Extracellular HMGB1 promotes differentiation of nurse-like cells in chronic lymphocytic leukemia

Abstract

Chronic lymphocytic leukemia (CLL) is a disease of an accumulation of mature B cells that are highly dependent on the microenvironment for maintenance and expansion. However, little is known regarding the mechanisms whereby CLL cells create their favorable microenvironment for survival. High-mobility group protein B-1 (HMGB1) is a highly conserved nuclear protein that can be actively secreted by innate immune cells and passively released by injured or dying cells. We found significantly increased HMGB1 levels in the plasma of CLL patients compared with healthy controls, and HMGB1 concentration is associated with absolute lymphocyte count. We therefore sought to determine potential roles of HMGB1 in modulating the CLL microenvironment. CLL cells passively released HMGB1, and the timing and concentrations of HMGB1 in the medium were associated with differentiation of nurse-like cells (NLCs). Higher CD68 expression in CLL lymph nodes, one of the markers for NLCs, was associated with shorter overall survival of CLL patients. HMGB1-mediated NLC differentiation involved internalization of both receptor for advanced glycation end products (RAGE) and Toll-like receptor-9 (TLR9). Differentiation of NLCs can be prevented by blocking the HMGB1-RAGE-TLR9 pathway. In conclusion, this study demonstrates for the first time that CLL cells might modulate their microenvironment by releasing HMGB1.

Figure 1

Detection of extracellular and intracellular HMGB1 protein in CLL samples. (A) HMGB1 and (B) DNA concentrations were determined in the plasma of healthy (n = 14) and CLL samples (untreated, n = 22; relapsed, n = 7; partial remission, n = 7). Significantly increased HMGB1 and DNA concentrations in CLL samples were expressed as median ± interquartile range and differences analyzed by Student t test. (C) Correlation between HMGB1 and DNA concentration in 36 cases of CLL plasma. Dots in the box are 5 cases with the highest concentrations of both HMGB1 and DNA and represent the group of patients with the poorest prognosis. (D-E) Correlations between plasma (D) HMGB1 or (E) DNA concentration with ALC. ALC information was collected at the time of blood draw (supplemental Table 3). (F-H) Determination of HMGB1 expression in CLL and normal B cells by western blotting. (G) Representative results from 2 cases of CLL and 4 normal B-cell samples. (H) HMGB1 expression in long-term cryopreserved CLL samples. Fifty micrograms of proteins was loaded into each lane. Mouse anti-HMGB1 antibody (Sigma-Aldrich) was used at 1:3000 dilution. β-actin was used as a loading control. Numbers below each pair of bands are ratios of HMGB1 to β-actin. (I) Detection of HMGB1 intracellular localization in freshly isolated CLL cells by immunofluorescent staining. DAPI indicates nuclear localization. The isotype control images are demonstrated in supplemental Figure 1A-B.

Figure 2

Determination of HMGB1 and CD68 expression in tissue microarrays. (A) Representative samples of HMGB1 expression in (i-ii) reactive and (iii-vi) CLL lymph nodes. Images were taken with a Leixa DM2500 microscope: i, iii, and v, original magnification ×200); ii, iv, and vi, original magnification ×400. (B) Overall HMGB1 expression. HMGB1 expression was determined by a computerized image analysis Ariol system using pathologist-trained visual parameters. All cores were reviewed manually before scoring, and only intact cores were used for scoring. Each datum represents an average of triplicate cores. There was a statistically significant decrease in expression of percentage of cells expressing HMGB1 in CLL compared with reactive LNs. (C) Cytoplasmic HMGB1 expression. Numbers of cytoplasmic HMGB1-positive cells were counted blindly by 2 reviewers in 4 high power fields (hpfs) and expressed as a mean value from triplicate cores. (D) Overall survival of CLL patients based on low (<50) and high (>50) numbers of cells containing cytoplasmic HMGB1 in 4 counted hpfs, with the cutoff determined using X-tiles software. (E) Expression of CD68 in (i-ii) reactive and CLL-LN with (iiii) lower and (iv) higher CD68 expression. (F) Statistical comparison of CD68 expression in CLL-LN and RA-LN. (G) Overall survival of CLL patients based on low (<10%) and high (>10%) expression of CD68.

Figure 3

NLC in vitro differentiation. Expression of (A) CD68, (B) vimentin, (C) CD163, and (D) CD14 of in vitro differentiated NLCs. Fresh CLL cells or CLL LN single cells were cultured in 4-well chambered slides for 1 to 2 weeks. Fresh CLL cell slides were fixed the day of separation from blood. Slides were costained with a rabbit anti-HMGB1 antibody (green) and DAPI. (E) Determination of NLC marker proteins by flow cytometry. CLL mononuclear cells were cultured for 2 weeks. Cells were fixed/permeabilized and then stained with anti–CD14-FITC, anti–CD68-PE, and anti–CD163-Allophycocyanin or relative isotype controls. The NLC population that is CD19-AF-488 negative and contains larger sizes (forward scatter) and high granulation (side scatter) was selected for marker protein analysis. Empty peaks were those stained with isotype controls and solid peaks were cells stained with specific antibodies. The flow cytometry profiles in the lower panel represent negative expression of 3 markers from selected small CD19-AF-488–positive CLL cells. Data shown are 1 of the typical CLL samples from 3 individual cases studied.

Figure 4

Association between HMGB1 release and NLC differentiation. (A) Association between cell viability and NLC differentiation. Cell viability of each sample was monitored every 2 or 3 days using a Cell Viability Analyzer (Beckman Coulter). The box represents the day on which NLCs were first identified. (B) Observation of NLCs by MTT staining. After 1 or 2 weeks of culture, the cells in suspension were removed. Cells were then cultured with fresh medium containing 0.5 mg/mL MTT for 2 hours. (C) HMGB1 passive release. CLL cells were cultured for 2 days on chambered slides. After fix/permeabilization, cells were costained with rabbit anti-COX IV (red) and mouse anti-HMGB1 (green) antibodies. Arrows indicate a typical phenomenon of HMGB1 passive release from the cells. (D-E) Time course of HMGB1 release determined by (D) ELISA and (E) western blotting in 2 representative cases of 6 independent cases examined. HMGB1 release into culture medium was monitored over 19 days. Conditioned medium was collected every 2 or 3 days and stored at −80°C. Ten or 40 µL of conditioned medium was used for ELISA assay or western blotting, respectively. Western blotting was probed with an anti-HMGB1 mouse antibody first and then reprobed with an anti-LDH rabbit antibody. (G) Determination of HMGB1 release by western blotting. Four CLL samples were cultured for 15 days, and conditioned medium was collected in every 5 days. Numbers of NLCs were quantified by MTT staining at the 15th day. The amounts of NLCs indicated as −, +, ++, +++, and ++++ were 0%, 10%, 25%, 50%, and >75% coverage of the well. (F,H) DNA concentration in the conditioned medium. Two hundred microliters of conditioned medium was used to determine DNA concentration. Boxes in A, D, E, and F indicate the time when NLCs were observed. Numbers in each chart are CLL sample IDs.

Figure 5

Activation of RAGE/TLR9 in the CLL microenvironment in vivo and in vitro. (A) Detection of RAGE and TLR9 expression in CLL cells by western blotting. Ten untreated and 6 relapsed CLL peripheral mononuclear cells were stored at −80°C and lysed directly from the pellets. Fifty micrograms of proteins was loaded onto each lane. Rabbit anti-RAGE antibody and mouse anti-TLR9 antibody were used at 1:3000 dilution. Hsp70 was used as a loading control. (B) Detection of RAGE and TLR9 expression in CLL-LN and RA-LN by immunohistochemical staining as described in the Materials and methods. Images were taken with a Leixa DM2500 microscope (original magnification, ×200). Immunofluorescent costaining of HMGB1/TLR9 and RAGE/TLR9 in (C) fresh CLL cells and differentiated NLCs after (D) 1 and (E) 2 weeks of in vitro culture. Costained primary antibodies used were mouse anti-TLR9 (red)/rabbit anti-HMGB1 (green) and mouse anti-TLR9 (red)/rabbit anti-RAGE (green) antibodies. Secondary antibodies for costaining were Alexa Fluor 488 goat anti-rabbit IgG and Alexa Fluor 546 donkey anti-mouse IgG. Arrows in D show TLR9 and RAGE aggregation at the contacting sites with CLL cells. Empty and solid triangles in E indicate specified vesicles containing HMGB1/TLR9 and RAGE/TLR9. (F) Expression of HMGB1, RAGE, and TLR9. Proteins were extracted from fresh and 2-week cultured CLL cells and NLCs. Protein expression was detected by western blotting using mouse anti-HMGB1, anti-TLR9, or a rabbit anti-RAGE antibody. CD68 and vimentin were used as markers for NLCs, and β-actin was used for a loading control. (G) Co-IP was performed using mouse anti-TLR9 antibody, and the blots were probed with a rabbit anti-RAGE, rabbit anti-HMGB1, and mouse anti-TLR9 antibodies, respectively. This is a typical result from 3 individual experiments performed.

Figure 6

Nuclear translocation of STAT3 and NF-κB. (A-B) Immunofluorescent staining of (A) STAT3-P^S727 and (B) NF-κB in cells before and after cell culture for 1 or 2 weeks. Slides were costained with mouse (A) STAT3-P^S727 or (B) NF-κB antibody, showing the red color and a rabbit anti-HMGB1 antibody, showing green. Yellow color indicates nuclear colocalization between STAT3-P^S727 or NF-κB with HMGB1. (C) Detection of STAT3 and NF-κB activation by cellular fractionation. Cytoplasmic (C) and nuclear (N) proteins were extracted using NE-PER Nuclear and Cytoplasmic Extraction Reagents. Fifty micrograms of proteins was loaded onto each lane. Mouse anti-STAT3-P^S727 or NF-κB antibody was used for determination of nuclear translocation of each protein, and mouse anti-Rb and rabbit anti-LDH antibodies were used as marker for nucleus and cytoplasm, respectively. CD68 was used as a marker for NLCs.

Figure 7

Blocking differentiation of NLCs. (A,C,E) CLL cells were cultured for 1 week individually with different neutralizing antibodies or inhibitors. Suspending CLL cells were removed gently prior to adding the MTT solution. After incubation with MTT for 2 hours, a large amount of formazan was formed inside NLCs. Images of NLCs were taken under phase-contrast microscopy, which are displayed as larger dark cells on the bottom of plates. (B,D,F) To quantify the amount of NLCs, formazan in the cells was dissolved with isopropanol, and OD values were measured by spectrophotometry. Reduced numbers of NLCs were expressed as percentage of control. Data shown are mean ± standard deviation from 3 individual cases CLL patients with triplicate OD values. Significant changes were analyzed by the Student t test. (A-B) Blocking HMGB1, RAGE, or TLR9. Fresh CLL cells (5 × 10⁶/mL) in 24-well plates were incubated with 10 µg/mL of anti-HMGB1 or anti-RAGE neutralizing antibody or 100 nM G-iODN for 1 week. (C-D) Inhibition of HMGB1 by EP. Cells were treated with 1.0, 2.5, and 5.0 mM of EP for 1 week. (D) Viability of CLL cells were determined by a Vi-Cell XR Cell Viability Analyzer, and MTT data represent the amount of NLCs before and after treatment with EP. (E-F) Blockade of RAGE, IL-6, or TLR4 by neutralizing antibodies. CLL cells were incubated with 10 µg/mL of anti-RAGE, anti–IL-6, or TLR4 neutralizing antibody for 1 week.