Tel.: +39-0522-295777, Fax: +39-0522-295743
Cancer Cell Biology
Resveratrol-mediated apoptosis of hodgkin lymphoma cells involves SIRT1 inhibition and FOXO3a hyperacetylation
Article first published online: 7 AUG 2012
Copyright © 2012 UICC
International Journal of Cancer
Volume 132, Issue 5, pages 1013–1021, 1 March 2013
How to Cite
Frazzi, R., Valli, R., Tamagnini, I., Casali, B., Latruffe, N. and Merli, F. (2013), Resveratrol-mediated apoptosis of hodgkin lymphoma cells involves SIRT1 inhibition and FOXO3a hyperacetylation. Int. J. Cancer, 132: 1013–1021. doi: 10.1002/ijc.27748
- Issue published online: 20 DEC 2012
- Article first published online: 7 AUG 2012
- Accepted manuscript online: 26 JUL 2012 12:54AM EST
- Manuscript Accepted: 11 JUL 2012
- Manuscript Received: 28 OCT 2011
- G.R.A.D.E. Onlus Association, Reggio Emilia, Italy
- Cancer League, France
- Hodgkin lymphoma;
- germinal-center B lymphocytes
Resveratrol (RSV), a plant-derived stilbene, induces cell death in Hodgkin lymphoma (HL)-derived L-428 cells in a dose-dependent manner (IC50 = 27 μM, trypan blue exclusion assay). At a lower range (25 μM), RSV treatment for 48 hr causes arrest in the S-phase of the cell cycle, while at a higher concentration range (50 μM), apoptosis can be detected, with activation of caspase-3. The histone/protein deacetylase SIRT1 has been described as a putative target of RSV action in other model systems, even though its role in cancer cells is still controversial. Here we show that RSV, at both concentration ranges, leads to a marked increase in p53, while a decrease of SIRT1 expression level, as well as enzyme activity, only occurred at the higher concentration range. Concomitantly, however, treatments at both concentration ranges resulted in a marked increase in K373-acetylated p53 and lysine-acetylated FOXO3a. Immunohistochemical stainings of human lymph nodes show a preferential distribution of SIRT1 in the germinal center of the follicles while the mantle zone shows nearly no staining to few positive cells. The classical HL-affected lymph nodes show a strong positivity of the diagnostic Hodgkin Reed-Sternberg cells. Notably, both the HL-derived cell lines and the Hodgkin Reed-Sternberg cells of the affected lymph nodes derive from germinal center-derived B cells. The study of SIRT1 distribution and expression on a larger number of biopsies might disclose a novel role for this histone/protein deacetylase as therapeutic target.
Plant polyphenols are a class of natural molecules well known for their wide range of beneficial properties.1–3 Resveratrol (RSV) is a polyphenol belonging to the class of the stilbenes present in many vegetables and fruits including grapes. The manifold properties of RSV span from chemopreventive action to antioxidant activity, promotion of tissue differentiation, modulation of adipogenesis and antiproliferative effect in several tumoral experimental models.4–15
The information currently available concerning the effects of this natural chemopreventive agent on human lymphoma cells and experimental models is limited. Although there are reports on the antiproliferative activity of RSV on different leukemia and lymphoma cell lines,16–18 no data are available to date in the literature concerning the effects of RSV on Hodgkin lymphoma (HL) cell lines. HL is a tumor frequently affecting adolescents and young adults and represents about 11% of all the diagnosed lymphomas in the United States.19, 20 Furthermore, no studies addressed the involvement of the histone/protein deacetylase SIRT1 (one of the key RSV targets) and of its downstream substrates in lymphomas.
Therefore, the initial aims of our work have been the assessment of the potential antiproliferative activity of RSV in the HL-derived L-428 cell line. Next, we examined various apoptosis criteria and cell cycle changes following the RSV treatment. We demonstrate the dose-dependent increase of the proportion of early and late apoptotic cells and the involvement of some key apoptotic mediators including caspase-3 (casp-3).
The second goal of our work was the assessment of SIRT1 involvement in the RSV-mediated apoptosis of lymphoma cells. SIRT1 is a class III NAD+-dependent histone/protein deacetylase and is the mammalian homolog of yeast Sir2.21, 22 This enzyme plays a role in a great variety of processes related to cancer. This fact is not surprising considering that the acetylation/deacetylation balance is a key feature of epigenetic gene expression modulation also playing a role in carcinogenesis.21, 23, 24 Furthermore, the function of many proteins is post-translationally regulated through an acetylation/deacetylation mechanism. Two relevant examples are the oncosuppressor p53 and the forkhead family of transcription factors (FOXO) that are SIRT1 substrates. Acetylated-p53 is the active form of p53, and acetylation of FOXO3 increases in response to oxidative stress. Through its GADD45 and Bim downstream targets, FOXO3 controls the balance between stress resistance and apoptosis.22, 25, 26
SIRT1 has been reported to be a key target of RSV in several human tissues and solid tumor models although some of the data are still controversial, especially the evidences concerning the postulated role of RSV as a SIRT1 activator in cancer.27–30 We therefore decided to investigate whether SIRT1 is also involved in the RSV-mediated antiproliferative effect in the L-428 cell line. We observed that RSV treatment led to arrest of the cell cycle and apoptosis. SIRT1 protein level and deacetylase enzyme activity decreased at a higher concentration range of RSV.
Limited evidence is available concerning the expression and distribution of SIRT1 in human lymph nodes and lymphomas to date and is completely lacking in classical HL.31 Thus, we decided to investigate its expression in human lymph node biopsies and assessed a series of 30 paraffin-embedded tissue sections by immunohistochemistry in order to gather preliminary information on the in vivo distribution of SIRT1. The data obtained on this initial group of patients suggest a selective localization of SIRT1 in the germinal center of reactive lymph nodes and in the malignant cells of HL lymph nodes. This observation, not yet reported in the literature, might disclose a novel role of SIRT1 during lymphocyte differentiation and malignant transformation in the germinal center and therefore merits further investigation.
Material and Methods
Cell lines and human tissue sections
L-428, Hodgkin lymphoma-derived cell line was purchased from DSMZ (German Collection of Microorganisms and Cell Cultures Braunschweig, Germany). Cells were grown in RPMI medium (Euroclone, Milano, Italy) supplemented with penicillin/streptomycin (Euroclone), L-glutamine (Euroclone) and 10% heat inactivated fetal calf serum (FCS, Gibco-Invitrogen, San Giuliano Milanese, Italy) in a humidified incubator at 37°C with a 5% CO2 atmosphere. Formalin-fixed paraffin-embedded tissue sections were stored and classified by the Pathology Unit of the IRCCS Arcispedale S. Maria Nuova, Reggio Emilia, Italy, and used in accordance with to the Ethics Committee requirements.
Cell counts and fluorescence microscopy
Trypan blue exclusion assay was used to prepare the dose-response curve after RSV treatment. The data were analyzed and the IC50 values calculated with Graph Pad Prism 5.0 (GraphPad software, La Jolla, CA). Hoechst dye (Hoechst 33342, Sigma-Aldrich, Milano, Italy) was used to stain L-428 nuclei. Briefly, 3 x 105 cells were fixed with 2% paraformaldeyde in PBS buffer and, after washings, cells were seeded on a microscope glass through cytospin at 113 g for 5 min (Cytospin 4, Thermo Scientific, Milano, Italy). Spots were stained with Hoechst 5 μg/ml (10 min at room temperature in the dark) and images were acquired through a fluorescence microscope (Olympus BX41, Milano, Italy. Image acquisition with Cytovision software Leica, Wetzlar, Germany).
Cell cycle analysis was carried out by propidium iodide staining (Sigma-Aldrich). Briefly, 2 x 105 cells were fixed with ethanol 70% and stained with a solution of 20 μg/ml propidium iodide + 0.2 mg/ml RNAse A (Qiagen, Milano, Italy). Control and stained samples were acquired through a Cytomics FC500 flow cytometer (Beckman Coulter Inc., Fullerton, CA). DNA histograms were analyzed by a modified sum of Gaussians method as previously described.32 Annexin V/propidium iodide (AnnV/P.I.) apoptosis assay (Bender Medsystems, Annexin V-FITC Apoptosis Detection Kit, Milano, Italy) was performed accordingly to manufacturer's instructions and samples were acquired through a Cytomics FC500 flow cytometer (Beckman Coulter).
Immunoblottings, immunoprecipitation and antibodies
Total cell extracts were prepared as follows. Cells were centrifuged (244 g for 6 min, Eppendorf 5810R centrifuge, Milano, Italy) and the supernatant discarded. The samples were washed twice with PBS 1X pH 7.2 and, after centrifugation, the buffer was removed carefully. Cells were resuspended in cell lysis buffer (Tris HCl 50 mM pH 8.0, KCl 50 mM, EDTA 10 mM, NP40 1%) supplemented with Protease Inhibitor Cocktail (P2714, Sigma-Aldrich). Cell lysis was accomplished through pipetting and incubating at 4°C for 40 min with shaking. Samples were stored at −20°C O/N and centrifuged, the day after, at 17,900g and 4°C for 20 min. The cell lysate was then quantified by Bio Rad Dc Protein assay (Bio Rad, Segrate, Milano, Italy), aliquoted and stored at −20°C until use. SDS-PAGE under reducing conditions was performed in 10% polyacrylamide gels, which were either hand-made or precast and proteins were blotted on Hybond-ECL nitrocellulose membrane (GE Amersham, Milano, Italy). Primary antibodies: mouse anti-p53 mAb (clone PAB1801, Zymed, San Francisco, CA) diluted 1:2,000; mouse anti-β actin (clone AC-15, Sigma-Aldrich) diluted 1:20,000; rabbit anti-Bax mAb (clone E63, Epitomics, Milano, Italy) diluted 1,000; mouse anti-pro+active casp-3 (clone 31A1067, Abcam, Cambridge, UK) diluted 1:500; rabbit anti-SIRT-1 (clone E104, Epitomics) diluted 1:1,000; rabbit polyclonal anti-FOXO3a (clone ab47409, Abcam) diluted 1:1,000; rabbit anti-acetyl K373 p53 (clone EP356(2)AY, Epitomics) diluted 1:1,000; rabbit anti-acetyl Lys (#9441, Cell Signaling, Milano, Italy) diluted 1:1,000. Secondary antibodies: anti-mouse IgG HRP-linked and anti-rabbit IgG HRP-linked (GE Amersham) were diluted 1:20,000 and 1:10,000 respectively. The protein extract for immunoprecipitation was prepared with the following lysis buffer: Tris 50 mM pH 8.0, KCl 50 mM, EDTA 10 mM, NP40 1%. Cell lysates were cleared by centrifugation as described above before starting the immunoprecipitation procedure. Immunoprecipitations with anti-FOXO3a were performed on total cell lysates by means of Protein A agarose (Thermo Scientific). Briefly, 200 μl of the cell lysate was incubated with 20 μg of antibody and kept at 4°C O/N rotating. The following day the samples were incubated 2 hr with protein A agarose, washed, and the immunoprecipitated product eluted. Samples were then loaded onto 10% polyacrylamide gels and SDS-PAGE performed. ECL western blotting detection reagent (GE Amersham) was used according to the manufacturer's instructions.
Colorimetric and fluorimetric enzymatic assays
Caspase-3 enzymatic activity was measured through a colorimetric assay on the total protein extracts used for immunoblottings according to the manufacturer's instructions (Caspase-3 colorimetric assay, Merk-Millipore, Milano, Italy). SIRT1 deacetylase activity was measured with a fluorimetric assay according to the manufacturer's instructions (SIRT1 assay kit, Sigma-Aldrich). The data were acquired through a fluorometer Tecan Infinite M200 (Tecan, Männendorf, CH).
Immunocytochemistry and immunohistochemistry
Immunochemistry on L-428 cells was performed on fresh, cytospin prepared slides (113 g for 5 min, Cytospin 4, Thermo Scientific). Immunohistochemistry was performed on 4 μm formalin-fixed deparaffined tissue sections. Staining was performed by using rabbit monoclonal anti-SIRT-1 antibody (clone E104, Epitomics) diluted 1:50 or mouse anti-Ki67 (MIB-1) (DAKO, Milano, Italy) diluted 1:100. The tissue sections were pre-treated with heat antigen retrieval in sodium citrate buffer for 25 min and the immunohistochemical staining was developed with UltraVision Quanto Detection System HRP (Thermo Scientific). The immunohistochemical stains have been evaluated independently by two Pathologists (I.T. and R.V.).
Software and statistics
UN-SCAN-IT software (Silk Scientific, Orem, UT) was used to perform the quantification of the immunoblotting bands. Graphs for the densitometric analysis after immunoblotting, for the growth curves and for the IC50 calculation were prepared with Graph Pad Prism 5.0 (GraphPad software, La Jolla, CA). Flow cytometric data were analyzed through the CXP analysis software (Beckman Coulter, Fullerton, CA). Fluorescence microscopy images were obtained using CytoVision software (Leica). Adobe Photoshop and Illustrator CS4 (Adobe, San Jose, CA) were used to format the figures.
Resveratrol inhibits L-428 cell growth in a dose-dependent fashion and induces apoptosis
We first investigated the antiproliferative activity of RSV in L-428 cells after 48-hr incubation, and we observed a clear dose-dependent growth inhibition by trypan blue exclusion assay (Fig. 1a). The average IC50, calculated from five independent replicas, was 27 μM. Hoechst staining of cell nuclei following RSV treatment showed the appearance of apoptotic bodies, starting at 10 μM, and this effect became evident at 50 μM (Fig. 1b).
Resveratrol-mediated accumulation of L-428 cells in the S phase and induction of early and late apoptosis
We next investigated the effects of RSV on the cell cycle in L-428 cells, and found that treatment with 25 μM leads to accumulation of the cells in the S phase with a concomitant decrease in the G1 and G2/M phases (Fig. 2a). Treatment with 50 μM RSV resulted in the appearance of a sub-G1 peak (indicative of apoptosis) and a further decrease of the cells in the G2/M phase (indicative of a decreased number of dividing cells, Fig. 2a).
To confirm the onset of apoptosis following RSV treatment, cells either treated or not were double-stained with Annexin V (Ann V)-propidium iodide (P.I.) and analyzed by flow cytometry. The proportion of early (Ann V+/P.I.−) and late (Ann V+/P.I.+) apoptotic cell populations increased after 48-hr incubation with RSV in a dose-dependent fashion (Fig. 2b). Specifically, a 15.5-fold-increase of the Ann V+/P.I.− and a 2.1-fold increase of the Ann V+/P.I.+ populations were observed after treatment with 25 μM, while at the dose of 50 μM RSV, the increase was 19.4-fold for Ann V+/P.I.− and 4.4-fold for the Ann V+/P.I.+ populations in independent experiments. Interestingly, only a slight increase in the necrotic Ann V−/P.I.+ population was visible after 25 or 50 μM RSV treatment, consistent with the very low toxicity to human cells observed with this natural drug.14
p53, Bax and caspase-3 are involved in resveratrol-mediated apoptosis
We further characterized the molecular changes induced during the antiproliferative action by RSV (Fig. 3a) and show that after a 48 hr treatment with 50 μM RSV there was a marked increase in the protein expression levels of p53 and cytosolic Bax (Fig. 3b). Caspases are the executioners of apoptosis and casp-3 (one of the effector caspases) is activated by proteolytic cleavage into two smaller fragments once the cell death process has been triggered.33 As demonstrated by the western blot in Figure 3b, RSV treatment at 50 μM resulted in a decrease of the pro-casp-3 level. The fact that pro-casp-3 was not affected under treatment with 25 μM RSV suggests that cell death induced in L-428 cells at this RSV concentration is not a classical type of caspase-dependent apoptosis. We also show that the 50 μM RSV- and 1 μM doxorubicin-treated L-428 cells displayed a higher casp-3 enzymatic activity (about 87 and 109%, respectively, Fig. 3c) thus paralleling the decrease of pro-casp-3. These observations are consistent with the ones previously reported for myeloma and Burkitt's lymphoma cell lines by Shimizu et al.16
Resveratrol treatment leads to decreased SIRT1 activity and hyperacetylation of p53 and FOXO3a in L-428 cells
Many researchers reported that one of the targets for RSV activity is the class III NAD+-dependent histone and lysine dacetylase SIRT1.28, 29 SIRT1 is believed to be a downstream mediator for BRCA1 transcription factor in breast cancer and is thought to be activated by RSV in breast cancer cell lines.27, 29 We therefore decided to assess whether there is an involvement of SIRT1 in RSV-mediated apoptosis of HL lymphoid cells, a model where SIRT1 had not been investigated before. SIRT1 localizes, as expected, in the nuclei of L-428 cells while it is absent in the cytoplasm (Fig. 4a). Here we show that the treatment with 50 μM RSV partially decreases SIRT1 protein levels. Interestingly, also the enzyme activity of SIRT1 as measured with a fluorimetric assay is inhibited (Figs. 4b and 4c).
Acetyl-p53 is the active form of the oncosuppressor p53 which is a substrate of SIRT1.21, 22, 34 It is therefore logical to expect an increase of the acetylated form concomitant to the apoptosis onset and to the increase of total p53 shown in Figure 3b. RSV treatment resulted in the increase of acetyl-p53 consistent with the deacetylase activity inhibition previously shown (Fig. 4d). FOXO3a is also a well known substrate for SIRT1 and its activity is post-translationally modulated by acetylation/deacetylation processes.25, 34, 26 Immunoprecipitation followed by western blot with anti-acetyl Lys antibody shows the dose-dependent increase of acetyl-FOXO3a over the total immunoprecipitated FOXO3a (Fig. 4d).
These data collectively suggest that this polyphenol is a SIRT1-inhibitor, at least at high doses, and support the recently published biochemical data that show that RSV does not interact with SIRT1 and could not be a SIRT1 activator.35 The role of SIRT1 in cancer is context-dependent and some authors suggest that SIRT1 signaling in cancer cells may promote malignant growth.36 This is consistent with the hypothesis that RSV-mediated growth inhibition involves SIRT1 inhibition.
SIRT1 co-localizes with the germinal center of the normal lymph nodes and is expressed in the nuclei of the malignant HRS cells
Since there are no data available in the literature concerning the expression of SIRT1 in normal lymph nodes and in HL-affected lymph nodes, we next investigated the expression of SIRT1 on a total of 30 paraffin embedded lymph node sections by immunohistochemistry. We found that all the 12 reactive, non-neoplastic lymph nodes examined show a marked SIRT1 positivity localized in the follicles of the organ and, specifically, in the germinal centers (GC) (Fig. 5a). GC is the area where activated B lymphocytes concentrate and proliferate after an antigenic stimulation and is a key source of effector B cell population and humoral immunity.37 By comparison, we show the staining pattern of the proliferation index Ki67 that is strongest in the centroblasts of the dark zone of the GC (Fig. 5b).
GCs are clearly identifiable in normal and reactive lymph nodes whereas they disappear in case of a full colonization by HL.38 Here we show that, in the 18 HL examined, most of the Hodgkin Reed Sternberg (HRS) cells are strongly positive for SIRT1 and that surrounding cells are moderately positive to negative for this marker (Fig. 5c). HRS cells are clearly identifiable in that they are large and often bi-nucleated. They derive from immature B lymphocytes of the GC that do not undergo apoptosis despite their defects.39
Resveratrol (RSV) is a natural polyphenol with promising anticancer features, as shown by preclinical data and recent pioneering clinical trials.40, 41 Much evidence has been published on the RSV antiproliferative activity in vitro and its putative molecular targets, however, no data were available concerning HL cell lines, and some critical information is missing about the mechanism of action of RSV in lymphoid cells. Our findings demonstrate the antiproliferative effects of RSV on the HL cell line L-428 in terms of growth inhibition, cell cycle changes and finally, apoptosis induction. We show that RSV exerts a marked antiproliferative activity with an IC50 of 27 μM and that this is achieved through an S phase cell cycle arrest at 25 μM followed by apoptosis and cytotoxicity at 50 μM after 48 hr of incubation. Fluorescence microscopy shows the appearance and the dose-dependent increase of apoptotic bodies (Figs. 1 and 2). At the molecular level, this is accompanied by total p53 and cytosolic Bax increase and by casp-3 activation. Casp-3 is an aspartic acid-specific cysteine protease synthesized as inactive proenzyme (pro-casp-3) and cleaved during the activation process by the initiator caspase-933. Caspase activation is considered to be the sign of final commitment to cell death and a hallmark of apoptosis. We demonstrate the RSV-mediated dose-dependent decrease of the 32 kDa pro-casp-3 level and the concomitant increase in casp-3 activity, consistent with the observed apoptosis onset (Fig. 3).
Among the several suggested targets involved in the RSV mechanism of action there is the Sirtuin deacetylases family member SIRT1.21, 22 Many authors reported that RSV is an activator of the class III NAD+-dependent histone/protein deacetylase SIRT1.28, 29, 42 However, some recent literature demonstrates that the interactions between RSV, SIRT1 and apoptosis are indeed more complex.27, 43 In particular, a recently published biochemical study addresses the issue of the interaction between RSV (and other RSV-analogues) and SIRT1.35 The conclusion of this research is that RSV is not a direct activator of SIRT1 and therefore some mediators are probably involved in this interplay. Furthermore, a recent review suggests that SIRT1 may promote malignant cell-growth after transformation of the cell since the pathways that serve to protect normal cells may also prevent the death of cancer cells.36
These data prompted us to study the involvement of this key deacetylase during the RSV antiproliferative activity in the L-428 model system. We localized SIRT1 by immunocytochemistry and, as expected, the positivity is nuclear in L-428 (Fig. 4a). We show that the treatment with RSV partially decreases SIRT1 protein level (Fig. 4b). Next, we investigated whether the enzymatic activity of SIRT1 is affected by RSV and found that a limited but consistent inhibition is evident in total cell extracts after 48 hr treatment with 50 μM RSV (Fig. 4c). Furthermore, the study of the acetylation of two key SIRT1 downstream mediators (p53 and FOXO3a) revealed that RSV treatment increases acetyl-p53 (K373) and acetyl-FOXO3a suggesting that deacetylase inhibition is a mechanism involved in RSV-mediated apoptosis of lymphoma cells (Fig. 4d). Since SIRT1-mediated deacetylation inactivates p53 and FOXO3a transcriptional activity, our results are consistent with the fact that RSV-induced acetylation activates the oncosuppressor p53 and the FOXO3a transcription factor during apoptosis onset.25, 26 However, it remains to be investigated whether RSV can exert a direct effect on SIRT1 or if instead some mediators are involved.
The immunohistochemical data collected on 12 normal lymph nodes and 18 HL-affected lymph nodes show an interesting SIRT1 staining pattern (Fig. 5). SIRT1 immunostain in reactive, non-neoplastic lymph nodes is localized in areas such as the germinal centers (GC) and the mantle zone. The strongest intensity of immunostaining is seen in the GC; in particular, SIRT1 highlights the so-called “dark zone” of the follicles, where most of the centroblasts resides. The centroblasts are the most proliferating cells of the GC and are characterized by the higher Ki67 (proliferating index) compared to the “light zone” of the follicle (where the Ki67 is lower)38 (Fig. 5b). A fainter intensity of staining for SIRT1 is seen in some cells of the mantle zone and the marginal zone of the non-neoplastic lymph nodes (Fig. 5a). GC reactions provide a cellular milieu for the affinity maturation of antibody responses but they also bear the risk of generating autoreactive B-cell clones and B-cell lymphomas.37
In HL samples, SIRT1 immunostain is strongly positive in the nuclei of both the diagnostic Reed-Sternberg and Hodgkin cells. In this setting, SIRT1 is also positive in a proportion of the reactive lymphocytes that surround the neoplastic cells (Fig. 5c).
Albeit preliminary, our observations are consistent with the hypothesis that SIRT1 is a marker of GC-derived B lymphocytes being expressed in the nuclei of a HL-derived cell line, in the GC of the normal lymph nodes and in the tumoral cells of the HL-affected lymph nodes (Figs. 4a, 5a and 5c).
Overall, the existing literature describing the relationship between cancer, RSV and SIRT1 is still controversial and far from being definitive. We hypothesize that, due to the wide distribution of SIRT1 among human tissues and due to the pleiotropic activity of this enzyme, the relationship between RSV and its target SIRT1 could be model system-specific. To better determine the SIRT1 cell-specific distribution and role in lymph nodes, a greater number of specimens will be investigated and analyzed in terms of correlation with pathological characteristics and prognosis of the patients. The collected data may provide novel insights concerning the potential of SIRT1 as a therapeutic target.
The authors are grateful to Dr. Massimo Broggini and Dr. Paolo Ubezio (Istituto di Ricerche Farmacologiche M. Negri, Milano, Italy) for their useful support with instrument and data analysis.
- 20Howlader NNA, Krapcho M, Neyman N, et al., eds. SEER Cancer Statistics Review, 1975-2008: National Cancer Institute, Bethesda, MD. SEER website, 2011.
- 38Hematopathology. In: Elsevier. ed., 2011. 100–6., , , et al.
- 39Molecular biology of Hodgkin lymphoma. Hematology 2009: 491–96..