocytes was found in the interfollicular compartment, some of them showing cytologic atypica and increased size (Figure 3E). Moreover, occasionally T cells formed rosettes around large blasts (Figure 3F), which have been reported in AITL. Sinusoidal dilatation was also evident, as well as vessels filled with T cells. A prominent feature present in most AITL cases, but absent in the tumors of Roquin^san/+ mice, is prominent expansion of the FDC network,^3,7,8,27  as determined by an absence of CD21⁺ dendritic cells in lymph node sections (data not shown). Together, this histologic appearance is highly reminiscent of AITL.

In contrast, histologic examination of nonenlarged lymph nodes from tumor-bearing Roquin^san/+ mice revealed that nodal architecture was preserved (Figure 3G). These lymph nodes displayed both primary and secondary follicles with reactive GCs as well as a marked increase of mature plasma cells in the interfollicular area (Figure 3H).

In AITL, the neoplastic T cells account for only a small fraction of the lymphoid infiltrate and are admixed with a large number of reactive immune cell types, including small lymphocytes, eosinophils, histiocytes, plasma cells, and large B cells. To quantify tumor composition, we enumerated T, B, and myeloid cells subsets by flow cytometry.

There was a significantly increased proportion of B cells in tumor lymph nodes compared with normal lymph nodes from either Roquin^san/+ or Roquin^+/+ mice (Figure 4A). Analysis of T_FH cells revealed that the frequency of these cells was increased in “normal” Roquin^san/+ lymph nodes compared with wild-type Roquin^+/+ mice (Figure 4B,D-E), and this was accompanied by a significant increase in frequency of GC B cells (Figure 4C). In contrast, the tumor lymph node displayed frequencies similar to Roquin^+/+ mice. Comparing an enlarged tumor node with its contralateral nonenlarged lymph node, although the proportion of T_FH cells identified as either PD-1^hiCXCR5⁺ (Figure 4B; P < .001) or PD-1^hiBcl-6⁺ (Figure 4D-E; not significant) within CD4⁺ T cells was decreased, the total T_FH-cell number was significantly increased (Figure 4F; P < .05). This supports the hypothesis that, in tumor lymph nodes, an initial expansion of T_FH cells is diluted because of a large reactive infiltrate resulting in lymphadenopathy. This is also consistent with the observation that T_FH cells contribute to only a small fraction of the cellularity of human AITL-affected lymph nodes.^9,10  Analysis of expression of the proliferation marker, Ki-67, by flow cytometry indicated that approximately 30% the T_FH cells in the tumor lymph nodes of Roquin^san/+ mice were proliferating (Figure 4G). This was significantly higher (P < .05) than the proportion of proliferating T_FH cells in either the nontumor lymph node from tumor-bearing Roquin^san/+ mice or lymph nodes from Roquin^+/+ mice. These results are also consistent with the abnormal CD3ϵ⁺ proliferating cells observed in the histologic sections (Figure 3E). In contrast, GC B cells in the tumor lymph node appeared to be less proliferative compared with both the Roquin^san/+ and Roquin^+/+ controls (Figure 4H), although the differences were not significant.

An increased frequency of myeloid cells and dendritic cells was also detected by flow cytometry (Figure 5A), with monocyte-derived dendritic cells being the most dysregulated subset with a 2- to 3-fold expansion (P = .0535) in the tumor lymph node compared with wild-type mice (Figure 5B). Monocyte-derived dendritic cells were expanded to an intermediate level (P = .1099) in the contralateral nontumor lymph nodes (Figure 5A-B). In contrast, CD8⁺ and CD8⁻ dendritic cells were proportionally decreased, significantly and nonsignificantly, respectively, in both tumor and nontumor lymph nodes from Roquin^san/+ mice (Figure 5B).

Together, these data suggest that dysregulation of T_FH cells and subsequent GC expansion are an obligatory consequence of the Roquin^san mutation and preclude tumor development. As the lymph node disorganization progresses, it is probable that GCs collapse and neoplastic T_FH cells are diluted amid large numbers of reactive cells. This scenario is consistent with the histologic data showing hyperplasic expansion of GCs in the nonenlarged lymph nodes leading to complete effacement of the nodal architecture in enlarged tumor lymph nodes.

Expansion of 1 or several T-cell clones is a common feature of AITL tumor development.^28–30  We examined clonality of both T and B cells in tumor lymph nodes by PCR amplification of TCR-β and IgH genes, respectively. Clonal rearrangements were found in the TCR-β gene in 12 of 15 cases (Table 2). In contrast, a clonal peak of the IgH gene was found in only 1 case (Table 2). Five cases were also examined for clonal rearrangements in the TCR-γ gene; however, no PCR products were amplified (data not shown).

To determine whether the observed clonality corresponded to expanded T_FH-cell clones, we performed additional studies on T_FH (CD4⁺CXC5⁺PD-1^hi) and CD4⁺ non-T_FH cells (CD4⁺CXCR5⁻PD-1^lo) purified by flow cytometry. T_FH cells from Roquin^san/+ mice with and without tumors were found to be more clonal than T_FH cells from Roquin^+/+ mice (Figure 6A; Table 3). Furthermore, T_FH cells from Roquin^san/+ mice exhibited a more restricted TCR-βV repertoire with the total number of TCR-βV chains detected in each sample being much less than T_FH cells from Roquin^+/+ mice (Table 3).

Of particular interest, T_FH cells also exhibited more clonality than CD4⁺ non-T_FH cells from the same lymph nodes (Figure 6B; Table 3). Importantly, the clonal peaks detected in the T_FH samples often (68%) composed at least 5% of the total TCR-βV repertoire, and each sample had at least 1 clone that accounted for more than 10% of the repertoire, suggesting that this clone is highly overexpressed. In contrast, in the non-T_FH samples, only 19% of clonal peaks had a frequency more than or equal to 5%, and only 1 of 5 samples had a clone that accounted for more than 10% of the repertoire. Together, these data indicate that lymph nodes from Roquin^san/+ mice typically display expanded T_FH-cell clones, and this may possibly precede tumor development.

To further investigate a possible role of T_FH cells in driving or maintaining tumors in Roquin^san/+ mice, we introduced genetic manipulations that reduce the number and/or function of T_FH cells, and evaluated the tumor incidence in these mice. Reduction or deficiencies in CD28, SAP, and ICOS¹⁹  have already been shown to decrease the numbers of T_FH cells in Roquin^san/san mice. SAP deficiency in Roquin^san/san mice results in a 4-fold reduction in T_FH cells and abrogates spontaneous GC formation but does not alter T_H1 or T_H2 subsets.²⁰  CD28 deficiency corrects excessive T_FH accumulation in Roquin^san/san mice to levels comparable with Roquin^+/+ mice.²⁰  Similarly, ICOS hemizygosity also reduces T_FH-cell numbers.¹⁷ 

Tumor incidence in Roquin^san/+ mice that were also deficient in Cd28, Sap, or Icos was significantly reduced compared with wild-type Roquin^san/+ mice (Figure 7). Strikingly, mice lacking SAP were completely protected from tumor development. This is particularly significant because, of all 3 molecules, SAP deficiency has the most selective effect on T_FH-cell numbers. These results strongly suggest that T_FH cells play a crucial role in tumor development and support the idea that neoplastic transformation of T_FH cells in Roquin^san/+ mice results in AITL-like disease.

We next asked whether loss of heterozygosity (LOH) within T_FH cells so that they only express the mutant form of ROQUIN could explain the occurrence of AITL-like disease in only approximately 50% of Roquin^san/+ mice. To test this, we generated gDNA samples from low numbers of FACS-purified T_FH cells (10-100) from tumor and nontumor lymph nodes and investigated LOH using nested PCR. All samples remained heterozygous for the Roquin^san allele. To exclude the possibility that LOH was occurring in only a subset of T_FH cells and thus masked by the pooling of multiple cells, we extended the study to examine LOH in single cells. Although we were able to detect loss of either the Roquin^san or the Roquin⁺ allele at the single-cell level, this phenonomen was not exclusive to or increased in tumor-derived T_FH cells compared with either naive T cells or T_FH cells from Roquin^san/+ nontumor controls (data not shown). Thus, we conclude that, although LOH may occur in individual lymphocytes in Roquin^san/+ mice, this is not causally related with the development of AITL-like disease.

AITL is an aggressive T-cell malignancy with characteristic autoimmune-like manifestations related to B-cell reactivity that make it a complex disease with no appropriate mouse model to date. Here we have described that heterozygosity for Roquin^san recapitulates most of the clinical, histologic, and cellular features associated with AITL, including lymphadenopathy, hypergammaglobulinemia, increased T_FH-cell numbers,^9,10  and clonal expansion of T cells.^28–30  Thus, this work establishes Roquin^san/+ mice as a potentially useful mouse model of this peculiar type of peripheral T-cell lymphoma.

Notably, AITL is characterized by a large histopathologic spectrum with 3 patterns described: hyperplasic GCs and no increase in FDC (pattern 1), depleted follicles and loss of normal lymph node architecture (pattern 2), and complete effacement of lymph node architecture, with prominent FDCs (pattern 3) in full-blown AITL. All 3 patterns display polymorphic infiltrates and vascular proliferation. The majority of patients present with pattern 2 or 3 (consequently known as classic AITL).^11,31–33  Although these different morphologic variants do not appear to influence survival, progression from pattern 1 to pattern 2 or 3 is well documented, suggesting that pattern 1 is an early phase of the disease.^11,31,32 Roquin^san/+ enlarged lymph nodes, which do not contain hyperplasic GCs, share features with classic AITL (pattern 2 or 3). By contrast, hyperplasic GCs were a frequent feature of contralateral “nontumor” lymph nodes in the same mice, suggesting that these may already be displaying the early form of disease (pattern 1). The proportion of Ki67⁺ T_FH cells increased from Roquin^+/+ to nontumor lymph nodes and again in tumor lymph nodes from the same mouse, suggesting that increased proliferation correlates with tumor development.

Attempts to transplant AITL-like tumors were unsuccessful consistent with the reported poor transplantability of most peripheral T-cell tumors and B-cell lymphomas of GC origin.^34–37  The difficulty in transplanting these tumors probably relates to the specialized environment of the GC, containing unique niches and stromal elements that may be required to support tumor growth^1–3  and to the fact that terminally differentiated T_FH cells do not recirculate because of the down-regulation of CCR7.³⁸ 

The absence of an expanded FDC network was arguably the one discordant feature between the tumors of Roquin^san/+ mice and AITL. Most AITL patients are infected by EBV, which does not infect mice. It is therefore possible that the FDC expansion is a consequence of EBV infection.³⁹  Although EBV infection is normally associated with B cells, FDCs express high levels of CD21, the surface receptor for this virus,⁴⁰  and EBV infection has been associated with several FDC tumors.^41,42  The role of EBV in FDC expansion in AITL remains to be determined. Alternatively, it is possible that EBV infection and/or FDC expansion in humans may drive/support T_FH growth and/or survival, providing the right environment for neoplasia to develop. Indeed, in human AITL samples, neoplastic T cells are intimately associated with FDC networks.⁴³  In contrast, the cell-autonomous accumulation of T_FH cells in the presence of the Roquin “san” mutation may bypass the requirement of an expanded FDC network.

Increased clonality among T_FH cells compared with non-T_FH cells from the same lymph node, even in nontumor lymph nodes from Roquin^san/+ mice, is an intriguing finding. It suggests that ongoing antigen-specific responses to either self or foreign antigens are probably driving expansion of nonmalignant T_FH clones, some of which may subsequently become neoplastic, thus mimicking early stages of AITL.

Our data provide experimental evidence in support of the notion emerging from histopathologic studies that T_FH cells may be the cellular counterparts of AITL.^9,13,44  Indeed, blockade in T_FH-cell development completely prevented AITL-like tumor development. AITL patients do not respond well to cytotoxic chemotherapeutic treatment,^4,5  and alternative strategies are needed to improve patient outcome.^3,6  Novel therapies that neutralize or deplete T_FH cells by targeting ICOS, PD-1, SAP, or the PI3K signaling pathway^45,46  may prove to be more specific and effective and improve prognosis. Because AITL neoplastic cells represent only a small proportion of the tumor,^3,7,8,11  conventional T-cell-lymphoma mouse models,^47,48  in which disease is the result of uncontrolled malignant cell proliferation, may not be useful in testing new therapeutic compounds. In contrast, the Roquin^san/+ model of AITL should be useful for preclinical testing of these or other novel therapeutic compounds.

In addition, Roquin^san/+ mice may help further understanding of AITL disease development mechanisms and pathways. Despite all Roquin^san/+ mice exhibiting dysregulated T_FH cells, only 53% of mice develop AITL-like disease. This suggests that additional, possibly mutagenic, events are required for neoplastic transformation. Dissecting the pathway(s) that lead to lymphadenopathy and AITL-like disease in Roquin^san/+ mice may thereby provide important insights into human disease development and progression.

The authors thank M. Townsend for cryosectioning, A. Prins for histology, Dr H. Vohra and M. Devoy for cell sorting, and Debbie Howard for technical assistance.

This work was supported by a Viertel Senior Medical Research Fellowship and National Health and Medical Research Council program and project grants (C.G.V.) and a National Health and Medical Research Council Overseas Biomedical Fellowship (J.I.E.).

Contribution: J.I.E. performed and analyzed experiments and wrote the manuscript; T.C., S.-M.R.-P., J.L.M., X.H., M.N.-G., J.F.G., and S.M.-M. and performed and analyzed experiments; M.-H.D.-L. and P.G. provided critical analysis of the manuscript and clinical interpretation of the data; G.W. designed experiments and interpreted data; M.C.C. helped conceive the original study and had intellectual input; and M.A.P. and C.G.V. conceived the study, analyzed data, provided intellectual input, and revised the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Carola G. Vinuesa, Department of Pathogens and Immunity, John Curtin School of Medical Research, Australian National University, Building 131, Garran Road, Canberra, ACT 0200, Australia; e-mail: carola.vinuesa@anu.edu.au.