FOXP2-positive diffuse large B-cell lymphomas exhibit a poor response to R-CHOP therapy and distinct biological signatures

Abstract

FOXP2 shares partially overlapping normal tissue expression and functionality with FOXP1; an established diffuse large B-cell lymphoma (DLBCL) oncogene and marker of poor prognosis. FOXP2 is expressed in the plasma cell malignancy multiple myeloma but has not been studied in DLBCL, where a poor prognosis activated B-cell (ABC)-like subtype display partially blocked plasma cell differentiation. FOXP2 protein expression was detected in ABC-DLBCL cell lines, and in primary DLBCL samples tumoral FOXP2 protein expression was detected in both germinal center B-cell-like (GCB) and non-GCB DLBCL. In biopsies from DLBCL patients treated with immunochemotherapy (R-CHOP), ≥ 20% nuclear tumoral FOXP2-positivity (n = 24/158) correlated with significantly inferior overall survival (OS: P = 0.0017) and progression-free survival (PFS: P = 0.0096). This remained significant in multivariate analysis against either the international prognostic index score or the non-GCB DLBCL phenotype (P < 0.05 for both OS and PFS). Expression of BLIMP1, a marker of plasmacytic differentiation that is commonly inactivated in ABC-DLBCL, did not correlate with patient outcome or FOXP2 expression in this series. Increased frequency of FOXP2 expression significantly correlated with FOXP1-positivity (P = 0.0187), and FOXP1 co-immunoprecipitated FOXP2 from ABC-DLBCL cells indicating that these proteins can co-localize in a multi-protein complex. FOXP2-positive DLBCL had reduced expression of HIP1R (P = 0.0348), which is directly repressed by FOXP1, and exhibited distinct patterns of gene expression. Specifically in ABC-DLBCL these were associated with lower expression of immune response and T-cell receptor signaling pathways. Further studies are warranted to investigate the potential functional cooperativity between FOXP1 and FOXP2 in repressing immune responses during the pathogenesis of high-risk DLBCL.

Keywords: FOXP2; diffuse large B-cell lymphoma; survival.

Figure 1. FOXP2 is expressed in ABC-DLBCL cell lines

(A) Western blotting of nuclear lysates from GCB-DLBCL, ABC-DLBCL and myeloma cell lines with anti-FOXP2 (clone 73A/8) and anti-NPM (clone NA24) antibodies. An overexposed image (middle panel) shows very weak FOXP2 expression in OCI-Ly3 and just detectable expression in RCK8. *Indicates an upper band that is thought to be non-specific. MIEU remained negative in other experiments with more abundant sample loading. (B) qRT-PCR for FOXP2 transcripts in GCB- and ABC-DLBCL cell lines. FOXP2 expression was normalized according to the expression of 18S rRNA housekeeping gene and expressed relative to JJN3 (set to 100%). (C) Immunohistochemical labeling of FOXP2 nuclear expression in DLBCL cell lines: showing negative (Karpas 422), weak (OCI-Ly3, RCK8) and strongly positive (OCI-Ly10, RIVA, SU-DHL-2) examples. Higher power insets are shown in the bottom right corners of each panel.

Figure 2. FOXP2 expression confers inferior survival in R-CHOP-treated DLBCL cases

(A) Immunohistochemical labeling of FOXP2 expression in primary DLBCL cases with variable intensity and frequency of expression. Left to right these were scored as negative, intensity 1/frequency 55%, intensity 2/frequency 80%, intensity 3/frequency 100%. (B) Frequency distribution of FOXP2 (left panel) and BLIMP1 (right panel) at each 10% cut-off level in DLBCL cases (n = 158). (C–H) Overall survival (OS) and progression free survival (PFS) of DLBCL cases (n = 158) according to FOXP2 frequency at 20% cut-off (C–D) or intensity (E–F), and BLIMP1 frequency at 40% cut-off (G–H).

Figure 3. FOXP2 expression is not associated with non-GCB DLBCL primary cases, and it confers worse OS in GCB or non-GCB DLBCL subtypes

(A–B) Relationship of FOXP2 (A) or BLIMP1 (B) protein expression frequency with GCB or non-GCB DLBCL subtype according to Hans (n = 158), Choi (n = 155) or Visco-Young (n = 157) immunohistochemical algorithms. (C–F) OS (C–D) and PFS (E–F) in GCB-DLBCL (n = 89) or non-GCB DLBCL (n = 69) according to FOXP2 20% cut-off. The DLBCL cases were stratified according to Hans algorithm.

Figure 4. FOXP2 is co-immunoprecipitated by FOXP1 and their co-expression confers worse survival in R-CHOP-treated DLBCL cases

(A) Correlation of FOXP1 and FOXP2 expression frequency in DLBCL (n = 157). (B) Relationship of FOXP2 frequency with FOXP1^hi/HIP1R^lo (n = 36), FOXP1^lo/HIP1R^hi (n = 58) or non-reciprocal (n = 62) DLBCL cases. (C–D) Co-immunoprecipitation of endogenous FOXP1 in RIVA and SU-DHL-2 cell lysates with anti-FOXP1 JC12 antibody or with non-specific IgG2a isotype and 130/B4 (“in house” anti-NFIL3) control antibodies. Input indicates non-immunoprecipitated cell lysates. JC12 blotting; top arrow indicates FOXP1_FL, bottom arrow indicates FOXP1_s (C). Co-immunoprecipitated FOXP2 was detected by Western blotting the JC12 immunoprecipitated proteins with the FOXP2-73A/8 antibody (D).

Figure 5. Genes most highly-correlated with FOXP2 protein expression in primary DLBCL cases

(A–C) Top 20 genes positively- or inversely-correlated with FOXP2 protein positive expression in all DLBCL cases (n = 39; A), GCB-DLBCL (n = 24; B) or ABC-DLBCL (n = 15; C) subtype. The cases were stratified according to FOXP2 protein positive (i.e. absolute positivity without regarding frequency cut-off) or negative (i.e. 0% frequency). FOXP2 protein was positive in six (of 24) GCB-DLBCL and five (of 15) ABC–DLBCL cases.

Figure 6. TCR signaling and immune response signatures were significantly reduced in FOXP2-positive ABC-DLBCL cases

(A) Gene ontology and pathway signatures suppressed in FOXP2-positive DLBCL cases compared with their FOXP2-negative counterparts. All DLBCL cases (n = 39) or ABC-DLBCL (n = 15) molecular subtype were analyzed separately. The bar graph was constructed on a negative log₁₀ scale with higher -log₁₀ value denoting more significant FDR value. The linear FDR values were shown next to or within each bar, and the horizontal dotted line denotes linear FDR threshold of 0.05. GO: Gene Ontology; PID: Pathway Interaction Database; RID: Reactome ID. (B) Heatmap visualization of genes involved in TCR and immune response signatures compared with FOXP2 microarray transcript expression (probe: 1555516_at) in FOXP2 protein positive or negative GCB-DLBCL (n = 24) or ABC-DLBCL (n = 15) cases. Line graphs comparing FOXP2 transcript expression with the averaged expression values of TCR/immune response genes (the region shaded in turquoise represents the standard deviation of the averaged values) were plotted corresponding to each case present on the heatmap.