Abstract

   Interleukin-2 (IL-2) regulates immune homeostasis by fine-tuning the
   balance between effector and regulatory T (T[reg]) cells. To identify
   regulators of IL-2 signaling, we performed genome-wide CRISPR-knockout
   screening in IL-2–dependent cells derived from a patient with adult T
   cell leukemia (ATL) and found enrichment of single guide RNAs targeting
   PRDM1, which encodes B lymphocyte–induced maturation protein 1
   (BLIMP1). BLIMP1 inhibits IL-2 production by T cells; however, its role
   in IL-2 signaling remains unknown. Here, we show that overexpressing
   Prdm1 down-regulated IL-2 signaling, whereas Prdm1-deficiency enhanced
   IL-2 signaling in mouse CD4^+ T cells and T[reg] cells with augmented
   IL-2 signaling in T cells from influenza-infected mice and during
   adoptive T cell transfer–induced colitis. Deleting PRDM1 in human CD4^+
   T cells and T[reg] cells also increased IL-2 signaling. Furthermore,
   CD4^+ T cells from patients with ATL expressed less BLIMP1 and had
   enhanced IL-2 signaling, whereas overexpressing PRDM1 in ATL cells
   suppressed IL-2 signaling. Thus, BLIMP1 inhibits IL-2 signaling during
   normal and pathophysiological responses, suggesting that manipulating
   BLIMP1 could have therapeutic potential.
     __________________________________________________________________

   BLIMP1 is a negative regulator of IL-2 signaling in mouse and human T
   cells.

INTRODUCTION

   Interleukin-2 (IL-2) is a pleiotropic cytokine produced by
   antigen-activated T cells that play an important role in immune
   regulation, having effects on both effector (T[EFF]) and regulatory T
   (T[reg]) cells ([58]1–[59]5). IL-2 is a T cell growth factor that binds
   to its high-affinity IL-2 receptor (IL-2R) comprising three subunits,
   IL-2Rα (CD25), IL-2Rβ (CD122), and the common cytokine receptor
   γ-chain, IL-2Rγ (CD132) ([60]6–[61]12). IL-2 signaling results in the
   activation of signal transducers and activators of transcription 5
   (STAT5), phosphatidylinositol 3-kinase (PI3K)–AKT, and
   mitogen-activated protein kinase–extracellular signal–regulated kinase
   1/2 (ERK1/2) signaling pathways ([62]10, [63]13, [64]14), which in turn
   leads to the up-regulation of genes involved in the cell survival and
   proliferation ([65]15–[66]17). IL-2 signaling also regulates immune
   tolerance, with Il2^−/−, Il2ra^−/−, and Il2rb^−/− mice developing
   autoimmunity, while adoptive transfer of T[reg] cells into Il2rb^−/−
   mice prevents autoimmunity, consistent with the nonredundant role of
   IL-2 in the development of T[reg] cells ([67]18–[68]23). IL-2 signaling
   initiates a negative feedback loop to limit IL-2 production and
   IL-2–dependent responses. The repression of Il2 transcription is
   achieved in part by IL-2–dependent induction of B lymphocyte–induced
   maturation protein 1 (BLIMP1) and T cell receptor (TCR)–induced T-bet
   in T[EFF] cells ([69]24, [70]25).

   BLIMP1 is a transcription factor that is encoded by the gene Prdm1
   (positive regulatory domain zinc finger protein 1), originally found in
   B cells ([71]26). BLIMP1 is induced by TCR activation and
   IL-2–dependent signaling in T cells, while naive T cells have lower
   levels of Prdm1 mRNA expression ([72]27–[73]29). BLIMP1 inhibits Il2
   transcription and IL-2 production and thus participates in a negative
   autoregulatory feedback loop that is critical for controlling
   IL-2–dependent responses, particularly in situations in which it is
   triggered by continuous antigen exposure that results in expansion of
   antigen-reactive T cells ([74]30–[75]33). BLIMP1 is highly expressed in
   chronically activated T cells, which do not produce IL-2 ([76]34).
   BLIMP1 inhibits IL-2 production not only by directly repressing Il2
   transcription but also by repressing transcription of Fos, which is a
   positive regulator of Il2 transcription ([77]24, [78]28, [79]35,
   [80]36). However, whether BLIMP1 additionally regulates IL-2 signaling
   in T cells has not been reported.

   To identify the genes whose expression might positively or negatively
   regulate IL-2 signaling, we performed a genome-wide CRISPR knockout
   screening using an IL-2–responsive cell line, ED40515(+), derived from
   a patient with adult T cell leukemia (ATL). ATL is an aggressive form
   of leukemia that develops in approximately 5% of individuals infected
   with human T cell lymphotropic virus-1 (HTLV-1) ([81]37). In the
   initial phase of ATL, leukemic CD4^+ T cells grow in an autocrine
   manner with persistent IL-2 expression and functional IL-2Rs ([82]38).
   The continuous proliferation of ATL cells during this phase can be
   inhibited using anti-Tac monoclonal antibody directed against IL-2Rα
   ([83]39). In the later phase, IL-2 production is lost; however, IL-2Rα
   expression persists, and the cells are no longer dependent on IL-2, in
   part due to a constitutively activated Janus kinase (JAK)–STAT
   signaling leading to IL-2–independent growth ([84]40). In our CRISPR
   knockout screen, PRDM1 (encoding BLIMP1) and PTEN were putatively
   identified as the two most significant negative regulators of
   IL-2–mediated cell proliferation. PTEN is a known negative regulator of
   IL-2R signaling ([85]41, [86]42), whereas BLIMP1 has not been shown to
   have such activity. Here, we describe the role of BLIMP1 as a negative
   regulator of IL-2 signaling in mouse and human T cells and in influenza
   infection and colitis mouse models, as well as in primary cells from
   patients with ATL.

RESULTS

CRISPR screening implicated BLIMP1 as a negative regulator of IL-2 signaling

   Considering the pivotal role of IL-2 signaling in sustaining immune
   homeostasis and balancing T[EFF] and T[reg] cell responses, it is
   important to understand how this signaling pathway is controlled. We
   therefore performed a genome-wide CRISPR knockout screen using
   IL-2–dependent ED40515(+) cells, which are CD4^+CD25^+ T[reg]–like
   cells derived from a patient with ATL, into which Cas9 was stably
   expressed. The CRISPR screen (fig. S1A) was designed to identify the
   genes that are important for IL-2–mediated cell proliferation and
   survival by comparative sequence analysis of single guide RNAs (sgRNAs)
   in control cells versus IL-2 expanded cells. We found a significant
   dropout of sgRNAs for genes known to be important for IL-2 signaling,
   including IL2RA, IL2RB, JAK3, and STAT5B, validating the screen. sgRNAs
   corresponding to PRDM1 were highly enriched, followed by those for PTEN
   (fig. S1B), making them the top hits and indicating that the deletion
   of these genes promoted IL-2–induced survival and proliferation of
   ED40515(+) cells. PTEN is known to inhibit IL-2R signaling in T[reg]
   cells ([87]41); however, the role of PRDM1 (BLIMP1) in the regulation
   of IL-2 signaling remains unknown.

BLIMP1 inhibits IL-2 signaling in mouse CD4^+ T cells

   To investigate the potential role of BLIMP1 in IL-2 signaling, Prdm1
   was overexpressed in mouse CD4^+ T cells. Purified CD4^+ T cells from
   the spleens of wild-type (WT) C57BL/6 mice were preactivated for 48
   hours and then retrovirally transduced with pRV-EV (empty vector
   control) or pRV-Prdm1 (Prdm1–expressing vector) ([88]Fig. 1A). There
   was a significant increase in BLIMP1 protein expression after
   retroviral transduction–mediated overexpression of Prdm1 in mouse CD4^+
   T cells ([89]Fig. 1B), with lower expression levels of CD25 ([90]Fig.
   1C) and CD122 ([91]Fig. 1D). There was markedly reduced IL-2 production
   ([92]Fig. 1E) and decreased phosphorylation of STAT5, AKT, and ERK
   ([93]Fig. 1, F and G) after Prdm1 overexpression in mouse CD4^+ T
   cells. Thus, overexpressing BLIMP1 represses the expression of key
   receptor subunits and phosphorylated proteins involved in IL-2
   signaling in mouse CD4^+ T cells.

Fig. 1. BLIMP1 inhibits IL-2 signaling in mouse CD4^+ T cells.

   [94]Fig. 1.
   [95]Open in a new tab

   (A to G) CD4^+ T cells were isolated from the spleens of WT C57BL/6
   mice and TCR-stimulated for 24 hours at 37°C. Cells were retrovirally
   transduced with either pRV-EV or pRV-Prdm1 and cultured with IL-2 (200
   IU/ml) for 72 hours at 37°C. Cells were harvested and stained for flow
   cytometry analysis. (A) Experimental setup for (B) to (G), n = 5 to 6
   individual mice per group. Created in BioRender. Roy, S. (2025)
   [96]https://BioRender.com/922i5p2. [(B) to (D)] Histograms with
   statistical representation for (B) BLIMP1, (C) CD25, and (D) CD122 are
   shown. (E) Representative flow plot and statistics showing the
   frequency of IL-2–producing CD4^+ T cells. (F) Histograms showing
   pSTAT5, pAKT, and pERK staining by flow cytometry. (G) Statistical
   representation of flow cytometry analysis of pSTAT5, pAKT, and pERK.
   Representative data (means ± SEM) from three independent experiments
   (n = 4 to 6 individual mice per group) are shown. Two-tailed unpaired
   Student’s t test was used for statistical analysis. (H to L) CD4^+ T
   cells were purified from spleens of Prdm1^fl/fl and Prdm1^fl/flCD4^cre
   mice. Cells were TCR-stimulated in the presence of IL-2 (200 IU/ml) for
   72 hours at 37°C, harvested, stained, and subjected to flow cytometric
   analysis. Shown are histograms and statistical analysis of protein
   expression for (H) CD25, (I) CD122, and [(J) and (K)] pSTAT5, pAKT, and
   pERK. (L) Representative flow plot and statistics showing the frequency
   of IL-2–producing CD4^+ T cells. Data are representative of means ± SEM
   from three independent experiments (n = 4 to 6 mice per group). A
   two-tailed unpaired Student’s t test was used for statistical analysis.

   Conversely, we next eliminated Prdm1 (BLIMP1) expression in mouse CD4^+
   T cells using Prdm1^fl/flCD4^cre conditional knockout (CKO) mice. CD4^+
   T cells from Prdm1^fl/fl (WT) and Prdm1^fl/flCD4^cre (Prdm1-CKO) mice
   were preactivated with anti-CD3 + anti-CD28 in the presence of IL-2 for
   72 hours at 37°C. Although there was only a modest increase in the
   levels of CD25 ([97]Fig. 1H) and CD122 ([98]Fig. 1I) protein, the
   phosphorylation of STAT5, AKT, and ERK was significantly enhanced in
   CD4^+ T cells from Prdm1-CKO mice as compared to WT mice ([99]Fig. 1, J
   and K). As expected, IL-2 production was also higher after deleting
   Prdm1 in mouse CD4^+ T cells ([100]Fig. 1L). Together, these results
   indicate that BLIMP1 negatively regulates IL-2 signaling in mouse CD4^+
   T cells, potentially in part by inhibiting the expression of CD25 and
   CD122.

Prdm1 knockout up-regulates IL-2 signaling in influenza virus–infected T
cells

   BLIMP1 is known to be essential for CD8^+ T cell differentiation and
   cytotoxic function during influenza infection ([101]43); however, the
   effect of BLIMP1 deficiency on CD4^+ T cell responses in influenza
   infection has not been reported. In addition, IL-2 signaling has been
   shown to be induced during influenza infection ([102]44). We therefore
   studied the role of BLIMP1 in regulating IL-2 signaling in CD4^+ T
   cells from influenza-infected mice. We intranasally infected
   Prdm1^fl/fl (WT) and Prdm1^fl/flCD4^cre (CKO) mice with influenza virus
   strain PR8 (a pathogenic mouse-adapted strain of H1N1 influenza) and
   harvested mediastinal lymph nodes (medLNs) and lungs 10 days later
   ([103]Fig. 2A). CD4^+ T cells were then stimulated with IL-2 and
   influenza-specific peptide (nucleoprotein NP[311–325]), and NP[311–325]
   tetramer–positive CD4^+ T cell populations including T follicular
   helper (T[FH]), T follicular regulatory (T[FR]), T[EFF], and
   conventional T[reg] (cT[reg]) cells were analyzed ([104]Fig. 2A). There
   was an increased frequency of influenza-specific total CD4^+ T cells in
   both the medLNs and lungs of Prdm1-CKO mice as compared to WT mice
   (fig. S2, A to D). Among the different CD4^+ T cell populations (gating
   strategy shown in fig. S3, A to C), the frequency of influenza-specific
   T[EFF] cells was higher in medLNs of Prdm1-CKO mice as compared to WT
   mice (fig. S3D). The percentage of total T[FH] and T[FR] cell
   populations in the medLNs were similar in influenza-infected WT and
   Prdm1-CKO mice (fig. S3, E and F), whereas the frequency of T[reg]
   cells trended slightly higher in influenza-infected Prdm1-infected mice
   but the increase was not statistically significant (fig. S3G).
   Nevertheless, we found higher expression of CD25 ([105]Fig 2, B and C),
   CD122 ([106]Fig. 2, D and E), and phospho-STAT5 (pSTAT5) ([107]Fig. 2,
   F and G) in all populations (T[FH], T[FR], cT[reg], and T[EFF] cells)
   in the medLNs of influenza-infected Prdm1-CKO mice as compared to WT
   mice. There was also increased frequency of T[EFF] and cT[reg] cells
   from the lungs of influenza-infected Prdm1-CKO mice (fig. S3, H and I).
   Strikingly, there was higher expression of CD25, CD122, and pSTAT5 in
   cT[reg] cells from the lungs of influenza-infected Prdm1-CKO mice than
   in WT mice ([108]Fig. 2, H to J), as well as higher production of
   interferon-γ (IFN-γ) and IL-2 but lower production of IL-10 by
   Prdm1-CKO cT[reg] cells (fig. S3, J to L), consistent with a
   dysregulated immune response to influenza infection. Overall, these
   results are consistent with BLIMP1 controlling influenza infection, at
   least in part, by regulating IL-2 signaling in CD4^+ T cells.

Fig. 2. Prdm1 knockout up-regulates IL-2 signaling in CD4^+ T cells during
influenza infection.

   [109]Fig. 2.
   [110]Open in a new tab

   (A to J) Prdm1^fl/fl (WT) and Prdm1^fl/flCD4^cre (Prdm1-CKO) mice were
   infected intranasally with 1000 viral focal units (VFU) of PR8
   influenza virus. At day 10, medLNs and lungs were harvested, and
   single-cell suspensions were prepared. Cells were stimulated with
   NP[311–325] peptide (1 μM) for 5 hours at 37°C, stained, and subjected
   to flow cytometry analysis. (A) Experimental setup for (B) to (J).
   Created in BioRender. S. Roy (2025) [111]https://BioRender.com/ja4wpce.
   Flow cytometry plots with statistical representation showing [(B) and
   (C)] CD25, [(D) and (E)] CD122, and [(F) and (G)] pSTAT5 staining in
   NP[311–325] tetramer^+ influenza-specific follicular T helper cells
   (T[FH])[,] follicular T[reg] cells (T[FR])[,] conventional T[reg] cells
   (cT[reg]), and T[EFF] cells from the medLNs of WT and Prdm1-CKO mice.
   [(H) to (J)] Histograms with statistical representation showing (H)
   CD25, (I) CD122, and (J) pSTAT5 staining in NP[311–325] tetramer^+
   influenza-specific cT[reg] cells from the lungs of WT and Prdm1-CKO
   mice by flow cytometry. Data are representative of means ± SEM from two
   independent experiments (n = 4 to 5 mice per group) are shown.
   Two-tailed unpaired Student’s t test was used for statistical analysis.

BLIMP1-deficient T cells have enhanced IL-2–STAT5–dependent gene expression

   We next sought to better understand the effects of Prdm1 deletion on
   IL-2–induced gene expression following influenza infection. Purified
   CD4^+ T cells from medLNs of influenza-infected WT (Prdm1^fl/fl) and
   Prdm1-CKO (Prdm1^fl/flCD4^cre) mice were treated with IL-2 for 24 hours
   after ex vivo influenza-specific peptide (NP[311–325]) stimulation and
   subjected to RNA sequencing (RNA-seq) analysis ([112]Fig. 3A). Gene
   ontology (GO) and hallmark pathway analysis revealed a differential
   regulation of cytokine-mediated signaling pathway in influenza-infected
   CD4^+ T cells between Prdm1-CKO and WT mice ([113]Fig. 3B), with
   enrichment of IL-2–STAT5 signaling in influenza-infected CD4^+ T cells
   from Prdm1-CKO mice compared to WT mice ([114]Fig. 3, C and D),
   suggesting that BLIMP1 repressed IL-2 signaling in influenza-infected
   CD4^+ T cells. The genes associated with IL-2–STAT5 signaling exhibited
   greater expression levels in CD4^+ T cells from Prdm1-CKO mice compared
   to those from WT mice ([115]Fig. 3E).

Fig. 3. Enhanced IL-2–STAT5–dependent gene expression in BLIMP1-deficient T
cells.

   [116]Fig. 3.
   [117]Open in a new tab

   (A to E) Mouse CD4^+ T cells were isolated from the medLNs of
   influenza-infected Prdm1^fl/fl (WT) and Prdm1^fl/flCD4^cre (Prdm1-CKO)
   mice. Cells were stimulated ex vivo with influenza-specific peptide
   NP[311–325] (1 μM), rested overnight, and stimulated with 200 IU of
   IL-2 for 24 hours. Cells were lysed, RNA was extracted, and RNA-seq was
   performed. (A) Schematic experimental design for (B) to (E). Created in
   BioRender. S. Roy (2025) [118]https://BioRender.com/ydnx0k6. (B) GO
   based pathway over-enrichment analysis showing the differentially
   regulated pathways in influenza-infected CD4^+ T cells from Prdm1-CKO
   mice and WT mice. (C) Preranked gene set enrichment analysis (GSEA)
   showing hallmark pathways that were up-regulated or down-regulated in
   influenza-infected CD4^+ T cells from Prdm1-CKO mice relative to WT
   mice. (D) Hallmark enrichment score of IL-2–STAT5 signaling pathway in
   influenza-infected CD4^+ T cells from Prdm1-CKO mice. (E) Heatmap
   showing top DEGs in the IL-2–STAT5 signaling pathway in
   influenza-infected CD4^+ T cells from WT and Prdm1-CKO mice. (F to I)
   CD4^+FOXP3^+ T[reg] cells were sorted from Foxp3^YFP-cre (WT) and
   Prdm1^fl/flFoxp3^YFP-cre (Prdm1-CKO) mice based on YFP expression.
   Cells were TCR-stimulated for 72 hours, rested overnight, stimulated
   with IL-2 (500 IU/ml) for 24 hours, and then lysed for RNA extraction
   followed by library preparation for RNA-Sequencing. (F) Schematic
   experimental design for (G) to (I). Created in BioRender. S. Roy (2025)
   [119]https://BioRender.com/ydnx0k6. (G) GO-based over-enrichment
   analysis showing the differentially regulated pathways in T[reg] cells
   from Prdm1-CKO mice as compared to WT mice. (H) Over-enrichment
   analysis showing differentially regulated hallmark pathways in T[reg]
   cells from Prdm1-CKO mice relative to WT mice. (I) Heatmap showing top
   DEGs of the IL-2–STAT5 signaling pathway in T[reg] cells from WT and
   Prdm1-CKO mice.

   As noted above, the CRISPR knockout screening was performed in an
   ATL-derived cell line, ED40515(+) cells. Because ATL cells typically
   have a T[reg]-like suppressor phenotype ([120]45, [121]46) and since
   the absence of BLIMP1 resulted in enhanced IL-2 signaling in cT[reg]
   cells during influenza infection, we next investigated the role of
   BLIMP1 in IL-2 signaling in ex vivo–isolated T[reg] cells. Splenic
   T[reg] cells from WT (Foxp3^YFP-cre) and Prdm1-CKO
   (Prdm1^fl/flFoxp3^YFP-cre) mice were isolated, preactivated with
   anti-CD3 + anti-CD28, rested overnight, cultured with IL-2 for 24
   hours, and then subjected to RNA-seq analysis ([122]Fig. 3F). The GO
   analysis showed that the cytokine-mediated signaling pathway was
   differentially regulated in T[reg] cells from Prdm1-CKO versus WT mice
   ([123]Fig. 3G, red arrow). In the Hallmark pathway analysis, IL-2–STAT5
   signaling was enriched with higher expression of genes associated with
   this pathway in T[reg] cells from Prdm1-CKO mice ([124]Fig. 3H, red
   arrow, and [125]Fig. 3I). Thus, BLIMP1 limits IL-2–STAT5 signaling in
   ex vivo–isolated splenic T[reg] cells as well as in influenza-infected
   mouse CD4^+ T cells.

Diminished BLIMP1 augments IL-2 signaling but attenuates T[reg] suppressive
activity

   We further assessed the effects of BLIMP1 on IL-2 signaling in ex
   vivo–isolated FOXP3-YFP^+ T[reg] cells at the protein level (gating
   strategy shown in fig. S4A) and found that Prdm1-deficient T[reg] cells
   had increased CD25 and CD122 expression ([126]Fig. 4, A and B).
   Consistent with this, pSTAT5 was elevated in Prdm1-deficient T[reg]
   cells ([127]Fig 4, C and D); however, there were no significant changes
   in the levels of pAKT and pERK (fig. S4, B and C) in keeping with the
   observation that T[reg] cells have defective PI3K/AKT signaling, which
   may explain their hypoproliferative response to IL-2 ([128]41). We also
   found that IL-10 production was lower ([129]Fig. 4, E and F), while
   FOXP3 and HELIOS expression were higher in Prdm1-deficient T[reg] cells
   (fig. S4, D and E).

Fig. 4. Absence of BLIMP1 augments IL-2 signaling and attenuates T[reg]
suppressive activity.

   [130]Fig. 4.
   [131]Open in a new tab

   (A to F) T[reg] cells were purified from the spleens of Foxp3^YFP-cre
   (WT) and Prdm1^fl/flFoxp3^YFP-cre (Prdm1-CKO) mice. Cells were
   TCR-stimulated in the presence of IL-2 (500 IU/ml) for 72 hours at
   37°C, harvested, and stained for flow cytometry analysis. Live cells
   were gated as CD4^+CD25^+FOXP3^+ T[reg] cells. Histograms and
   statistical analysis of protein expression for (A and B) CD25 and
   CD122, [(C) and (D)] pSTAT5, and [(E) and (F)] IL-10 by flow cytometry.
   Data are representative of means ± SEM from three independent
   experiments (n = 5 to 6 individual mice per group). A two-tailed
   unpaired Student’s t test was used for statistical analysis. (G to N)
   Adoptive T[reg] transfer from Foxp3^YFP-cre (WT) or
   Prdm1^fl/flFoxp3^YFP-cre (Prdm1-CKO) mice to Rag2^−/− mice with T cell
   transfer induced colitis. (G) Schematic representation for group I,
   Rag2^−/− mice with no T cell transfer; group II, Rag2^−/− mice with
   intravenous (i.v.) transfer of CD45.1^+ naïve CD4^+ T cells; group III,
   Rag2^−/− mice with intravenous transfer of CD45.1^+ naïve CD4^+ T cells
   and CD45.2^+ WT T[reg] cells from WT mice; and group IV, Rag2^−/− mice
   with intravenous transfer of CD45.1^+ naïve CD4^+ T cells and CD45.2^+
   CKO T[reg] cells from Prdm1-CKO mice. Created in BioRender. S. Roy
   (2025) [132]https://BioRender.com/debaxae. [(H) and (I)] Flow cytometry
   analysis for (H) IFN-γ and (I) IL-10 production by CD45.2^+FOXP3^+
   T[reg] cells in the mesLNs. (J) Schematic representation for ex vivo
   IL-2 stimulation of T[reg] cells purified from spleen and mesLNs of
   Rag2^−/− mice after colitis induction. Created in BioRender. S. Roy
   (2025) [133]https://BioRender.com/debaxae. [(K) and (L)] Flow cytometry
   analysis of protein expression for (K) CD25 and (L) pSTAT5 in
   CD45.2^+FOXP3^+ T[reg] cells from the mesLNs. [(M) and (N)] Flow
   cytometry analysis of protein expression for (M) CD25 and (N) pSTAT5 in
   CD45.2^+FOXP3^+ T[reg] cells from the spleen. Two-tailed unpaired
   Student’s t test was used for statistical analysis. Data are
   representative of mean ± SEM from two independent experiments (n = 4
   independent mice per group).

   To complement these data, we assessed the in vivo functional relevance
   of Prdm1 deletion in T[reg] cells using a T cell–adoptive transfer
   colitis mouse model. Colitis was induced in Rag2^−/− mice by
   intravenously injecting CD45.1^+ naïve CD4^+ T cells with or without
   cotransferring CD45.2^+ T[reg] cells from WT or Prdm1-CKO mice
   ([134]Fig. 4G). There was a gradual decline in the body weight and
   shortening of the colon length in the mice that received the Prdm1-KO
   T[reg] cells (group IV), similar to the colitis induced in mice that
   did not receive T[reg] cells (group II); in contrast, the mice that
   received WT T[reg] cells (group III) progressively gained weight (fig.
   S5, A and B). The colon histology images and scores indicated severe
   colitis in the mice receiving Prdm1-KO T[reg] cells (group IV), while
   transfer of WT T[reg] cells (group III) suppressed the development of
   colitis (fig. S5C). Mesenteric lymph nodes (mesLNs) and spleen were
   analyzed by flow cytometry for CD45.1^+Tbet^+ and CD45.2^+FOXP3^+
   populations (fig. S5D). Colitis was not suppressed in the mice
   receiving Prdm1-KO T[reg] cells, with greater accumulation of
   CD45.1^+Tbet^+IFN-γ^+ T cells in the mesLNs and spleen than in mice
   that received WT T[reg] cells (fig. S5, E to H). The increased
   intestinal inflammation in the mice receiving Prdm1-KO T[reg] cells was
   associated with higher levels of IFN-γ and lower levels of IL-10
   production in the CD45.2^+FOXP3^+ T[reg] cells in the mesLNs ([135]Fig.
   4, H and I) and spleen (fig. S5, I and J). Furthermore, ex vivo
   stimulation with IL-2 induced higher expression of CD25 and pSTAT5 in
   CD45.2^+FOXP3^+ T[reg] cells from the mesLNs ([136]Fig. 4, K and L) and
   spleen ([137]Fig. 4, M and N) from mice that received adoptively
   transferred Prdm1-CKO T[reg] cells as compared to WT T[reg] cells.
   These data indicate that in the absence of BLIMP1, T[reg] cells lose
   their ability to suppress inflammatory responses in the intestine,
   which could be attributed to exacerbated IL-2 signaling and reduced
   IL-10 production, accompanied by heightened T helper 1 (T[H]1) effector
   responses with increased IFN-γ production.

BLIMP1 restrains IL-2 signaling in human CD4^+ T cells and T[reg] cells

   We next assessed whether the findings in mouse CD4^+ T cells extended
   to human CD4^+ T cells by using CRISPR-Cas9 gene editing to delete
   PRDM1 in preactivated CD4^+ T cells isolated from buffy coats of
   healthy donors (HDs). The gene editing efficiency in deleting the PRDM1
   gene was validated by performing a T7 endonuclease I assay in human
   CD4^+ T cells (fig. S6A), and correspondingly, there was significantly
   reduced BLIMP1 protein in cells electroporated with PRDM1–guide RNAs
   (gRNAs) as compared to cells receiving control gRNAs (fig. S6B).
   CRISPR-Cas9–mediated deletion of PRDM1 in human CD4^+ T cells led to
   higher expression of CD25 and CD122 protein ([138]Fig. 5, A and B) and
   a significant increase in pSTAT5, pAKT, and pERK ([139]Fig. 5, C and
   D). As expected, there was also enhanced production of IL-2 in
   PRDM1-deleted human CD4^+ T cells (fig. S6C).

Fig. 5. BLIMP1 restrains IL-2 signaling in human CD4^+ T cells and natural
T[reg] cells.

   [140]Fig. 5.
   [141]Open in a new tab

   (A to D) Preactivated human CD4^+ T cells were electroporated with
   control gRNAs or PRDM1 gRNAs annealed with Cas9 as RNP complex,
   cultured with IL-2 (200 IU/ml) for 72 hours, and stained for flow
   cytometry. Histograms with statistical representation showing (A) CD25
   and (B) CD122 protein levels by flow cytometry. (C) Histograms
   representing pSTAT5, pAKT, and pERK, staining by flow cytometry. (D)
   Statistical representation of flow cytometry analysis of pSTAT5, pAKT,
   and pERK. Data are representative of means ± SEM from three independent
   experiments (n = 5 individuals). Two-tailed paired Student’s t test was
   used for statistical analysis. (E to H) Preactivated human natural
   T[reg] cells were electroporated with control gRNAs or PRDM1 gRNAs
   annealed with Cas9 as RNP complex, cultured with IL-2 (500 IU/ml) for
   72 hours, and stained for flow cytometry. Histograms with statistical
   representation showing (E) CD25 and (F) CD122 protein levels by flow
   cytometry. (G) Histograms representing pSTAT5, pAKT, and pERK staining
   by flow cytometry. (H) Statistical representation of flow cytometry
   analysis of pSTAT5, pAKT, and pERK. Data are representative of
   means ± SEM from three independent experiments (n = 5 individuals).
   Two-tailed paired Student’s t test was used for statistical analysis.
   (I to L) Natural T[reg] cells from HDs (D1 and D2) were in vitro
   expanded in the presence of IL-2 followed by ChIP-seq analysis of
   BLIMP1 binding using antibodies BLIMP1 Ab1 (clone #9115S) and BLIMP1
   Ab2 (clone #MA5-14879) as compared to the rabbit IgG control. (I)
   BLIMP1 ChIP-seq heatmaps (promoter versus distal regions). (J) Motif
   analysis for BLIMP1 peaks in the promoter and distal regions. (K) GO
   term enrichment for BLIMP1 peaks near TSS target genes. (L) Genome
   browser tracks showing BLIMP1-binding profiles at the indicated genes.

   We next examined the role of BLIMP1 in IL-2 signaling in natural T[reg]
   cells isolated from buffy coats of healthy individuals by deleting
   PRDM1 with CRISPR-Cas9. We again confirmed the knockout efficiency by
   the T7 endonuclease I assay in human T[reg] cells (fig. S6D). BLIMP1
   protein expression was markedly lower in the cells that received PRDM1
   gRNAs than in the cells that received control gRNAs (fig. S6E). We
   found higher levels of CD25 and CD122 protein expression after PRDM1
   deletion in human T[reg] cells ([142]Fig. 5, E and F). Consistent with
   the findings in mouse T[reg] cells, PRDM1 deletion in human T[reg]
   cells also induced more pSTAT5 but only minimal changes in the
   expression of pAKT and pERK ([143]Fig. 5, G and H). The protein
   expression levels of FOXP3 (fig. S6F) and HELIOS (fig. S6G) were
   elevated, while IL-10 production was diminished (fig. S6H) after PRDM1
   deletion in human T[reg] cells. These results in the human CD4^+ T
   cells and T[reg] cells recapitulated those in the mouse, establishing
   that BLIMP1 restrains IL-2 signaling in CD4^+ T cells, particularly
   T[reg] cells, in both mice and humans.

   To gain mechanistic insights into the regulation of IL-2 signaling by
   BLIMP1, we performed BLIMP1 chromatin immunoprecipitation sequencing
   (ChIP-seq) in human natural T[reg] cells. Natural T[reg] cells were
   isolated from healthy human donors, expanded in vitro in the presence
   of IL-2 and subjected to ChIP-seq analysis. BLIMP1 binding using two
   antibodies: BLIMP1 Ab1 (clone #9115S) and BLIMP1 Ab2 (clone #MA5-14879)
   was assessed as compared to rabbit anti–immunoglobulin G (IgG) control.
   The identification of BLIMP1 peaks revealed notably enriched genomic
   loci, visualized as intensity signals on the heatmaps. Approximately
   50% of the BLIMP1 ChIP-seq peaks were localized within 2 kb of 5′ of
   the transcription start site (TSS) and thus were defined as binding in
   the promoter region, while the peaks more than 2 kb from the TSS were
   defined as distal regions ([144]Fig. 5I). We performed de novo motif
   analysis to identify and explain the differential BLIMP1 binding and
   chromatin accessibility at the ChIP-seq peaks. The analysis revealed
   similar and highly enriched BLIMP1-binding motifs in both the promoter
   and distal regions; interferon regulatory factor (IRF) and E26
   transformation-specific (ETS) were the second and third most
   significant motifs in the promoter region, and IRF and Runt-related
   transcription factor (RUNX) motifs were the second and third most
   significant in the distal regions ([145]Fig. 5J). Overall, there were
   1041 genes with BLIMP1 motifs, 526 with IRF motifs, 860 with ETS
   motifs, and 378 with RUNX motifs (table S1). BLIMP1 and IRF motifs were
   found for IL2, IL2RA, and IL10; ETS motifs for IL2RA and IL10; and RUNX
   motifs for IL2RA, indicating that more than one motif was identified
   for some key genes and suggesting that there may be cooperative effects
   among IRF, ETS, and RUNX family proteins with BLIMP1, an area for
   future investigation. GO analysis of BLIMP1 binding peaks proximal to
   TSS of target genes revealed enrichment of the cytokine-mediated
   signaling pathway, evidenced by the most significant P value ([146]Fig.
   5K, red arrow). BLIMP1 binds near the promoter regions of IL2 and IL2RB
   but in the enhancer regions of IL2RA and IL10 ([147]Fig. 5L). Since
   BLIMP1 negatively regulates expression of IL-2, IL-2Rα, and IL-2Rβ,
   these findings are consistent with direct negative regulation of IL2,
   IL2RA, and IL2RB by BLIMP1. The binding to the IL10 gene is interesting
   given that BLIMP1 is a positive regulator of IL-10 expression.
   Together, the ChIP-seq analysis and our functional data suggest that
   BLIMP1 directly binds and regulates the key genetic elements involved
   in IL-2 signaling in human T[reg] cells.

CD4^+ T cells from patients with ATL have amplified IL-2 signaling

   The results in mouse cells, including influenza infection and T
   cell–induced colitis models, as well as in primary human cells,
   indicate that BLIMP1 negatively regulates IL-2 signaling that helps to
   modulate the immune response. Since the CRISPR knockout screening was
   performed in an IL-2–dependent ATL cell line, we next analyzed primary
   cells from patients with acute ATL. Purified CD4^+ T cells from frozen
   peripheral blood mononuclear cells (PBMCs) of HDs and patients with
   acute ATL were stimulated with anti-CD3 + anti-CD28 and IL-2 for 24
   hours at 37°C ([148]Fig. 6A). Cells were gated as CD3^+CD4^+
   CD25^+CCR4^+ “ATL-like T[reg]” cells [based on the phenotype of ATL
   cells with fluorescence minus one, (FMO) controls] and analyzed by flow
   cytometry (fig. S7, A to D). As expected, the frequency of these cells
   were much lower in HDs, while most ATL cells had this phenotype (fig.
   S7D). T[reg]-like cells from patients with ATL showed higher expression
   of CD25 and CD122 as compared to HDs ([149]Fig. 6, B to E). Consistent
   with ATL cells having T[reg]-like suppressor activity ([150]45,
   [151]46), there was higher FOXP3 expression in cells from patients with
   ATL than in HDs ([152]Fig. 6, F and G). We also found more
   IL-10–producing T[reg]-like cells in patients with ATL in comparison to
   HDs, consistent with the suppressive properties of ATL cells (fig.
   S7E). Cells from patients with ATL showed augmented IL-2 signaling as
   manifested by enhanced phosphorylation of STAT5, AKT, and ERK
   ([153]Fig. 6, H and I). Notably, BLIMP1 protein expression was lower in
   cells ex vivo treated with IL-2 from patients with ATL as compared to
   HDs either after 24 hours ([154]Fig. 6, J and K) or 48 or 72 hours
   (fig. S7, F and G), inversely correlating BLIMP1 expression with
   activated IL-2 signaling in cells from patients with acute ATL. We also
   investigated other potential regulators of IL-2 signaling such as Tbet,
   PTEN, and THEMIS ([155]3, [156]41, [157]47). We observed a modest
   increase in expression of PTEN but decreased expression of Tbet and
   THEMIS in ATL cells as compared to HDs (fig. S7, H and I), suggesting
   their possible role in ATL; however, the impact of BLIMP1 was clearly
   more evident.

Fig. 6. Patients with acute ATL showed augmented IL-2 signaling as compared
to HDs.

   [158]Fig. 6.
   [159]Open in a new tab

   Purified CD4^+ T cells from frozen PBMCs of HDs and patients with acute
   ATL were ex vivo stimulated with anti-CD3 and anti-CD28 and IL-2 for 24
   hours at 37°C and stained for flow cytometry. Cells were gated as CD3^+
   CD4^+ CD25^+ CCR4^+ T cells (ATL-like T[reg] cells). (A) Schematic
   representation of the experiment (B to K). Created in BioRender. S. Roy
   (2025) [160]https://BioRender.com/ae8e1h0. Representative histograms
   with cumulative statistics for flow cytometry analysis of [(B) and (C)]
   CD25 and [(D) and (E)] CD122 expression. [(F) and (G)] Representative
   histograms and cumulative statistics for FOXP3 expression by flow
   cytometry. [(H) and (I)] Flow cytometric analysis of pSTAT5, pAKT, and
   pERK showing representative histograms and cumulative statistics. [(J)
   and (K)] Representative histograms and statistics for BLIMP1 expression
   by flow cytometry. Data are representative of means ± SEM from three
   independent experiments (n = 4 to 5 individuals). Two-tailed unpaired
   Student’s t test was used for statistical analysis. (L) Schematic
   representation of PRDM1 overexpression in ATL cells. Created in
   BioRender. S. Roy (2025) [161]https://BioRender.com/ae8e1h0. (M)
   Histograms showing protein expression for BLIMP1, CD25, CD122, pSTAT5,
   pAKT, and pERK in ATL cells transduced with pLVX-EV (empty vector
   control) or pLVX-PRDM1 (vector expressing PRDM1). (N) Statistical
   analysis showing mean fluorescent intensity (MFI) for BLIMP1, CD25,
   CD122, pSTAT5, pAKT, and pERK, protein levels in ATL cells transduced
   with pLVX-EV (empty vector control) or pLVX-PRDM1 (vector expressing
   PRDM1). Data are representative of means ± SEM from two independent
   experiments (n = 5 individuals). Two-tailed paired Student’s t test was
   used for statistical analysis.

   Because the above data indicate that the activated IL-2 signaling in T
   cells from patients with acute ATL was associated with lower BLIMP1
   protein levels, we further investigated a mechanistic role for BLIMP1
   by transducing either control or PRDM1-expressing lentiviruses in ATL
   cells and examining the effect of IL-2 ([162]Fig. 6L). Increasing
   BLIMP1 expression in ATL cells resulted in significantly lower CD25 and
   CD122 protein levels and decreased phosphorylation of STAT5, AKT, and
   ERK ([163]Fig. 6, M and N). These results suggest that BLIMP1
   expression reciprocally regulates the activation of IL-2 signaling in
   patients with acute ATL.

PRDM1 expression inversely correlates with IL-2 signaling in patients with
ATL

   We next performed single-cell RNA-seq (scRNA-Seq) analyses in CD4^+ T
   cells from HDs and patients with acute ATL that were stimulated with
   anti-CD3 and anti-CD28 in the presence of IL-2 for 4 hours at 37°C
   ([164]Fig. 7A). An integrated scRNA-seq dataset from three HD samples
   and five ATL samples was annotated with Azimuth. The CD4^+ T cells were
   annotated as “CD4 TCM” or central memory T cells (n = 26978), “T[reg]”
   cells (n = 11455), and “Other” T cells (n = 404) (which was a mix of
   cells annotated as CD4^+ proliferating, naïve, and effector memory
   cells) ([165]Fig. 7B) (see table S2 for the markers used by Azimuth to
   define the populations). All five ATL samples showed a higher
   proportion of T[reg] cells compared to the three HD samples ([166]Fig.
   7C). An integrated dataset for T[reg] cells from the five ATL and three
   HD samples showed higher abundance of T[reg] cells in ATL as compared
   to HD ([167]Fig. 7D). Unsupervised clustering revealed seven different
   subpopulations (clusters 0 to 6) of T[reg] cells (fig. S7J). The T[reg]
   population in ATL samples was more diverse compared to HD samples (fig.
   S7J). More than 70% of T[reg] cells in HD samples were clustered
   together in cluster 4, which was nearly absent in ATL. This suggested
   the presence of one major T[reg] subpopulation in HD samples, whereas
   ATL samples had multiple T[reg] subpopulations (fig. S7, J and K). The
   seven subclusters of T[reg] cells in HD and ATL showed differential
   gene expression profile with top 32 genes average expression being
   represented (fig. S7L). We next evaluated PRDM1 expression in T[reg]
   cells and found higher PRDM1 expression in HD as compared to ATL
   ([168]Fig. 7E). The integrated dataset from HD and ATL showed a higher
   proportion of PRDM1^− versus PRDM1^+ subpopulations in T[reg] cells in
   ATL but almost equal proportions of PRDM1^− and PRDM1^+ subpopulations
   in T[reg] cells in HD ([169]Fig. 7F). Approximately 78% of T[reg] cells
   in ATL were PRDM1^−, whereas the PRDM1^+ and PRDM1^− subpopulations
   were almost equivalent in HD ([170]Fig. 7G). Thus, most T[reg]-like
   cells in ATL lack PRDM1 expression. Using the differentially expressed
   genes (DEGs) as input (see table S3), the Qiagen Ingenuity Pathway
   Analysis (IPA) in PRDM1^− versus PRDM1^+ populations within T[reg]
   cells revealed up-regulation of pathways including JAK-STAT signaling
   and IL-2 signaling pathways in the PRDM1^− subpopulation of cells from
   patients with ATL ([171]Fig. 7H; see table S4 for the genes used by the
   Qiagen IPA to identify the pathways). This suggests that diminished
   PRDM1 expression potentially contributes to the constitutively
   activated JAK-STAT signaling and IL-2 signaling that are often found in
   ATL cells ([172]40).

Fig. 7. PRDM1 expression inversely correlates with activated IL-2 signaling
in patients with acute ATL.

   [173]Fig. 7.
   [174]Open in a new tab

   Purified CD4^+ T cells from HDs (n = 3) and patients with acute ATL
   (n = 5) were stimulated with anti-CD3 and anti-CD28 in presence of IL-2
   for 4 hours at 37°C. The cells were harvested and subjected to
   scRNA-seq. (A) Schematic representation of the experimental setup for
   (B to H). Created in BioRender. S. Roy (2025)
   [175]https://BioRender.com/9zdi679. (B) Uniform Manifold Approximation
   and Projection (UMAP) plot showing all cells split by HD cells (left)
   and ATL cells (right). (C) Bar plot showing proportion of the cells in
   each cluster annotated by Azimuth for each HD and ATL sample. (D) UMAP
   plot showing distribution of T[reg] cells split by HD cells (left) and
   ATL cells (right). (E) Violin plot showing PRDM1 expression in T[reg]
   cells in HDs and patients with ATL. (F) UMAP plot showing distribution
   of PRDM1^− versus PRDM1^+ T[reg] populations in HDs and patients with
   ATL. (G) Pie chart showing percentage of cells in HDs and patients with
   ATL that express any amount of PRDM1. (H) Bar plot showing top pathways
   (by z score) from Qiagen IPA using differentially expressed genes
   derived from PRDM1^− versus PRDM1^+ populations within T[reg] cells
   cluster in patients with ATL.

DISCUSSION

   The immunomodulatory effects of IL-2 on both T[EFF] cells and T[reg]
   cells make it a promising target for immunotherapy ([176]48). IL-2 has
   been exploited to selectively expand T[reg] cells in autoimmune
   diseases, while in cancer, expanding T[EFF] cells is desirable
   ([177]49–[178]54). Considering the immunostimulatory and
   immunosuppressive consequences of IL-2 signaling in different diseases,
   a deep understanding of its regulation is important. We therefore
   investigated IL-2 signaling by performing a genome-wide CRISPR knockout
   screen using IL-2–dependent ED40515(+) cells derived from patients with
   HTLV-I–associated ATL.

   The screen showed a dropout of sgRNAs for the genes IL2RA, IL2RB, JAK3,
   and STAT5B, which are required for IL-2 signaling, validating the
   screen. In contrast, all four sgRNAs for both PRDM1 and PTEN were
   significantly enriched. PTEN is a known inhibitor of IL-2R signaling in
   T[reg] cells where it inhibits PI3K-AKT signaling and IL-2–mediated
   cell proliferation, while the JAK-STAT5 signaling remains intact
   attributing to the hypoproliferative phenotype of T[reg] cells in
   response to IL-2 ([179]41, [180]42). BLIMP1 (encoded by PRDM1 in human
   and Prdm1 in mouse) is a transcription factor that inhibits IL-2
   production by binding directly to the Il2 promoter in T cells
   repressing the cell proliferation ([181]24, [182]35). In turn, IL-2
   induces BLIMP1 expression, establishing a negative feedback loop
   limiting its own production ([183]24, [184]35). It is known that IL-2
   primarily acts on the responding cells in a paracrine fashion by
   inducing phosphorylation of STAT5 ([185]55, [186]56). Our CRISPR
   knockout screen implicated BLIMP1 as a negative regulator of
   IL-2–mediated proliferation. Here, we demonstrated that the
   overexpression of Prdm1 in mouse CD4^+ T cells inhibited IL-2
   signaling, whereas deletion of Prdm1 in mouse CD4^+ T cells resulted in
   enhanced IL-2 signaling, as evidenced by increased protein expression
   of CD25 and CD122 as well as increased phosphorylation of STAT5, AKT,
   and ERK. These findings are also consistent with higher expression of
   Il2ra and Il2rb mRNA expression in LCMV-infected CD8^+ T cells from
   Prdm1-knockout mice ([187]57), although IL-2 signaling was not examined
   in that study. Previously, BLIMP1-deficient CD8^+ T cells were shown to
   have an impaired effector response to influenza virus ([188]43). IL-2
   signaling has been shown to be highest at the peak of influenza
   infection in various CD4^+ T cell populations, such as T[FH], T[FR],
   T[EFF], and T[reg] cells ([189]44). Here, we demonstrate that BLIMP1
   deficiency results in augmented IL-2 signaling in T[FH], T[FR], T[EFF],
   and cT[reg] cells after infection with influenza virus, with the most
   robust effect in the T[reg] cells. Viral antigen has been shown to
   induce the T[reg] cells in influenza virus–infected mice ([190]58); in
   mice with BLIMP1-deficient T cells, following influenza infection,
   there was increased frequency of T[reg] cells and T[EFF,] with
   decreased IL-10 but enhanced IFN-γ production and increased IL-2
   signaling in medLNs and lungs. This creates a dysregulated
   microenvironment that is ineffective in controlling the inflammatory
   responses in the medLNs and lungs of influenza virus–infected mice that
   succumb to the deleterious effects of augmented IL-2 signaling as well
   as the abrogated IL-10 production. This is consistent with a known
   positive regulation of IL-10 gene expression by BLIMP1 ([191]59), in
   addition to negative effects of BLIMP1 on IL-2 production and IL-2
   signaling as we show in this study. Thus, BLIMP1 is critical in
   modulating effector and regulatory functions of T[EFF] and T[reg] cells
   during influenza virus infection.

   Selectively knocking out Prdm1 in Foxp3-expressing T[reg] cells
   resulted in higher IL-2 signaling, with increased expression of T[reg]
   transcription factors FOXP3 and HELIOS but reduced IL-10 production.
   Moreover, adoptive transfer of Prdm1-deficient T[reg] cells into
   Rag2^−/− mice failed to suppress the T cell–induced colitis, likely due
   to their defective IL-10 production ([192]60–[193]63). In the absence
   of BLIMP1, transferred CD45.2^+FOXP3^+ T[reg] cells acquired an
   effector phenotype, producing more IFN-γ but less IL-10, which
   attenuates their ability to suppress the colonic intestinal
   inflammation. It is known that increased IFN-γ production after
   adoptive transfer of naïve CD4^+ T cells into Rag2^−/− mice contributes
   to the development of colitis ([194]64, [195]65). In addition, a prior
   study showed that Tbet, which induces IFN-γ from T[reg] cells,
   contributes to colitis ([196]66). IL-2 can induce Tbet and IL-12Rβ2 and
   promote T[H]1 differentiation with enhanced IFN-γ production ([197]67),
   indicating a mechanism for how increased IL-2 signaling may enhance
   IFN-γ production in T[reg] cells and limit their suppressive function,
   observations that support the idea that IFN-γ derived from
   BLIMP1-deficient T[reg] cells may contribute to inflammation in colitis
   and possibly in other autoimmune conditions. Furthermore, the enhanced
   IL-2 signaling in the transferred Prdm1-knockout T[reg] cells likely
   contributed to the T cell–induced colitis by expanding the number or
   function of these poorly suppressive “effector” T[reg] cells. Although
   the contribution of IL-2 signaling in colitis is unclear, the increased
   IL-2 signaling due to diminished BLIMP1 potentially provides an
   additional explanation for the aggravated inflammation during colitis
   in addition to the effects of lower levels of IL-10 and CTLA-4
   ([198]60–[199]63). Consistent with this idea, a recent study reported
   that the duplication of IL2RA locus results in excessive IL-2 signaling
   and that this predisposed to colitis in patients with inflammatory
   bowel disease ([200]68). Other studies have shown that overexpression
   of CD25 occurs in various autoimmune diseases and that blocking it can
   improve inflammation and clinical outcome ([201]69, [202]70). Thus,
   optimal BLIMP1 expression is vital for maintaining balanced IL-2
   signaling and intestinal homeostasis.

   In addition to augmented IL-2 signaling after Prdm1 deletion in the
   mouse, PRDM1 deletion also resulted in increased IL-2 signaling in
   human CD4^+ T cells and natural T[reg] cells. ChIP-seq showed binding
   of BLIMP1 to key genes associated with IL-2 signaling in human T[reg]
   cells. Since IL-2 activates JAK1 and JAK3 ([203]71), JAK inhibitors
   including ruxolitinib and tofacitinib has been used for the treatment
   of immunosuppressive diseases ([204]72). The combination of ruxolitinib
   and navitoclax (an inhibitor of Bcl2/Bcl-xL) showed promising antitumor
   efficacy and prolonged survival in a mouse model of ATL ([205]73).
   Although combination chemotherapy and immunotherapy has been used in
   the treatment of ATL, long-term success has been very limited,
   indicating a need for new therapeutic strategies. Acute ATL is the most
   common and aggressive form of ATL and is characterized by clonal
   expansion of HTLV-1–infected cells. Here, we demonstrated that patients
   with acute ATL have enhanced IL-2 signaling with reciprocally lower
   BLIMP1 expression and that overexpression of BLIMP1 could repress IL-2
   signaling. scRNA-seq analysis established that PRDM1 expression
   inversely correlates with the activation of JAK-STAT5 and IL-2
   signaling in ATL-like T[reg] cells from patients with acute ATL.
   Consequently, diminished induction of BLIMP1 correlates with augmented
   JAK-STAT and IL-2 signaling in ATL. In summary, we propose here that
   BLIMP1 serves as a pivotal regulatory node in the IL-2 signaling
   pathway, and thus modulating the activity of BLIMP1 may allow better
   control of IL-2–mediated diseases by calibrating immune responses and
   maintaining immune homeostasis.

METHODS

Healthy human donors and patients

   Buffy coats from healthy human volunteers were obtained from National
   Institutes of Health (NIH) Blood Bank, Bethesda, USA. Because Buffy
   coats are produced as a by-product of a volunteer whole-blood donation
   for transfusion and are thus already in existence (preexisting
   material), would otherwise be discarded if not distributed for research
   use, and are irreversibly anonymized before distribution, they meet the
   FDA 45 CFR46 2018 Revisions of the Common Rule and are exempt from the
   requirement for Institutional Review Board (IRB) review and informed
   consent. Frozen PBMCs from blood samples of patients with acute ATL
   were obtained from patients enrolled under a clinical trial approved by
   the NCI IRB, protocol number 97-C-0143. The original samples were
   collected under an approved protocol of the NCI. All cells provided to
   us were deidentified and were not considered human subjects research
   and did not require IRB approval. PBMCs from HDs were isolated and
   stored in liquid nitrogen to be used as controls. The information on
   the samples from patients with ATL is listed in table S5.

Mice

   C57BL/6 (B6) mice were obtained from the Charles River Laboratory
   (strain #027) and B6-CD45.1 mice (B6.SJL-Ptprc^aPepc^b/BoyJ) were
   obtained from the Jackson laboratory (strain #002014). B6-Rag2^−/−
   (C57BL/6-Rag2^−/−) mice ([206]74) were from Taconic. Prdm1^fl/fl mice
   ([207]75, [208]76) (the Jackson Laboratory strain #008100) were crossed
   with CD4^cre mice (the Jackson Laboratory, strain #022071) and
   Foxp3^YFP-cre mice (the Jackson Laboratory, strain #016959) to generate
   Prdm1^fl/flCD4^cre and Prdm1^fl/flFoxp3^YFP-cre mice, respectively. All
   mice were housed in pathogen-free BSL2 AALAC-accredited facilities at
   the NIH, Bethesda, USA. All mice used for experiments were 8 to 12
   weeks old and both age and sex matched. All animal experiments were
   performed using protocols approved by the National Heart, Lung and
   Blood Institute (NHLBI) Animal Care and Use Committee and followed NIH
   guidelines for use of animals in intramural research.

Mouse flow cytometry antibodies

   The following antibodies were used to perform flow cytometry on mouse
   cells (all antibodies are from BioLegend unless indicated otherwise):
   CD4 (clone RM4-5), CD25 (clone PC-61), CD122 (clone TM-β1), CD19 (clone
   6D5), CD45RB (clone C363-16A), CD45 (clone 30-F11), CD45.1 (clone A20),
   CD45.2 (clone 104), PD-1 (clone 29F.1A12), CXCR5 (clone L138D7), Bcl-6
   (clone 7D1), Tbet (clone 4B10), BLIMP1 (clone 5E7), FOXP3 (clone
   MF-14), HELIOS (clone 22F6), IL-2 (clone JES6-SH4), IL-10 (clone
   JESS-16E3), IFN-γ (clone XMG1.2), live/dead near infrared (IR)
   ([209]L10119, Thermo Fisher Scientific), pSTAT5 (clone 47/Stat5(pY694),
   BD Biosciences), pAKT (clone M89-61, BD Biosciences), and pERK (clone
   6B8B69). The influenza PR8-specific NP311-325 tetramer conjugated to
   BV421 was provided by the NIH tetramer core facility.

Human flow cytometry antibodies

   The following antibodies were used to perform flow cytometry on human
   cells (all antibodies are from BioLegend unless indicated otherwise):
   anti-CD3 (clone OKT3), CD4 (clone OKT4), CD25 (clone BC96), CCR4 (clone
   L291H4), CD122 (clone TU27), live/dead near IR ([210]L10119 Thermo
   Fisher Scientific), BLIMP1 (clone 6D3), FOXP3 (clone 150D), HELIOS
   (clone 22F6), IL-2 (clone MQ1-17H12), IL-10 (clone JES-9D7), Tbet
   (clone 4B10), THEMIS (clone REA463), PTEN (clone A2B1), pSTAT5 [clone
   47/Stat5(pY694), BD Biosciences], pAKT (clone M89-61, BD Biosciences),
   and pERK (clone 6B8B69).

Antibodies and reagents

   The list of antibodies and reagents used in this study is provided in
   table S6.

CRISPR-Cas9 knockout screening in ED40515(+) cell line

   ED40515(+) cell line with stable Cas9 expression were generated as
   described ([211]77). For sgRNA library construction, the human CRISPR
   pooled genome-wide Brunello sgRNA library ([212]78) (Addgene, #73178)
   was cloned in pLKO-based sgRNA vector (Addgene, #52628) and transformed
   in Stbl4 bacteria (Invitrogen). Lentiviruses were produced in 293FT
   cells by transfecting the sgRNA plasmid library with packaging vectors
   pCMV-VSV-G (Addgene, #8454) and pCMV-dR8.2 dvpr (Addgene, #8455) in a
   10:1:10 ratio using Lipofectamine 3000 reagents according to the
   manufacturer’s instructions. Viral supernatants were harvested and
   concentrated using a Lenti-X concentrator (CloneTech), and the viral
   titer was determined by Lenti-X GoStix (Takara). ED40515(+) cells were
   transduced with the concentrated lentiviral supernatants by
   centrifugation at 3500 rpm for 45 min at 30°C, and the transduced cells
   were selected using puromycin (2 μg/ml) after 48 hours. Cells were
   transduced with a multiplicity of infection of 0.3 to achieve 500×
   coverage of the library after 5 days of puromycin selection. Cells were
   then cultured for 21 days in the presence of IL-2 (200 IU/ml; #Ro
   23-6019, Roche) and doxycycline (200 ng/ml; #D5207, Sigma-Aldrich) to
   induce Cas9 expression. DNA was isolated on day 21 followed by library
   amplification and sequencing on Illumina NextSeq500 platform as
   described ([213]77, [214]79). An average of 150× (126 to 199×)
   sequencing depth was achieved, and data analyzed as described
   previously ([215]77). Sequences were aligned to the sgRNA library,
   allowing for 1–base pair (bp) mismatch using custom scripts and Bowtie
   2 version 2.2.9 with the following parameters: -p 16 -f --local -k 10
   --very-sensitive-local -L 9 -N1. Pairwise differential sgRNA comparison
   between day 0 and day 21 was performed using R package edgeR.
   Differentially expressed sgRNAs were identified and displayed with
   volcano plots to show the biological difference (log[2]FC), and
   statistical significance [−log[10]FDR (false discovery rate)]. Positive
   selection [log[2]FC(D21/D0) > 0] and negative selection
   [log2FC(D21/D0) < 0] indicated enriched to depleted sgRNAs of potential
   interest.

Mouse CD4^+ T cell isolation and culture

   CD4^+ T cells were isolated from mouse spleens using a negative
   selection kit (#19852, Stem Cell Technologies) according to the
   manufacturer’s protocol. Purified CD4^+ T cells were activated with
   plate-bound anti-mouse CD3 (2 μg/ml, #BE0001-1; clone 145-2C11,
   BioXCell) and soluble anti-mouse CD28 (1 μg/ml, #BE0015-1; clone 37.51,
   BioXCell) and cultured in RPMI 1640 medium (#11875093, Gibco)
   supplemented with 10% fetal bovine serum (FBS) (#100-106, GeminiBio), 2
   mM l-glutamine (#25030-149, Gibco), penicillin (50 U/ml) + streptomycin
   (50 μg/ml) (#15070-063, Gibco), and 50 μM 2-mercaptoethanol
   (#21985-023, Gibco) for 72 hours in the presence of recombinant human
   IL-2 (100 IU/ml; #Ro 23-6019, Roche) at 37°C after which the cells were
   harvested and assayed.

Mouse CD4^+Foxp3^+ T[reg] isolation and culture

   CD4^+ T cells were isolated from mouse spleen and lymph nodes by
   negative selection as described above. Foxp3^+ T[reg] cells were sorted
   from enriched CD4^+ T cells suspension based on YFP expression using a
   FACSAria cell sorter (BD Biosciences). Sorted CD4^+Foxp3^+ T[reg] cells
   were activated using plate-bound anti-CD3 (2 μg/ml) and soluble
   anti-CD28 (1 μg/ml) and cultured in complete RPMI 1640 medium with
   recombinant human IL-2 (500 IU/ml) for 72 hours at 37°C following which
   the cells were harvested and analyzed.

Human CD4^+ T cell isolation and culture

   PBMCs were isolated from buffy coats using lymphocyte separation medium
   (#MT25072CI, Corning) based on density gradient centrifugation. CD4^+ T
   cells were purified from PBMCs by negative selection following the
   kit’s protocol (#17952, Stem Cell Technologies). CD4^+ T cells were
   then activated with plate-bound anti-human CD3 (2 μg/ml, #BE0001-2;
   clone OKT3, BioXCell) and soluble anti-human CD28 (1 μg/ml; #BE0248;
   clone 9.3, BioXCell) and cultured in RPMI 1640 medium with 10% FBS, 2
   mM l-glutamine, and penicillin (50 U/ml) + streptomycin (50 μg/ml) in
   the presence of recombinant human IL-2 (100 IU/ml) for 72 hours at
   37°C.

Human CD4^+CD25^+CD127^low natural T[reg] isolation and culture

   PBMCs were isolated from buffy coats of healthy human donors as
   described above. CD4^+CD25^+CD127^low natural T[reg] cells were
   purified by magnetic separation according to the manufacturer’s
   instructions (#18063, Stem Cell Technologies). Purified natural T[reg]
   cells were activated using plate-bound anti-CD3 (2 μg/ml) and soluble
   anti-CD28 (1 μg/ml) and then cultured in complete TexMACS medium
   (#130-097-197, Miltenyi Biotec) supplemented with 5% human AB serum
   (#100-812-100, GeminiBio), 2 mM l-glutamine, and penicillin (50 U/ml) +
   streptomycin (50 μg/ml) for 7 days at 37°C in the presence of
   recombinant human IL-2 (500 IU/ml). The media were replenished every
   alternate day containing recombinant human IL-2 (500 IU/ml).

Retroviral transduction

   The cDNA was polymerase chain reaction (PCR) amplified for Prdm1 gene
   from purified mouse CD8^+ T cells using forward
   (ATCGAGATCTATGAGAGAGGCTTATCTC) and reverse
   (CTAGAGATCTTTAAGGATCCATCGGTT) primers, cloned into pRV-EV ([216]80),
   using restriction enzyme BglII (#RO144S, NEB) and transfected with the
   pCL-Eco retrovirus packaging plasmid (#12371, Addgene) into 293T cells.
   Retroviral supernatants carrying the pRV-EV (empty vector) or pRV-Prdm1
   (vector-expressing Prdm1) plasmids were mixed with polybrene (8 μg/ml;
   #TR-1003, Sigma-Aldrich) and transduced into preactivated mouse CD4^+ T
   cells by centrifugation at 3500 rpm for 45 min at 30°C. The retroviral
   supernatants were replaced with new medium containing human recombinant
   IL-2 (100 IU/ml), and the cells were cultured for 2 days at 37°C and
   harvested for further analysis.

CRISPR-Cas9 deletion of PRDM1

   Four PRDM1-gRNA target sequences [PRDM1 complementary RNA 1 (crRNA1),
   crRNA2, crRNA3, and crRNA4] and four control gRNA sequences (control
   crRNA1, crRNA2, crRNA3, and crRNA4) were selected from the Brunello
   library that was used for the CRISPR-Cas9 knockout screening (table S7)
   and purchased from Integrated DNA Technologies (IDT). First, 80 μM
   crRNA (IDT) and 80 μM trans-activating CRISPR RNA (tracrRNA)
   (#10007810, IDT) were mixed in a 1:1 ratio and incubated for 30 min at
   37°C to generate 160 μM crRNA-tracrRNA duplexes (gRNAs). Cas9 nuclease
   (40 μM, #1081059, IDT) was slowly added to the crRNA-tracrRNA duplexes
   (gRNAs) and incubated for 15 min at 37°C to generate ribonucleoprotein
   (RNP) complexes. For each reaction, 10 × 10^6 stimulated human CD4^+ T
   cells or T[reg] cells were pelleted and resuspended in 20 μl of P3
   primary cell nucleofection solution (#V4XP-3032, Lonza) and 2 μl of
   Cas9-RNP complexes and 0.75 μl of 100 μM electroporation enhancer
   (#1075916, IDT) were added, and the entire volume was transferred to a
   nucleocuvette strip (Lonza). The cells were electroporated using the
   EH-115 program on a Amaxa 4D-nucleofector (Lonza), 80 μl of prewarmed
   RPMI 1640 medium was added to each well after electroporation, and the
   cells were allowed to recover for 30 min at 37°C. The cells were then
   cultured in TexMACS medium supplemented with 5% human AB serum, 2 mM
   l-glutamine, and penicillin (50 U/ml) + streptomycin (50 μg/ml) for 3
   to 4 days at 37°C in the presence of IL-2 (200 IU/ml) for human CD4^+ T
   cells and IL-2 (500 IU/ml) for human T[reg] cells. Gene editing
   efficiency was estimated by performing a T7 endonuclease I assay
   (#M0302S, New England Biolabs) per the manufacturer’s protocol using
   the primers listed in table S8.

Intracellular staining and flow cytometry

   The cells were resuspended in staining buffer [phosphate-buffered
   saline (PBS) with 2% FBS and 0.02% sodium azide] and stained for
   live/dead using LIVE/DEAD Fixable Dead Cell Stain Kit (#[217]L10119,
   Invitrogen) and surface markers for 20 min at room temperature. Cells
   were fixed and permeabilized with Cytofix/Cytoperm (#554714, BD
   Biosciences) for intracellular cytokine staining and
   Foxp3/transcription factor staining buffers (#00-5523-00, eBioscience)
   for intracellular staining of transcription factors according to the
   instructions in the kit. Before staining, cells were stimulated with
   phorbol 12-myristate 13-acetate (50 ng/ml; #P1585, Sigma-Aldrich),
   ionomycin (1 μg/ml; #13909, Sigma-Aldrich), and 1× of protein transport
   inhibitor (#00-4980-03, eBioscience) for 5 hours at 37°C. For pSTAT5,
   pAKT, and pERK staining, the cells were rested overnight in complete
   medium at 37°C and then stained for live/dead and stimulated with
   recombinant human IL-2 for 10 min at 37°C. The cells were immediately
   fixed in 2.5% paraformaldehyde (#043368-9 M, Thermo Fisher Scientific),
   permeabilized in 100% cold methanol (#326950010, Thermo Fisher
   Scientific), and stained with the respective surface markers and
   phospho flow antibodies for 1 hour at room temperature. After staining,
   the cells were acquired using a BD LSR Fortessa X-20 Cell Analyzer flow
   cytometer (BD Biosciences) and data were analyzed with FlowJo software
   (TreeStar). All the antibodies were from BD Biosciences or BioLegend.

PR8 influenza infection mouse model

   One thousand viral focal units of PR8 influenza virus recombinant
   strain [A/Puerto Rico/8/1934 (H1N1)] ([218]81, [219]82) in 100 μl of
   PBS was administered intranasally into Prdm1^fl/fl (WT) and
   Prdm1^fl/flCD4^cre (CKO) mice after anesthesia with ketamine/xylazine.
   Infection was allowed to develop for 10 days after which the mice were
   euthanized, and medLNs and lungs were collected for analysis.
   Lymphocytes were isolated from the tissues as described ([220]83).
   Briefly, the lungs were perfused with 1 ml of PBS, harvested, minced,
   and digested with collagenase (1 mg/ml, Sigma-Aldrich) and
   deoxyribonuclease (1 mg/ml, Sigma-Aldrich) in 3 ml of RPMI 1640 medium
   for 45 min at 37°C. Digested lungs and medLNs were passed through 40 μM
   cell strainer (BD Biosciences) to prepare single-cell suspension.
   Lymphocytes from lungs were obtained on a 44/67% Percoll gradient
   (Sigma-Aldrich) after centrifugation at 2000 rpm for 20 min at 4°C. To
   measure antigen-specific cytokine production, the cells were stimulated
   with 1 mM of influenza NP (311-325) peptide (#AS-62420, AnaSpec, Inc)
   and 1× of protein transport inhibitor (#00-4980-03, eBioscience) for 5
   hours at 37°C. Antiviral T cell responses were evaluated by staining
   the cell suspensions with influenza PR8-specific NP311-325 (MHC class
   II) tetramer conjugated to BV421 (provided by NIH Tetramer Core
   Facility) for 30 min at 37°C followed by surface and intracellular
   staining and assessment by flow cytometry.

Adoptive T[reg] transfer in T cell–induced colitis mouse model

   Naïve (CD4^+CD25^−CD45RB^hi) T cells were sorted from the spleen and
   lymph nodes of CD45.1 congenic mice using a FACSAria cell sorter.
   Simultaneously, (CD4^+Foxp3^+) T[reg] cells were sorted from the spleen
   and lymph nodes of CD45.2 WT (Foxp3^YFP-cre) and CD45.2 CKO
   (Prdm1^fl/flFoxp3^YFP-cre) mice, respectively. Colitis was induced by
   the intravenous injection of 5 × 10^5 naïve CD4^+ T cells into the
   Rag2^−/− mice. To evaluate the in vivo T[reg] suppressive activity in
   the presence or absence of BLIMP1, 1 × 10^5 T[reg] cells from WT or
   Prdm1-CKO mice were cotransferred intravenously with 5 × 10^5 naïve
   CD4^+ T cells into the Rag2^−/− mice to monitor the development of
   colitis. The recipient mice were weighed weekly and euthanized 8 weeks
   after the cell transfer when the signs of colitis such as diarrhea
   appeared or when they lost more than 20% of their initial body weight.
   At the end of the experiment, mesLNs and spleen were examined for the
   protein expression levels by intracellular staining. Proximal, middle,
   and distal colon samples were excised, fixed in 10% formalin
   (#HT501128, Sigma-Aldrich), embedded in paraffin for cross-sectioning
   and stained with hematoxylin and eosin for histology imaging. The
   tissue section images for analysis were obtained using a Hamamatsu
   NanoZoomer slide scanner. Colon pathology scores were determined
   blindly on the basis of the criteria that includes the presence and
   number of crypt abscesses and submucosal inflammation (0 to 3), lamina
   propria cellularity/inflammatory infiltration (0 to 3), epithelial
   hyperplasia and goblet cell depletion (0 to 3), and the percentage of
   tissue involvement (0 to 3); where 0 = normal, 1 = mild, 2 = moderate,
   and 3 = severe ([221]74, [222]80, [223]84).

RNA-seq and analysis

   Purified CD4^+ T cells from medLNs of influenza-infected WT
   (Prdm1^fl/fl) and Prdm1 CKO (Prdm1^fl/flCD4^cre) mice were rested
   overnight in complete medium and restimulated with recombinant human
   IL-2 (200 IU/ml) for 24 hours at 37°C. Cells were harvested and total
   RNA was extracted using Direct-zol RNA MiniPrep kit (#R2052, Zymo
   Research) according to the kit’s protocol. Two hundred nanograms of
   total RNA was used to prepare RNA-seq libraries using the KAPA RNA
   HyperPrep Kit (#KK8542, KAPABIOSYSTEMS) per the manufacturer’s
   protocol. The libraries were barcoded (indexed) and sequenced on an
   Illumina NovaSeq platform. Sequenced reads (50 bp, single end) were
   obtained with the Illumina CASAVA pipeline and mapped to the mouse
   genome (mm10/GRCm38) using Bowtie 2.2.6 ([224]85) and TopHat 2.2.1
   ([225]86). Raw counts that fell on exons of each gene were calculated
   and normalized by using RPKM (reads per kilobase per million mapped
   reads). DEGs were identified using the R Bioconductor package “edgeR”
   ([226]87), using thresholds FDR < 0.05 and |log[2]FC| > =1, and
   expression heatmaps were generated with the R package “pheatmap”
   ([227]88). For gene set enrichment analysis, RNA-seq–based gene
   expression data were compared with molecular signature gene sets using
   R package “fgsea” ([228]89).

   Splenic T[reg] cells from WT (Foxp3^YFP-cre) and Prdm1-CKO
   (Prdm1^fl/flFoxp3^YFP-cre) mice were isolated, preactivated with
   anti-CD3 + anti-CD28, rested overnight, and cultured with IL-2 for 24
   hours. Cells were harvested, total RNA was extracted, and 100 ng of
   total RNA was used to prepare RNA-seq libraries as above. Fifty-bp PE
   reads were produced on an Illumina NovaSeq platform. RNA-seq data were
   processed using the RNA-seek workflow v1.8.0
   ([229]https://doi.org/10.5281/zenodo.5223025) and the NIH HPC Biowulf
   cluster. ([230]http://hpc.nih.gov). In brief, the reads were trimmed
   using cutadapt v1.18 ([231]90) and aligned to the mm10 reference genome
   and GENCODE release M21 using STAR v2.7.6a ([232]91) in two-pass basic
   mode. Expression levels were quantified with RSEM v1.3.0 ([233]92).
   Differential analysis was completed with limma 3.52.4 ([234]93) with
   the following prefiltering parameters: minimum CPM of 0.5 in two
   libraries. Genes were considered significant if they had an FDR less
   than 0.05 and an absolute fold change greater than 2. Pathway analysis
   was assessed on differential genes using ClusterProfiler 4.4.4
   ([235]94) on the GO biological processes database from GO.db 3.15.0
   ([236]95) and the Hallmark mouse database from R package msigdbr 7.5.1
   ([237]96). Related significant GO pathways were collapsed using the
   simplify function. The resulting figures were created using the
   enrichplot package 1.16.2. Heatmaps of the FPKM-normalized data were
   created with ComplexHeatmap 2.12.1 ([238]97).

ChIP-seq and analysis

   Purified human natural T[reg] cells (CD4^+CD25^+CD127^low) from HDs
   were cultured and expanded using T[reg] expansion kit, human
   (#130-095-345, Miltenyi Biotec) in the presence of recombinant human
   IL-2 (500 IU/ml) according to the manufacturer’s instructions. ChIP-seq
   was performed as described earlier ([239]98). Briefly, the cells were
   fixed in 1% formaldehyde for 10 min and sonicated to prepare chromatin
   followed by immunoprecipitation with rabbit anti-IgG (#2729S, Cell
   Signaling Technology) and two different antibodies for BLIMP1:
   anti-BLIMP1 Ab1 (#9115S, Cell Signaling Technology) and anti-BLIMP1 Ab2
   (#MA5-14879, Thermo Fisher Scientific). DNA was end-repaired using an
   End-It DNA-Repair kit (Epicentre), and was indexed, amplified, and
   sequenced on an Illumina HiSeq-2500.

   BLIMP1 ChIP-seq raw reads were processed into BEDPE and bigWig files as
   described previously ([240]99). Replicates were combined for peak
   calling using cLoops2’s callPeaks module ([241]100) with parameters
   -eps 150 and -minPts 10 against the IgG control. Only overlapped peaks
   from both donors and both BLIMP1 antibodies are used for the following
   analysis. Peak targets and distance to nearest gene transcription start
   sites were obtained with anoPeaks.py script from the cLoops2 package,
   and target genes GO terms enrichment analysis was performed by
   findGO.pl in the HOMER package ([242]101). We considered terms with at
   least ten associated target genes and excluded terms containing more
   than 1000 genes, consistent with previous studies ([243]99, [244]102).
   Motif analysis was performed using findMotifsGenome.pl in the HOMER
   package. Genome-browser–like images were presented by the cLoops2 plot
   module.

Ex vivo culture of CD4^+ T cells from patients with ATL

   Frozen PBMCs from HD and patients with acute ATL were thawed and rested
   overnight in complete TexMACS medium supplemented with 5% human AB
   serum, 2 mM l-glutamine, and penicillin (50 U/ml) + streptomycin (50
   μg/ml) at 37°C. CD4^+ T cells were isolated from the PBMCs and
   stimulated using plate-bound anti-CD3 (2 μg/ml) and soluble anti-CD28
   (1 μg/ml) and cultured in complete medium in the presence of
   recombinant human IL-2 (500 IU/ml) for 24 hours at 37°C, after which
   the cells were harvested for the protein expression analysis by flow
   cytometry.

Lentiviral transduction in ATL cells

   Human PRDM1 cDNA was PCR-amplified from mRNA extracted from purified
   human CD8^+ T cell using the following primers: forward
   (5′-ctagGAATTCatgttggatatttgcttg-3′) and reverse
   (5′-gattGCGGCCGCttaaggatccattggttc-3′). The PCR product was digested
   with restriction enzymes EcoRI and NotI, purified, and ligated into the
   pLVX-EF1a-IRES-ZsGreen1 lentiviral vector. The ligation mixture was
   transformed into Escherichia coli DH5α cells, and positive clones were
   selected. Recombinant plasmids were confirmed by sequencing to verify
   correct insertion and orientation of the PRDM1 cDNA. Lentiviruses were
   generated by transfecting the PRDM1 plasmids with the pCMV-VSV-G
   (#8454, Addgene) and pCMV-dR8.2 dvpr (#8455, Addgene) packaging
   plasmids into 293T cells. Lentiviral supernatants carrying the pLVX-EV
   (empty vector) or pLVX-PRDM1 plasmids were mixed with polybrene (8
   μg/ml; #TR-1003, Sigma-Aldrich) and transduced into preactivated mouse
   CD4^+ T cells from frozen PBMCs of patients with ATL by centrifugation
   at 3500 rpm for 45 min at 30°C. The lentiviral supernatants were
   replaced with new medium containing human recombinant IL-2 (100 IU/ml),
   and the cells were cultured for 2 days at 37°C and harvested for
   further analysis.

scRNA-seq and analysis

   Purified CD4^+ T cells from frozen PBMCs of HDs and patients with acute
   ATL were stimulated briefly with plate-bound anti-CD3 (2 μg/ml) and
   soluble anti-CD28 (1 μg/ml) in the presence of recombinant human IL-2
   (500 IU/ml) for 4 hours at 37°C. The cells were harvested and loaded
   into separate channels of a single-cell Chip G with reverse
   transcriptase reagent mixture and 5′ gel beads according to the
   manufacturer’s protocol (10X Genomics; Pleasanton, CA). The chips were
   next loaded into the 10X Genomics Chromium Controller for single-cell
   partitioning, immediately followed by emulsion recovery from the chip
   and incubated in an MJ Research PTC-200 Thermal Cycler for the reverse
   transcription reaction; cDNA isolation and library preparation were
   completed per the manufacturer’s protocol. Isolated cDNA was amplified
   (11 cycles) and was used for the preparation of a gene expression
   library with the Chromium Next GEM Single Cell 3’ Reagent kits v3.1
   (10X Genomics). The quality of the cDNA and library was evaluated using
   the D1000 high-sensitivity kits on a 2200 TapeStation system (Agilent,
   Santa Clara, CA). The gene expression libraries were sequenced on the
   Illumina HiSeq platform. Sequencing data were processed with 10X
   Genomics Cell Ranger version 7.2.0 ([245]103). Cell Ranger “mkfastq”
   function was used to generate fastqs and “count” was used to generate
   gene expression data using the Human reference (GRCh38) - 2020-A
   transcriptome reference. Analyses were primarily done in Seurat version
   5.0.1 ([246]104). Cells containing <500 unique molecular identifiers
   (UMIs) and > 10% mitochondrial gene percentage were filtered out.
   Features found in <3 cells were also removed. Doublets were identified
   using the scDblFinder R package version 1.16 ([247]105) and were also
   filtered out. PRDM1^+ cells were defined as having any detectable
   expression of PRDM1. Data from each individual sample were normalized
   and scaled. Dimensionality reduction (dims = 1:20) was performed with
   “RunPCA” and “RunUMAP” functions in Seurat. Shared nearest-neighbor
   graphs–based clustering was performed using the Louvain algorithm
   (resolution = 0.2) with “FindClusters” function in Seurat. All
   individual samples were integrated together using Seurat’s
   “FindIntegrationAnchors” method with “rpca” reduction and k.weight set
   to 70. The integrated data were then reprocessed with Seurat as
   described above—normalizing, scaling, dimensionality reduction, and
   clustering. Cell type annotation was performed with Azimuth R package
   version 0.4.6 using the provided human PBMC reference ([248]106). Cells
   annotated as “CD4 T” in the level 1 cell type annotation from Azimuth
   were used for downstream analyses. Cells with the level 2 cell type
   annotation as CD4 naïve (n = 321), CD4 proliferating (n = 65), and CD4
   effector memory (n = 18) were merged into the “Other” category as these
   cells were present in fewer numbers. Differential expression analyses
   were performed using the Wilcoxon ranked sum test with the
   “FindMarkers” function in Seurat for the following groups of cell
   populations: ATL versus HD in all cells and PRDM1^+ versus PRDM1^− in
   T[reg] cells. Pathway enrichment analyses were conducted using Qiagen
   IPA software (QIAGEN Inc., [249]https://digitalinsights.qiagen.com/IPA)
   version 107193442 ([250]107). The log[2]FC and P value from each
   differential expression comparison was used as input to the IPA
   software. A P value cutoff of 0.05 was used in the IPA software before
   running pathway enrichment analysis. Pathways from IPA were subset to
   just those with a −log(P value) of >6 and ratio of >0.5, where the
   ratio is the number of genes in the differential expression results
   that are also in the pathway divided by the total genes in the pathway.

Quantification and statistical analysis

   [251]Figures 1A, [252]2A, [253]3A, [254]3F, [255]4G, [256]4J, [257]6A,
   [258]6L, and [259]7A and fig. S1A were created with [260]BioRender.com.
   GraphPad Prism 10 (La Jolla, CA, USA) was used for statistical
   analysis. Two-tailed Student’s t test for comparison of means between
   two groups, and one-way analysis of variance (ANOVA) for comparison of
   means between more than two groups were used. All statistical tests
   were followed by Tukey’s multiple comparison’s posttest. Differences
   were considered statistically significant if the P value was <0.05. The
   data depicted in the bar graphs and scatter dot plots are represented
   as means ± SEM.

Acknowledgments