Metformin and tenovin‐6 synergistically induces apoptosis through LKB1‐independent SIRT1 down‐regulation in non‐small cell lung cancer cells

Abstract Sirtuin 1 (SIRT1) is known to play a role in a variety of tumorigenesis processes by deacetylating histone and non‐histone proteins; however, antitumour effects by suppressing SIRT1 activity in non‐small cell lung cancer (NSCLC) remain unclear. This study was designed to scrutinize clinicopathological significance of SIRT1 in NSCLC and investigate effects of metformin on SIRT1 inhibition. This study also evaluated new possibilities of drug combination using a SIRT1 inhibitor, tenovin‐6, in NSCLC cell lines. It was found that SIRT1 was overexpressed in 300 (62%) of 485 formalin‐fixed paraffin‐embedded NSCLC tissues. Its overexpression was significantly associated with reduced overall survival and poor recurrence‐free survival after adjusted for histology and pathologic stage. Thus, suppression of SIRT1 expression may be a reasonable therapeutic strategy for NSCLC. Metformin in combination with tenovin‐6 was found to be more effective in inhibiting cell growth than either agent alone in NSCLC cell lines with different liver kinase B1 (LKB1) status. In addition, metformin and tenovin‐6 synergistically suppressed SIRT1 expression in NSCLC cells regardless of LKB1 status. The marked reduction in SIRT1 expression by combination of metformin and tenovin‐6 increased acetylation of p53 at lysine 382 and enhanced p53 stability in LKB1‐deficient A549 cells. The combination suppressed SIRT1 promoter activity more effectively than either agent alone by up‐regulating hypermethylation in cancer 1 (HIC1) binding at SIRT1 promoter. Also, suppressed SIRT1 expression by the combination synergistically induced caspase‐3‐dependent apoptosis. The study concluded that metformin with tenovin‐6 may enhance antitumour effects through LKB1‐independent SIRT1 down‐regulation in NSCLC cells.


| INTRODUC TI ON
Lung cancer is the most common cause of cancer-related death in the world. Despite significant advances in its diagnosis and treatment, its prognosis remains extremely poor. 1 Currently, a number of agents targeting various molecular pathways are under development or being used in lung cancer treatment. Molecular therapies targeting epithelial growth factor receptor (EGFR), 2 vascular endothelial growth factor (VEGF), 3 and echinoderm microtubule associated protein-like 4 (EML4) and anaplastic lymphoma kinase (ALK) fusion oncogene 4 have been demonstrated to possess significant efficacies against lung cancer. 5 However, failure to achieve longlasting efficacy with a single agent has been observed because cancer cells can acquire resistance during long-term treatment with a single agent such as EGFR-tyrosine kinase inhibitor (TKI) or ALK inhibitor. 6,7 Therefore, this study designed a combination treatment with a new therapeutic target to achieve more effective response than single-agent lung cancer treatment. However, combination therapy using a therapeutic dosage of each individual drug is generally more toxic than single-agent therapy. 8,9 To overcome this problem, this study asked whether a combination treatment at lower concentrations instead of concentrations of each single agent commonly used in vitro could have synergistic effects.
Metformin is an oral antidiabetic drug used to treat type II diabetes. It is also being tested as an anticancer agent because of its ability to suppress cancer growth in vitro and in vivo. [10][11][12][13][14] Metformin is well-known to regulate cell growth through inhibition of mammalian target of rapamycin complex 1 (mTORC1) signalling pathway by activating the AMP-activated protein kinase (AMPK). 15,16 AMPK, a highly conserved intracellular energy sensor and modulator of cell growth, is activated upon decline in adenosine triphosphate (ATP). AMPK is activated by serine/threonine kinase LKB1, a major kinase phosphorylating AMPK under conditions of energy stress. 17 Metformin is known to trigger its activation through LKB1-dependent phosphorylation of AMPK under conditions of low intracellular ATP. 18 LKB1 is inactivated by somatic mutation in approximately 30% of NSCLCs. 19 However, the molecular mechanism involved in the antitumour effect of metformin that is dependent on LKB1 status remains unclear in NSCLC cells. Recent studies have indicated that metformin may sensitize cancer cells to chemotherapy agents in lung cancer. 20,21 For example, metformin and EGFR-TKI have a synergistic effect in treating NSCLC patients with diabetes mellitus type 2. 22 Moreover, metformin can reverse crizotinib resistance by inhibiting type I insulin-like growth factor receptor (IGF-1R) signalling in crizotinib-resistant human lung cancer cells. 23 Metformin and sorafenib can synergistically inhibit tumour growth by activating the AMPK pathway in NSCLC cells both in vitro and in vivo. 24 Thus, combination of metformin with other chemotherapy agents may improve treatment outcome for NSCLC patients. SIRT1, also known as NAD + -dependent deacetylase sirtuin-1, is a homolog of the silent information regulator 2 (Sir2) gene in yeast.
It is involved in diverse cellular processes including metabolism, senescence and tumour initiation and progression, by modulating the deacetylation of histone and non-histone proteins. 25,26 SIRT1 is overexpressed in several human cancers. It is known to play a role in cancer drug resistance by modulating several targets and in the activation of AMPK. [27][28][29] SIRT1 mainly regulates various transcription factors such as tumour suppressor p53, forkhead box protein O1 (FOXO1) and forkhead box class O 3a (FOXO3a) of forkhead transcription factors, peroxisome proliferator-activated receptor-γ coactivator (PGC)-1α, histone acetyltransferase p300 and nuclear factor kappa B (NFkB) in the nucleus. 29,30 Thus, inhibition of SIRT1 expression could have promising therapeutic potential for NSCLC.
This study examines the hypothesis that SIRT1 may be an important target for metformin.
HIC1 is an epigenetically regulated sequence-specific transcriptional repressor in many cancers including prostate, pancreatic and oesophageal cancers. [31][32][33] Inactivation of HIC1 expression is known to up-regulate SIRT1 expression and allow cells to bypass apoptotic cell death. 34,35 HIC1 is also known to play a critical role in DNA damage response. 36 38 The SIRT1 promoter has three HIC1 binding sites at −1116, −1039 and −8 bp regions from the transcription start site (NCBI Refseq: NT_030059.14). 39,40 HIC complexes can differentially bind on two mutually exclusive HIC1 binding sites (distal site and proximal site) on the SIRT1 promoter. 41 Occupancy of distal sites by HIC1 complex was regulated by serum starvation time. Although the mechanism by which HIC affects SIRT1 down-regulation has been explored, little is known about the mechanism involved in the regulation of anticancer activity of metformin in NSCLC cells by SIRT1.
Tenovin-6 is a small-molecule inhibitor of both SIRT1 and SIRT2 that can inhibit cell growth in various cancer types. 42,43 Tenovin-6 is known to enhance cytotoxic effects of 5-fluorouracil and oxaliplatin in colon cancer cells. 44 It has shown very encouraging in vivo effects against cancers in animal experiments. 45,46 Moreover, tenovin-6 is more water-soluble than tenovin-1. 45 Tenovin-6 can inhibit protein deacetylating activities of SIRT1 and SIRT2 and promotes p53 acetylation in cancer cells. 47,48 Although its effect is limited owing to its low specificity. It also induces apoptosis and results in dysregulated autophagy. 49 However, these inhibitors are not considered sufficiently potent to improve patient prognosis. Therapeutic application of SIRT1 inhibitors needs to be investigated in combination with other agents. Therefore, this study determined whether tenovin-6 might be suitable for administration to cancer cells together with metformin because of its potent anticancer effects and water solubility.
The objective of this study was to analyse clinicopathological significance of SIRT1 overexpression using 485 formalin-fixed paraffin-embedded NSCLC tissues. In addition, this study investigated a possible molecular mechanism of the anticancer effect of metformin plus SIRT1 inhibitor, tenovin-6 in NSCLC cells irrespective of LKB1 status.

| Immunohistochemistry
The construction of tissue microarrays (TMAs) from paraffin blocks prepared from the NSCLC samples and immunohistochemical staining of SIRT1 were performed as previously described. 52 A rabbit antihuman SIRT1 polyclonal antibody (Santa Cruz Biotechnology) was used as the primary antibody. All available slides were evaluated in a blinded fashion by two authors (EY Cho and D-H Kim) to reduce interobserver variability. SIRT1 was considered to be overexpressed if immunoreactivity was found in at least 10% of all nuclei. Expression levels of SIRT1 protein were calculated by multiplying the intensity score (0, none; 1, weak; 2, moderate; 3, strong) and the proportion score of positive staining tumour cells (0, absent; 1, 0% to 10%; 2, 10% to 50%; 3, 50% to 80%; 4, >80%). The cut-off value for overexpression was determined by comparison with an internal control consisting of 32 normal lung cores.

| Cell viability assay
Cells were seeded into six-well plates at a density of 2 × 10 5 cells/ mL and then treated with metformin and/or tenovin-6 for 48 hours.

| Soft agar colony formation assay
After A549 cells were treated with metformin and/or tenovin-6 for 48 hours, cells were trypsinized. Cell suspension was mixed with 0.3% soft agar in growth medium and layered (1000 cells/well in 6-well plates) on top of 0.6% base agar with growth medium. After 2 weeks, cells were stained with a nitro blue tetrazolium chloride (Sigma-Aldrich) solution (1 mg/mL in PBS) overnight at 37°C. Colonies containing more than 50 individual cells and those with diameter greater than 0.5 μm were counted using Image J software (National Institutes of Health, Bethesda, MD, USA). All experiments were performed in triplicate.

| Quantitative real-time RT-PCR
Total RNA was extracted from cultured cells using an RNeasy Mini Kit (Qiagen, Valencia, CA, USA) and cDNA was synthesized using

| Immunofluorescence
After cells were grown on glass coverslips and treated with 10 mmol/L

| Chromatin immunoprecipitation assay
The chromatin immunoprecipitation (ChIP) assay was performed with an EZ-ChIP kit (Millipore) and salmon sperm DNA/protein A agarose (Millipore) according to the manufacturer's instructions. A549 cells were cultured in RPMI1640 medium containing 10 mmol/L metformin and/or 10 µmol/L tenovin-6 for 48 hours.
Cells were cross-linked with 1% formaldehyde (Sigma-Aldrich) for 10 minutes at 37°C and lysed in SDS lysis buffer. Lysates were then sonicated to shear cross-linked DNA to fragments of 200 to 1000 base pairs in length. These DNA fragments were immunoprecipitated with an antibody against HIC1 or normal rabbit IgG (Santa Cruz Biotechnology). Purified DNA was then subjected to PCR and qPCR.
Primer sequences used to amplify three HIC1 binding sites in the SIRT1 promoter region are shown in Table 2.

| Luciferase reporter assay
SIRT1 promoter plasmid (pSIRT1-Gluc) containing Gaussia Luciferase (GLuc) as a reporter (vector pEZX-PG02) was purchased from GeneCopoeia (Rockville, MD, USA). This pSIRT1-Gluc plasmid was cotransfected into A549 cells with a wild-type HIC1 expression construct. After transfection, cells were treated with 10 mmol/L metformin and/or 5 µmol/L tenovin-6. Luciferase activity was measured using a Gaussia luciferase assay kit (Promega) according to the manufacturer's instructions. manufacturer's instructions. Apoptotic cells were detected by confocal microscopy. Immunoblot analysis was also performed to detect activated caspase-3 and poly-ADP-ribose polymerase (PARP) cleavage as markers of apoptosis induction. To detect caspase activity, a Caspase-Glo ® 3/7 Assay (Promega) was used according to the manufacturer's instructions.

| Statistical analysis
Associations of SIRT1 overexpression with continuous (or categorical) variables were analysed using the t test (or Wilcoxon rank-sum test) or Pearson's chi-square test (or Fisher's exact test). Multivariate logistic regression analysis was performed to identify independent risk factors affecting SIRT1 overexpression. This study also evaluated the effect of SIRT1 overexpression on patient survival using the Kaplan-Meier method and compared significant differences in survival between the two groups by the log-rank test. Cox proportional hazards regression analysis was performed to estimate hazard ratios of independent prognostic factors for survival, after adjusting for potential confounders. All statistical analyses were two-sided with a type I error rate of 5%.

| SIRT1 overexpression correlates with poor overall and recurrence-free survival in NSCLC patients
This study analysed the association of SIRT1 overexpression with continuous and categorical variables in NSCLC patients.
Clinicopathological characteristics of the 485 participants are described in Table 3. Positive staining for SIRT1 protein is shown in Figure 1A,B. It was overexpressed in 300 (62%) of 485 patients.  showed significantly reduced overall survival (P = 0.0005; Figure 1C) and poor recurrence-free survival (RFS; P = 0.006; Figure 1D).

| Metformin and tenovin-6 synergistically inhibit cell growth in NSCLC cells
This study showed that SIRT1 overexpression was associated with poor overall and recurrence-free survival in NSCLC. Thus, whether SIRT1 inhibitor tenovin-6 could enhance the anticancer effect of metformin by inhibiting SIRT overexpression in NSCLC cells was determined. First, this study compared effects of metformin-induced growth inhibition as a single agent and in combination with tenovin-6 in NSCLC cells. Concentrations of metformin and tenovin-6 used in this study were based on the MTS assay. IC 50 values for metformin and tenovin-6 in functionally LKB1-negative A549 cells were 28.7 mmol/L and 21.1 μmol/L respectively (data not shown).
However, this study used lower concentrations of metformin and tenovin-6 because high doses of metformin in vitro were controversial in clinical application. [57][58][59] Metformin ( Figure 1E) and tenovin-6 ( Figure 1F 10 μmol/L tenovin-6. Therefore, the combination of metformin and tenovin-6 showed synergism in suppressing cell growth. Consistent with this result, colony formation assay using A549 cells showed that the number of cell colonies was significantly decreased in metformin or tenovin-6 alone group than that in the control ( Figure 1H,I). In addition, combined treatment of metformin and tenovin-6 reduced colonies by 8% of initial plating density compared with control in A549 cells. This study also observed significantly decreased growth of wild-type LKB1 H1299 and H1650 as well as functionally LKB1-negative H460 under the same experimental conditions ( Figure 1J-L). These results confirmed that tenovin-6 sensitized the effect of metformin on controlling NSCLC cell growth irrespective of LKB1.

| Metformin and tenovin-6 synergistically down-regulate SIRT1 expression in NSCLC cells irrespective of LKB1 status
This study explored whether the antiproliferative effect of the combination of metformin with tenovin-6 was mediated by SIRT1 expression. Whether metformin regulated SIRT1 expression by metformin in functionally LKB1-deficient A549 cells was first investigated. SIRT1 mRNA (Figure 2A There was more SIRT1 reduction in the combined treatment trans-   Figure 3E).
Additionally, changes of SIRT1 expression by metformin and/or tenovin-6 after transfection with LKB1 wt in LKB1-deficient A549 cells were investigated. As expected, SIRT1 expression levels were remarkably suppressed in A549 cells with or without LKB1 wt by metformin and/or tenovin-6 treatment as compared with controls ( Figure 3F)

| Metformin and tenovin-6 synergistically induce p53 acetylation
SIRT1 inhibition is known to reduce cell survival through p53 acetylation. 61 Therefore, this study analysed whether SIRT1 inhibition by metformin and tenovin-6 could regulate p53 acetylation and downstream target genes. Effects of combination treatment on SIRT1 activity were assessed by examining p53 acetylation at lysine 382, a known SIRT1 deacetylation site. In A549 cells, metformin suppressed SIRT1 expression and induced p53 acetylation at lysine 382.
It also increased protein levels of p53, p21 and GADD45α in doseand time-dependent manners ( Figure 4A,B). In addition, immunofluorescent staining intensity of p53 acetylation at lysine 382 was significantly greater in metformin-treated cells than that in untreated cells ( Figure 4C). Combined treatment with metformin and tenovin-6 also increased p21 and GADD45α expression and p53 acetylation at lysine 382 in A549 cells in a dose-dependent manner ( Figure 4D).
Next, this study examined whether the treatment of metformin and/ or tenovin-6 directly act downstream of SIRT1 by testing the effect of SIRT1 overexpression. In addition, p53 acetylation was suppressed by ectopic expression of SIRT1 (lane 2 vs lane 1; Figure 4E).
As shown in Figure 4D To determine whether p53 stability was affected by accumulation of p53 acetylation at lysine 382 and increase in GADD45α protein level, the half-life of p53 was measured following metformin and/or tenovin-6 treatment ( Figure 4F,G). A549 cells were treated with 10 mmol/L metformin or 10 μmol/L tenovin-6 for 24 hours and then treated with CHX (25 μg/mL for 0, 0.5, 1, 2, 4 and 6 hours).
Metformin and tenovin-6 alone or in combination substantially increased the half-life of p53 in A549 cells. Thus, elevated p53 expression ( Figure 4A,B,E) may be a result of its increased half-life. Taken together, these results indicate that combination of metformin with tenovin-6 can synergistically enhance p53 acetylation and regulate its downstream targets by inhibiting SIRT1 in LKB1-deficient A549 cells.

| Metformin and tenovin-6 suppress SIRT1 expression by accumulating HIC1 binding to the SIRT1 promoter
To understand the possible mechanism underlying SIRT1 down-regulation by metformin and tenovin-6, this study analysed the binding of HIC1 to the SIRT1 promoter using chromatin immunoprecipitation. Changes in H1C1 mRNA levels induced by metformin or tenovin-6 were minimal. However, they were significantly affected by combination treatment with metformin and tenovin-6 ( Figure 5A).
There are three HIC1-binding sites in the human SIRT1 promoter To analyse effects of metformin and tenovin-6 on HIC1 binding to SIRT1 promoter, this study transiently transfected A549 cells with a SIRT1-luciferase vector (pSIRT1-Gluc) and then treated them with metformin and/or tenovin-6 ( Figure 5H) or cotransfected them with wild-type HIC1 ( Figure 5I). The combination treatment suppressed SIRT1 transcriptional activity in A549 cells with endogenous HIC1 ( Figure 5H). To analyse the effect of HIC1 on SIRT1 transcriptional activity after ectopic expression of wild-type HIC1, this study transiently cotransfected A549 cells with pSIRT1-Gluc and HIC1 wt followed by treatment with metformin and/or tenovin-6. SIRT1 luciferase activity in A549 cells with exogenous wild-type HIC1 was significantly suppressed in response to the combination treatment ( Figure 5I). Overall, these results indicated that metformin and tenovin-6 suppressed . Transfected cells were cultured in 10 mmol/L metformin and 10 µmol/L tenovin-6 alone or in combination for 48 h. Luciferase activity was then measured. (I) A549 cells were cotransfected with pSIRT1-Gluc (or pEZX-PG02-Gluc) and plasmids expressing wild-type HIC1 (HIC1 wt ). Transfected cells were treated with 10 mmol/L metformin and 10 µmol/L tenovin-6 alone or in combination for 48 h and luciferase activity was measured. "M," "T" and "MT" indicate metformin (10 mmol/L), tenovin-6 (10 μmol/L), and a combination of metformin and tenovin-6 respectively. Experiments shown in H and I were independently performed three times. Data are displayed as mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001

| Metformin and tenovin-6 synergistically promote the apoptotic pathway through SIRT1 downregulation in A549 cells
SIRT1 is known to repress p53-dependent transcription, thereby inhibiting p53-mediated apoptosis following DNA damage or oxidative stress. 62 This study evaluated the effects of a combination of 10 mmol/L metformin and 10 µmol/L tenovin-6 on apoptosis of A549 cells. The combination treatment increased mRNA levels of pro-apoptotic genes such as APAF1, BAK1, BAX, DDIT3, DR5, GADD45α, NOXA, PUMA and TNFRSF10A more effectively than either monotherapy alone in A549 cells ( Figure 6A). To determine whether metformin and tenovin-6 caused cell death by apoptosis, A549 cells were analysed by flow cytometry (FACs) following Annexin V-FITC and propidium iodide (PI) dual labelling ( Figure 6B,C).

The percentage of cells that underwent apoptosis as measured by
FACs was approximately two times higher in A549 cells treated with metformin than that in control cells (12.68% vs 5.12%, respectively; Figure 6B,C). Apoptosis was weaker for cells treated with tenovin-6 (5.80%) alone than that for cells treated with metformin (12.68%).
F I G U R E 6 Synergistic effect of metformin and tenovin-6 on apoptosis. (A) A549 cells were treated with 10 mmol/L metformin and 10 μmol/L tenovin-6 for 48 h and then mRNA expression levels of pro-apoptotic genes were measured by qRT-PCR. Fold change indicates mRNA levels relative to untreated control cells. Error bars indicate mean ± SD from triplicate experiments. (B-E) For apoptosis assay, A549 cells were treated with 10 mmol/L metformin and 10 μmol/L tenovin-6 for 48 h. Apoptosis was determined by annexin V-FITC/PI staining and measured by FACs (B). Apoptotic cells were gated as a percentage of annexin V-only-positive cells (C). Experiments shown in A and C were independently performed three times. Data are displayed as mean ± SD; *P < 0.05, **P < 0.01, ***P < 0.001. In addition, apoptotic cells were stained with annexin V conjugated to green fluorescent FITC dye (D) and analysed using a TUNEL assay (E). (F) Immunoblot and analysis of caspase-3/7 activity. A549 cells were treated with metformin (10 mmol/L) alone or in combination with tenovin-6 (10 μmol/L) for 48 h. Cell lysates were immunoblotted with antibodies against PARP and Caspase 3 to detect apoptosis. α-tubulin was used as a loading control. (G) Caspase-3/7 activity was measured using Caspase-Glo 3/7 assay kit. Results are displayed as mean ± SD; **P < 0.01, ***P < 0.001. (H) A549 cells were transfected with Flag-EGFP or Flar-SIRT1 and treated with or without metformin and tenovin-6 for 48 h. Cell lysates were then immunoblotted with caspase-3 and PARP antibodies for activated endogenous caspase-3/7 activity. Experiments were independently performed three times However, the combined treatment significantly increased apoptosis (22.96%). In addition, Annexin V staining ( Figure 6D) and TUNEL assays ( Figure 6E) showed the induction of apoptosis of A549 cells by the combination treatment.
To further confirm apoptosis induction by metformin and tenovin-6, this study measured cleaved forms of caspase-3 and PARP ( Figure 6F-6G). Caspase-3 and PARP were cleaved in the presence of metformin or tenovin-6 ( Figure 6F). However, the combination of metformin and tenovin-6 induced caspase-3 activation and PARP cleavage in A549 cells more effectively than either metformin or tenovin-6 alone. Furthermore, endogenous caspase-3/7 activity was 3.3 times higher in A549 cells treated with metformin and tenovin-6 than that in untreated A549 cells ( Figure 6G). Overexpression of Flag-SIRT1 restored the increase in caspase-3 activation and PARP cleavage (lane 4; Figure 6H). Adding metformin and tenovin-6 resulted in caspase-3 activation and PARP cleavage (lane 5; Figure 6H). These results suggest that the combined treatment of metformin and tenovin-6 can synergistically induce the apoptotic pathway through SIRT1 down-regulation in A549 cells.

| D ISCUSS I ON
The relationship between SIRT1 overexpression and overall survival of patients with NSCLC has been analysed in several studies.
A recent meta-analysis showed that SIRT1 overexpression was associated with reduced overall survival and that the unfavourable prognostic impact was independent of TNM stage, consistent with our finding. 63 80 Our study focuses on assessing synergistic effects between metformin and tenovin-6 in NSCLC cells irrespective of LKB1 status.
However, this study showed that tenovin-6, as well as metformin, suppressed SIRT1 transcriptional activity in A549 cells. Tenovin-6 was not only involved in the decrease of SIRT1 activity, but also involved in the decrease of SIRT1 expression. The mechanism of tenovin-6 needs to be determined in further study.
It has been reported that metformin can significantly inhibit tumour cells in vitro at higher concentrations. 81 (data not shown). However, we used low concentrations of metformin and tenovin-6. Our results reveal synergistic effects to regulate SIRT1 expression by the combination of metformin with tenovin-6.
A serious problem in treating lung cancer is that some patients continue to smoke even after their diagnoses. Continuous exposure to tobacco smoke may influence the efficacy of chemotherapeutic agents. 87 Therefore, in vitro studies with or without exposure to NNK (Nicotine-derived nitrosamine ketone) were performed. A549 cells were incubated with NNK for 2 days and then with metformin for 2 days. Metformin decreased both expression and phosphorylation of SIRT1 but increased p53 acetylation in a time-dependent manner (data not shown), suggesting that metformin might be effective even in smokers.
In summary, this study reveals that SIRT1 overexpression is associated with poor survival in NSCLC patients. This study also provides a mechanism for antitumour effects of targeting SIRT1 in NSCLCs. Results of this study showed that combination of metformin and tenovin-6 acted synergistically in inhibiting cell growth in

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N S
BBL and DHK designed the overall study and drafted the manu-