|Home | About | Journals | Submit | Contact Us | Français|
Numerous efficacious chemotherapy regimens may cause thrombocytopenia. Thrombopoietin receptor (TPO-R) agonists, such as eltrombopag, represent a novel approach for the treatment of chemotherapy-induced thrombocytopenia. The TPO-R MPL is expressed on megakaryocytes and megakaryocyte precursors, although little is known about its expression on other tissues.
Breast, lung, and ovarian tumor samples were analyzed for MPL expression by microarray and/or quantitative reverse transcription-polymerase chain reaction (qRT-PCR), and for TPO-R protein expression by immunohistochemistry (IHC). Cell line proliferation assays were used to analyze the in vitro effect of eltrombopag on breast, lung, and ovarian tumor cell proliferation. The lung carcinoma cell lines were also analyzed for TPO-R protein expression by Western blot.
MPL mRNA was not detectable in 118 breast tumors and was detectable at only very low levels in 48% of 29 lung tumors studied by microarray analysis. By qRT-PCR, low but detectable levels of MPL mRNA were detectable in some normal (14-43%) and malignant (3-17%) breast, lung, and ovarian tissues. A comparison of MPL to EPOR, ERBB2, and IGF1R mRNA demonstrates that MPL mRNA levels were far lower than those of EPOR and ERBB2 mRNA in the same tissues. IHC analysis showed negligible TPO-R protein expression in tumor tissues, confirming mRNA analysis. Culture of breast, lung, and ovarian carcinoma cell lines showed no increase, and in fact, showed a decrease in proliferation following incubation with eltrombopag. Western blot analyses revealed no detectable TPO-R protein expression in the lung carcinoma cell lines.
Multiple analyses of breast, lung, and ovarian tumor samples and/or cell lines show no evidence of MPL mRNA or TPO-R protein expression. Eltrombopag does not stimulate growth of breast, lung, or ovarian tumor cell lines at doses likely to exert their actions on megakaryocytes and megakaryocyte precursors.
Breast cancer is the most commonly occurring neoplasm and the second leading cause of cancer deaths in women. Lung cancer is the second most frequent cancer diagnosis in men and women, and remains the leading cause of cancer deaths. Although ovarian cancer affects fewer women than breast or lung cancer, it is one of the most lethal types of cancer . Current clinical guidelines recommend platinum-containing chemotherapy regimens, among others, for these malignancies in different disease stages [2-7]. Established treatment protocols may be associated with a range of adverse events (AEs), including thrombocytopenia, anemia, and neutropenia. Newer chemotherapy combinations including carboplatin plus pemetrexed or gemcitabine in non-small cell lung cancer (NSCLC) may also be associated with high rates of thrombocytopenia .
Thrombocytopenia may lead to significant clinical consequences including petechiae, gastrointestinal bleeding, and bleeding into the brain . Neutropenia, anemia, or thrombocytopenia resulting from bone marrow suppression can delay chemotherapy administration and/or may prompt dose reductions, with possible negative impact on disease control . The use of hematopoietic growth factors has ameliorated this problem to some degree with respect to red and white blood cell production [11,12]. However, there are concerns that some growth factors could induce proliferation of other cell types, including tumor cells [13,14].
Thrombopoietin (TPO) is a critical cytokine regulating thrombopoiesis. It is the endogenous ligand for the thrombopoietin receptor (TPO-R) MPL expressed on the surface of megakaryocytes, megakaryocyte precursors, and platelets [15-18]. TPO-R agonists are approved for the treatment of chronic immune thrombocytopenia (ITP). The interaction of the selective, nonpeptidyl TPO-R agonist, eltrombopag, with TPO-R triggers activation of the JAK-STAT and MAP kinase, but not the AKT, signal transduction pathways. This causes alterations in gene expression patterns to promote megakaryocytic differentiation and maturation, resulting in increased platelet counts [19,20]. There are differences between the amplitude and extent of signaling between TPO and eltrombopag . Acute myeloid leukemia (AML) blasts have been shown to express TPO-R  and in some reports TPO has induced proliferation of these blasts , although investigations in leukemia cell lines show conflicting results [24,25]. A variety of human non-megakaryocytic leukemia and lymphoma cell lines show decreased, rather than increased, proliferation upon incubation with eltrombopag . In vitro and in vivo analyses of bone marrow mononuclear cells from patients with AML and myelodysplastic syndromes (MDS) similarly showed no increase in proliferation with eltrombopag treatment [25,27].
The objective of this study was to determine whether solid tumors (i.e., breast, lung, and ovarian) express MPL mRNA or TPO-R protein, and whether eltrombopag affects the proliferation of solid tumor cell lines. Expression of MPL mRNA was assessed by microarray analysis in breast and lung tumor samples and by quantitative reverse transcription-polymerase chain reaction (qRT-PCR) in normal and malignant tissue samples from breast, lung, and ovary. TPO-R protein expression was determined by immunohistochemistry (IHC) on breast, lung, and ovarian tumor samples and by western blot on lung cell lines. Cell proliferation in response to eltrombopag was evaluated in breast, lung, and ovarian cancer cell lines.
In accordance with the Helsinki Declaration, all patients provided written informed consent for use of their samples, and the collection and use of the samples received Institutional Review Board (IRB) approval.
Archival tissue samples from 118 patients who had locally advanced or metastatic breast cancer and had failed treatment with anthracycline-, taxane-, and trastuzumab-containing regimens were studied. Specimens were obtained from patients who had histologically confirmed invasive breast cancer with Stage IIIB, Stage IIIC with T4 lesion, or Stage IV disease (GSK Study EGF100151), with documentation of ERBB2 overexpression (IHC 3+ or IHC 2+ with fluorescence in situ hybridization [FISH] confirmation) [28,29]. Microarray analysis was performed at Response Genetics, Inc (Los Angeles, CA, USA).
Frozen tissue samples from 29 patients who were presurgical, treatment-naive, with Stage IA or Stage IB, resectable NSCLC were studied (GSK Study VEG105290). Microarray analysis was performed at Weill Medical College of Cornell University (New York, NY, USA).
Archival breast tumor specimens were formalin-fixed and embedded in paraffin. The quality of tumor tissue RNA extraction was assessed by laser capture microdissection. Reverse transcription-polymerase chain reaction (RT-PCR) for 300 base pair (bp) of the β-actin gene was used for quality control of these samples. Samples with a cycle time (CT) of<32 at the 300bp level were acceptable for microarray analysis. mRNA and cDNA were prepared from archival breast tumor tissues and frozen NSCLC samples using the Affymetrix HG-U133 Plus2 array (Affymetrix, Inc., Santa Clara, CA, USA). RNA microarray gene signal intensity was normalized using the robust microchip analysis (RMA) method described by Irizarry et al . Samples with RMA<50 are below the level of detection by this method.
Samples of normal tissue cDNA and tumor tissue cDNA were obtained from Cytomyx (Lexington, MA, USA). Seven samples of normal breast tissue were studied, and 41 breast carcinoma samples, including tissue from cancers of Stage I, Stage II, and Stage IIIB were analyzed. Eight normal lung specimens and lung tissue samples from 40 tumors, ranging from Stages IA to IV were examined. Ovarian samples included 7 normal ovarian tissues and 41 ovarian tumors ranging from Stages I to IV.
MPL gene expression was measured using FAM-TAMRA-labeled primers and probes in a 7900HT thermal cycler using a standard 40-cycle profile with 9600 emulation. Following amplification, calculations were performed to determine the relative abundance of MPL normalized to 3 housekeeping genes, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β-actin (ACTB), and cyclophilin A (PPIA). Each result is presented as a normalized value of mRNA. EPOR, ERBB2, and IGF1R were also assessed to provide data for comparable receptors. Primers and probes were custom made by Integrated DNA Technologies (Coralville, IA, USA). Sequences for the primers and probes were previously described  and are listed in Additional file 1: Table S1.
CT values for genes of interest were normalized to the internal housekeeping genes run in the reaction using in-house software. Raw abundance value was calculated using the following equation: Abundance=10e[(40-CT)/3.35]. Samples were scaled relative to each other using the geometric mean of the set of valid housekeeping gene data points for that sample. Each data point was then expressed as the ratio of the housekeeping gene abundance in the sample to the average of that housekeeping gene in all samples and marked invalid if it had statistically inconsistent behavior with the other housekeeping genes in those samples of similar tissue types. Samples with a relative abundance of<50 are below the level of detection of this method.
Formalin-fixed, paraffin embedded (FFPE) controls and FFPE specimens of breast, lung, and ovarian cancer were procured by Mosaic Laboratories under an IRB-reviewed protocol that allows for use of remnant, de-identified, or anonymized human samples for in vitro analysis under the guidelines defining Exemption from Human Subject Research as defined by the Office of Human Research Protection. FFPE tissue samples of breast, lung, and ovarian cancer were also procured from OriGene (formerly Cytomyx) for analysis. A FFPE block of N2C-Tpo cells, a Tpo-dependent megakaryocytic leukemia cell line, and of normal bone marrow were used as positive controls for TPO-R expression.
Immunohistochemical analysis was performed in accordance with Mosaic Laboratories’ Standard Operating Procedures. Briefly, the procedure was performed manually using the Envision™ system and ancillary reagents (Dako, Carpinteria, CA). Specimens were sectioned at a thickness of 4 microns, mounted onto positive-charged glass slides, dried, deparaffinized, and rehydrated. Following rehydration, tissue sections were incubated in Envision peroxidase for 5minutes to quench endogenous peroxidase. Heat-induced epitope retrieval was performed using Rip Tide buffer for 40minutes at 95°C. Slides were incubated with anti-CD110 antibody (clone 1.78.1) (BD Biosciences) diluted in diluent (Dako) for 30minutes. Slides were then rinsed twice in Splash-T buffer for 5minutes each. Signal was visualized using Envision+Mouse HRP detection reagent (Dako) for 30minutes, followed by 3,3 diamino benzidine according to manufacturer’s instructions. Slides were rinsed with water, counterstained with hematoxylin (Dako), dehydrated through graded alcohols, cleared in xylene, and coverslipped for microscopic evaluation.
Staining was evaluated by a pathologist; evaluation of reactivity involved a combination of the following: cellular localization of staining, staining intensity, subcellular localization, and percentage of cells staining in the primary component of the tissue type of interest. The CD110 IHC assay was scored on a semi-quantitative scale, and the percentage of cancer cells staining at each of the following 4 levels was recorded: 0 (unstained), 1+ (weak staining), 2+ (moderate staining), and 3+ (strong staining). An H-score was calculated based on the summation of the product of percent of cells stained at each intensity, using the following equation: (3×% cells staining at 3+)+(2×% cells staining at 2+)+(1×% cells staining at 1+). H-score values range from 0–300. (Additional file 2: Table S2).
Cell lines were obtained from the American Type Culture Collection (Walkersville, MD, USA). All cells were maintained in log-phase growth in their respective media. The breast cancer cell lines (MCF-7, BT474, and HCC1937), lung cancer cell lines (A549, NCI-H226, NCI-H460, and NCI-H510), and ovarian cell lines (OVCAR4 and SKOV-3) were grown in RPMI 1640 medium with 10% fetal calf serum (FCS). The ovarian carcinoma cell line OVCAR3 was grown in RPMI 1640 medium with 20% FCS, 1% sodium pyruvate, and 1%v/v glutamine.
All experiments were performed with milled, monoethanolamine salt form SB-497115-GR (eltrombopag) resuspended in water to 10mg/mL, and diluted in Iscove’s Modified Dulbecco’s Medium (IMDM) with 1% FCS. Recombinant human TPO (rhTPO) was purchased from R&D Systems (Minneapolis, MN, USA) and diluted in IMDM.
Cells for the Cell Titer Glo® assay were plated at 1×103 cells/well in 96-well plates in medium containing 10% FCS and allowed to adhere for 24hours. Cells were treated with eltrombopag at 0, 0.1, 0.4, 1, 4, 10, and 40μg/mL. In breast and ovarian cancer cell lines, eltrombopag was also tested at 100μg/mL. rhTPO at 100ng/mL was also tested in these experiments. Cells were incubated for 72hours at 37°C in 5% CO2 after the addition of eltrombopag or rhTPO. Active cell determinations were performed using the Cell Titer Glo® reagent (Promega, Madison, WI, USA) according to the manufacturer’s protocol. Results were reported as relative luminescence units (RLU).
The calculated mean and standard deviations were produced using triplicate samples for each experiment. The IC50 was determined using XLfit version 4.2.1, utilizing the best fit model for each data set.
Lung cancer cell lines, A549, NCI-H226, NCI-H460, and NCI-H510, were grown as described above.
Western blots for TPO-R protein expression were performed on reduced cell lysates of log-phase growth lung cancer cell lines (50μg/well) on a NuPage 4-12% Bis-Tris gel (Invitrogen, Carlsbad, CA, USA) with MOPS running buffer. Precision Protein Dual Color Standards (Bio-Rad, Hercules, CA, USA) were used. The gels were transferred to nitrocellulose and stained with a rabbit polyclonal anti-TPO-R primary antibody (Upstate Biotechnology Inc., Lake Placid, NY, USA; Cat# 06–944) and analyzed with an Odyssey® infrared imager (LI-COR Biosciences, Lincoln, NE, USA).
None of the 118 advanced or metastatic breast cancer samples expressed detectable levels of MPL mRNA; all samples had RMA<50. EPOR mRNA was expressed at detectable levels in 89/118 (75%); all 118 (100%) demonstrated detectable levels of ERBB2 mRNA; and 102/118 (86%) expressed detectable levels of IGF1R mRNA (Table (Table1,1, Figure Figure1).1). ERBB2 mRNA was expressed at the highest level compared with MPL, EPOR, and IGF1R mRNA levels.
Of the 29 samples of NSCLC studied by microarray, 14 (48%) expressed low but detectable levels of MPL mRNA (Table (Table1,1, Figure Figure1).1). All 29 samples (100%) expressed EPOR mRNA; 28 of 29 (97%) expressed ERBB2 mRNA, and all 29 (100%) expressed IGF1R mRNA.
A more accurate quantitation of MPL expression was undertaken using qRT-PCR. MPL expression was lower in both normal and malignant breast tissue than expression of EPOR, ERBB2, and IGF1R (Table (Table2,2, Figures Figures2A2A and and3A).3A). One of 7 (14%) normal breast tissue samples studied and 1/40 (3%) breast tumor samples demonstrated detectable MPL mRNA expression (Table (Table2,2, Figures Figures2A2A and and3A).3A). EPOR mRNA expression was observed in 7/7 (100%) normal breast samples, and in 32/40 (80%) breast cancer samples. Four of 7 (57%) normal breast tissue samples and 33/40 (83%) breast tumor samples demonstrated detectable ERBB2 expression. IGF1R expression was detected in 2/7 (29%) normal breast tissue samples and in 15/40 (38%) breast cancer samples.
Detectable MPL expression was observed using qRT-PCR in 1/8 (13%) normal lung tissue samples, and in 1/40 (3%) lung tumor samples. The level of MPL expression was equivalent in the normal lung tissue and in the positive tumor sample (Table (Table2,2, Figures Figures2B2B and and3B).3B). The lung tumor sample that expressed detectable MPL mRNA also expressed EPOR mRNA at levels greater than did other lung tumor samples. EPOR mRNA expression was identified in all 8 (100%) normal lung samples, and in 28/40 (70%) lung tumor samples studied. ERBB2 mRNA expression was detected in all 8 (100%) normal samples and in 22/40 (55%) lung tumor samples. In contrast, IGF1R mRNA expression was identified at detectable levels in only 1/8 (13%) normal lung samples and 2/40 (5%) lung tumor samples studied (Table (Table2,2, Figures Figures2B2B and and33B).
Three of 7 (43%) normal ovarian tissue samples expressed detectable MPL mRNA and 3/41 (7%) ovarian tumor samples had detectable MPL mRNA. Relative expression of MPL was similar or greater in the positive normal ovary and ovarian tumor samples (Table (Table2,2, Figures Figures2C2C and and3C).3C). All 7 (100%) normal ovary samples and all 41 (100%) ovarian tumors studied expressed EPOR mRNA. ERBB2 expression was also observed in all 7 (100%) normal samples and in 39/41 (95%) tumor samples. IGF1R expression was not detected in the 7 normal ovarian samples, and was detectable in 7/41 (17%) ovarian tumors studied. Expression of MPL and IGF1R mRNA was low compared with EPOR and ERBB2 mRNA levels in normal ovary and ovarian tumor samples (Table (Table2,2, Figures Figures2C2C and and3C).3C). There was too little MPL mRNA expression in the tumor samples to determine any relationship between MPL expression levels and stage of cancer in these breast, lung, or ovarian samples (Figure (Figure33).
Immunohistochemical analysis examining multiple fields of FFPE samples of breast, lung, and ovarian patient tissue stained with the anti-TPO-R antibody (CD110) demonstrated little or no specific signal. One breast tumor section showed a 1+ signal in one of infiltrating inflammatory cell, but no staining of tumor cells was noted. (Figure (Figure44 and Additional file 2: Table S2). As expected, the positive controls of normal bone marrow megakaryocytes and the N2C-Tpo cell line (Figure (Figure44 bottom panels) demonstrated robust specific staining with the antibody.
Given that the expression of mRNA does not always reflect the expression of protein, several breast, lung, and ovarian cancer cell lines were chosen to assess the effects of eltrombopag on proliferation using the Cell Titer Glo assay. Three different breast cancer cell lines MCF-7, BT474, and HCC1937 were incubated with eltrombopag for 72hours to determine whether this compound stimulated proliferation of breast cancer cells. The MCF-7 cell line data are shown as a representative experiment in Figure Figure5A.5A. No increase in cell number was observed in the 3 breast cancer cell lines with eltrombopag (0.1-100μg/mL); in fact, all of them showed a decrease in cell number at eltrombopag concentrations>4μg/mL. Recombinant TPO did not increase or decrease proliferation of any of these breast cancer cell lines (data not shown). Table Table33 shows the IC50 of eltrombopag on these cell lines: 19.0μg/mL for MCF-7; 9.6μg/mL for BT474; and 10.7μg/mL for HCC1937.
Four lung carcinoma cell lines were studied to determine whether eltrombopag stimulated proliferation including 3 NSCLC (A549, from an alveolar cell carcinoma; NCI-H226, from a squamous cell carcinoma; and NCI-H460, from a large cell carcinoma) and a small cell lung carcinoma cell line, NCI-H510. There was no increase in cell number over 72hours of treatment with eltrombopag at concentrations from 0.1μg/mL to 40μg/mL in any of the 4 lung carcinoma cell lines. The response of NCI-H226 to eltrombopag is shown in Figure Figure5B.5B. All cell lines showed a decrease in cell number at higher concentrations of eltrombopag (>4μg/mL). Recombinant TPO did not affect the proliferation, either positively or negatively, of any of these cell lines (data not shown). As shown in Table Table3,3, the IC50 of eltrombopag on these cell lines was 9.0μg/mL for A549; 3.7μg/mL for NCI-H226; 8.1μg/mL for NCI-H460; and 10.3μg/mL for NCI-H510.
None of the ovarian carcinoma cell lines (OVCAR3, OVCAR4, and SKOV-3) demonstrated increased proliferation in response to 0.1μg/mL to 100μg/mL eltrombopag treatment over 72hours. For all 3 cell lines, cell number decreased at eltrombopag concentrations>4μg/mL. A representative experiment demonstrating the response of OVCAR3 cells is shown in Figure Figure5C.5C. Recombinant TPO had no positive or negative proliferative effect on these cell lines (data not shown). As shown in Table Table3,3, the IC50 of eltrombopag on these cell lines was 4.8μg/mL for OVCAR3, 11.0μg/mL for OVCAR4, and 49.7μg/mL for SKOV-3.
We had previously noted that NCI-H510, the cell line derived from small cell lung carcinoma, expressed high levels of MPL mRNA by qRT-PCR . We sought to determine whether mRNA expression correlated with TPO-R protein expression using Western blots of cell lysates of lung cancer cell lines treated with eltrombopag. In NCI-H510 cell lysates, no band corresponding to TPO-R protein was detected. Human platelets and the megakaryocytic cell lines N2C-Tpo and HEL92.1.7 had high levels of MPL mRNA and TPO-R protein (Figure (Figure6).6). NCI-H226 and NCI-H460 cell lines did not express detectable MPL mRNA by qRT-PCR, nor did they demonstrate TPO-R protein on Western blots. Western blots reveal a band corresponding to a protein smaller than TPO-R in NCI-510 and NCI-H226 cell lysates that is also present in N2C-Tpo and HEL92.1.7 lysates, but the identity of this band is unknown.
Thrombocytopenia is commonly observed during treatment with a number of chemotherapy regimens approved or in development for the treatment of breast, lung, and ovarian cancers [8,32-36]. Identification and characterization of novel, safe, and effective thrombopoietic agents to ameliorate thrombocytopenia remains an intense area of research . Eltrombopag interacts with TPO-R, stimulates megakaryopoiesis, and promotes platelet production and maturation. Although MPL is expressed on megakaryocytes and megakaryocyte precursors [15-18], little quantitative data are available on MPL expression in other tissues.
In this study, MPL expression was compared with expression of other genes (EPOR, ERBB2, and IGF1R), which have been widely reported [37-42]. Our analyses showed that MPL was not commonly expressed at detectable levels in tumor samples as measured by qRT-PCR. Using microarray analyses, MPL expression was lower than other cell surface receptors. Even when protein is expressed, it may not be folded properly to allow ligand interactions. For example, although MPL is expressed on leukemic blasts from patients with AML , no proliferation in response to eltrombopag in bone marrow mononuclear cells from patients with AML or MDS has been observed . Nor did eltrombopag stimulate proliferation in a variety of non-megakaryocytic leukemia and lymphoma cell lines; decreased proliferation was observed at physiologically achievable eltrombopag concentrations (≥ 4μg/mL) . The observed median Cmax for eltrombopag in patients with ITP is 11.4μg/mL at the 75-mg dose , while in patients with chemotherapy-induced thrombocytopenia, the observed median Cmax at the 75-mg dose (based on sparse data) is approximately 9.90μg/mL (unpublished data). Therefore, the dose at which tumor cell line growth declines is in a physiologically achievable range. Further studies are needed to determine whether eltrombopag affects tumor growth in vivo.
Eltrombopag did not stimulate growth of breast, lung, or ovarian cancer cell lines at doses likely to exert action on megakaryocytes and megakaryocyte precursors. Tumor samples of these types had very little, or no, MPL mRNA or TPO-R protein expression as determined by qRT-PCR or IHC, respectively.
ACTB: β-actin; AEs: Adverse events; AML: Acute myeloid leukemia; Bp: Base pair; CT: Cycle time; FCS: Fetal calf serum; FFPE: Formalin-fixed parafin embedded; FISH: Fluorescence in situ hybridization; GAPDH: Glyceraldehyde-3-phosphate dehydrogenase; IHC: Immunohistochemistry; IMDM: Iscove’s Modified Dulbecco’s Medium; IRB: Institutional Review Board; ITP: Chronic immune thrombocytopenia; MDS: Myelodysplastic syndromes; NSCLC: Non-small cell lung cancer; PPIA: Cyclophilin A; qRT-PCR: Quantitative reverse transcription-polymerase chain reaction; rhTPO: Recombinant human TPO; RLU: Relative luminescence units; RMA: Robust microchip analysis; RT-PCR: Reverse transcription-polymerase chain reaction; TPO: Thrombopoietin; TPO-R: Thrombopoietin receptor.
CLEM is employed by GlaxoSmithKline (GSK), holds stock, and receives research funding. KP is a former employee of GSK, held stock, and received research funding. JK is employed by GSK. DJF, LO, AMM, YL, YMK, and CM are employed by GSK and hold stock.
CLEM contributed to the conception and design of the experiments, analyzed and interpreted data, drafted and critically reviewed the manuscript. KP carried out gene profiling and cell proliferation assays. JK performed qRT-PCR and cell proliferation assays. DJF participated in IHC data acquisition, and analysis and interpretation. LO participated in data acquisition and manuscript revisions. AMM conducted research that contributed to results and helped in writing and editing the manuscript. YL participated in data acquisition and data analysis and interpretation. YMK participated in the interpretation of data and critical review of the manuscript. CM contributed to the conception and design of the experiments, interpreted the data, and drafted and critically reviewed the manuscript. All authors read and approved the final manuscript.
The pre-publication history for this paper can be accessed here:
Table S1. Sequences of primers and probes utilized in qRT-PCR analyses. qRT-PCR, quantitative reverse transcription-polymerase chain reaction.
Table S2. Pathologists review of immunohistochemistry slides.
Funding for this study was provided by GlaxoSmithKline (GSK). All listed authors meet the criteria for authorship set forth by the International Committee for Medical Journal Editors. The authors wish to acknowledge the following individuals for their contributions and critical review during the development of this manuscript on behalf of GSK: Rachel Brody and AOI Communications, L.P., for medical writing and editorial assistance, and Kimberly Marino and Rosanna Tedesco of GSK for critical review.