CADM1 Expression and Stepwise Downregulation of CD7 Are Closely Associated with Clonal Expansion of HTLV-I–Infected Cells in Adult T-cell Leukemia/Lymphoma Free

Human T-cell lymphotrophic virus type I (HTLV-I) is a human retrovirus that causes HTLV-I–associated diseases, such as adult T-cell leukemia–lymphoma (ATL), HTLV-I–associated myelopathy/tropical spastic paraparesis, and HTLV-I uveitis (1–3). In Japan, the estimated lifetime risk of developing ATL in HTLV-I carriers is 6% to 7% for males and 2% to 3% for females (4–6). It takes several decades for HTLV-I–infected cells to reach the final stage of multistep oncogenesis, which is clinically recognized as aggressive ATL (acute-type and lymphoma-type; ref. 7). Molecular interaction of viral genes [e.g., Tax and the HTLV-I basic leucine zipper (HBZ) gene] with the cellular machinery causes various genetic and epigenetic alterations (7–11). However, difficulties in purifying HTLV-I–infected cells in vivo seem to have hindered understanding of the genetic events that are directly involved in the multistep oncogenesis of ATL.

Upregulation or aberrant expression of cell surface markers, such as CCR4 and CD25, is useful for diagnosis of ATL and has been utilized for molecular-targeted therapy (12, 13). However, the expression levels of these markers vary among patients, which often make it difficult to identify ATL cells specifically based on the immunophenotype. Previously, we focused on downregulated markers in acute-type ATL cells, such as CD3 and CD7, and successfully purified ATL cells using the CD3 versus CD7 plot of CD4⁺ cells (14). Analysis of other clinical subtypes (indolent ATLs and HTLV-I asymptomatic carriers; AC) revealed that HTLV-I–infected and clonally expanded cells were purified similarly and that the subpopulations with downregulated CD7 grew concomitantly with the progression of HTLV-I infection (15). Although this type of flow-cytometric analysis was shown to be a useful tool, a substantial subpopulation of T cells shows downregulated expression of CD7 under physiologic (16, 17) and certain pathologic conditions, including autoimmune disorders, viral infection, and hematopoietic stem cell transplantation (18–23).

Recently, Sasaki and colleagues reported ectopic overexpression of the cell adhesion molecule 1/tumor suppressor in lung cancer 1 (CADM1/TSLC1) gene in primary acute-type ATL cells based on expression profile analysis (24, 25). CADM1 (/TSLC1) is a cell-adhesion molecule that was originally identified as a tumor suppressor in lung cancers (25, 26). In addition, numbers of CD4⁺CADM1⁺ cells have been found to be significantly correlated with the proviral load (PVL) in both ATLs and HTLV-I asymptomatic carriers (25, 27). Thus, CADM1 is a good candidate marker of HTLV-I–infected cells. In the present study, we incorporated CADM1 into our flow-cytometric analysis. In the CADM1 versus CD7 plot of CD4⁺ cells, HTLV-I–infected and clonally expanded cells were efficiently enriched in the CADM1⁺ subpopulations regardless of disease status. In these cells, stepwise CD7 downregulation (from dimly positive to negative) occurred with disease progression. The proportion of the three subpopulations observed in this plot [P, CADM1^{negative(neg)}CD7^{positive(pos)}; D, CADM1^posCD7^dim; and N, CADM1^pos CD7^neg] accurately reflected the disease status in HTLV-I infection. The analysis of comprehensive gene expression in each subpopulation revealed that the expression profile of CADM1⁺ subpopulations in indolent ATLs showed similarities with that in asymptomatic carriers with high PVL; yet, it was distinct from that in aggressive ATLs. These D and N subpopulations were indicative of HTLV-I–infected cells in the intermediate stage of ATL development.

TL-Om1, an HTLV-I–infected cell line (28), was provided by Dr. Sugamura (Tohoku University, Sendai, Japan). The MT-2 cell line was a gift from Dr. Miyoshi (Kochi University, Kochi, Japan) and ST-1 was from Dr. Nagai (Nagasaki University, Nagasaki, Japan). Peripheral blood samples were collected from in-patients and out-patients at our hospital, as described in our previous reports (14, 15). As shown in Supplementary Table S1, 26 cases were analyzed (10 cases of asymptomatic carrier; 5 cases of smoldering-type; 6 cases of chronic-type; and 5 cases of acute-type). All patients with ATL were categorized into clinical subtypes according to Shimoyama's criteria (12, 29). Patients with various complications, such as autoimmune disorders and systemic infections, were excluded. Lymphoma-type patients were also excluded because ATL cells are not considered to exist in the peripheral blood of this clinical subtype. Samples collected from six healthy volunteers (mean age 48.8 years; range 34–66 years) were used as normal controls. The present study was approved by the Institutional Review Board of our institute (the University of Tokyo, Tokyo, Japan). Written informed consent was obtained from all patients and healthy volunteers.

Peripheral blood mononuclear cells (PBMC) were isolated from whole blood by density gradient centrifugation, as described previously (14). An unlabeled CADM1 antibody (clone 3E1) and an isotype control chicken immunoglobulin Y (IgY) antibody were purchased from MBL. These were biotinylated (primary amine biotinylation) using biotin N-hydroxysuccinimide ester (Sigma-Aldrich). Pacific Orange-conjugated anti-CD14 antibody was purchased from Caltag-Invitrogen. All other antibodies were obtained from BioLegend. Cells were stained using a combination of biotin-CADM1, allophycocyanin (APC)-CD7, APC-Cy7-CD3, Pacific Blue-CD4, and Pacific Orange-CD14. After washing, phycoerythrin-conjugated streptavidin was applied. Propidium iodide (Sigma-Aldrich) was added to the samples to stain dead cells immediately before flow cytometry. A FACSAria instrument (BD Immunocytometry Systems) was used for all multicolor flow cytometry and fluorescence-activated cell sorting (FACS). Data were analyzed using FlowJo software (TreeStar). The gating procedure for a representative case is shown in Supplementary Fig. S1.

PVL in FACS-sorted PBMCs was quantified by real-time quantitative PCR (TaqMan method) using the ABI Prism 7000 sequence detection system (Applied Biosystems), as described previously (14, 30).

HTLV-I HBZ gene amplification was performed as described previously (25). Briefly, the 25-μL PCR mixture consisted of 20 pmol of each primer, 2.0 μL of mixed deoxynucleotide triphosphates (2.5 mmol/L each), 2.5 μL of 10× PCR buffer, 1.5 μL of MgCl₂ (25 mmol/L), 0.1 μL of AmpliTaq Gold DNA Polymerase (Applied Biosystems), and 20 ng of DNA extracted from cell lines and clinical samples. The PCR consisted of initial denaturation at 94°C for 9 minutes, 30 cycles of 94°C for 30 seconds, 57°C for 30 seconds, and 72°C for 45 seconds, followed by 72°C for 5 minutes. The β-actin gene (ACTB) was used as an internal reference control. The primer sequences used were as follows: HBZ forward, 5′-CGCTGCCGATCACGATG-3′; HBZ reverse, 5′-GGAGGAATTGGTGGACG-3′; ACTB forward, 5′-CGTGCTCAGGGCTTCTT-3′; and ACTB reverse, 5′-TGAAGGTCTCAAACATGATCTG-3′. Amplification with these pairs of oligonucleotides yielded 177-bp HBZ and 731-bp β-actin fragments.

FISH analysis was performed to detect HTLV-I proviral DNA in mononuclear cells that had been FACS-sorted on the basis of the CADM1 versus CD7 plot. These samples were sent to a commercial laboratory (Chromosome Science Labo Inc.), where FISH analysis was performed. Briefly, pUC/HTLV-I plasmid containing the whole-HTLV-I genome was labeled with digoxigenin by the nick translation method, and was then used as a FISH probe. Pretreatment, hybridization, and washing were performed according to standard laboratory protocols. To remove fluorochrome-labeled antibodies attached to the cell surface, pretreatment consisted of treatment with 0.005% pepsin and 0.1 N HCl. The FISH probe was detected with Cy3-labeled anti-digoxigenin antibody. Cells were counterstained with 4′, 6 diamidino-2-phenylindole. The results were visualized using a DMRA2 conventional fluorescence microscope (Leica) and photographed using a Leica CW4000 cytogenetics workstation. Hybridization signals were evaluated in approximately 100 nuclei.

For clonality analysis, inverse long PCR was performed as described previously (14). First, 1 μg genomic DNA extracted from the FACS-sorted cells was digested with PstI or EcoRI at 37°C overnight. RNase A (Qiagen) was added to remove residual RNA completely. DNA fragments were purified using a QIAEX2 Gel Extraction Kit (Qiagen). The purified DNA was self-ligated with T4 DNA ligase (Takara Bio) at 16°C overnight. After ligation of the EcoRI-digested samples, the ligated DNA was further digested with MluI, which cuts the pX region of the HTLV-I genome and prevents amplification of the viral genome. Inverse long PCR was performed using Tks Gflex DNA Polymerase (Takara Bio). For the PstI-treated group, the forward primer was 5′-CAGCCCATTCTATAGCACTCTCCAGGAGAG-3′ and the reverse primer was 5′-CAGTCTCCAAACACGTAGACTGGGTATCCG-3′. For the EcoRI-treated template, the forward primer was 5′-TGCCTGACCCTGCTTGCTCAACTCTACGTCTTTG-3′ and the reverse primer was 5′-AGTCTGGGCCCTGACCTTTTCAGACTTCTGTTTC-3′. Processed genomic DNA (50 ng) was used as a template. The reaction mixture was subjected to 35 cycles of denaturation (94°C, 30 seconds) and annealing plus extension (68°C, 8 minutes). Following PCR, the products were subjected to electrophoresis on 0.8% agarose gels. Fourteen patient samples were analyzed. For samples from which a sufficient amount of DNA was extracted, PCR was generally performed in duplicate.

Total RNA was extracted from each subpopulation in the CADM1 versus CD7 plot using TRIzol (Invitrogen) according to the manufacturer's protocol. Details of the clinical samples used for microarray analyses are shown in Supplementary Table S1. Treatment with DNase I (Takara Bio) was conducted to eliminate genomic DNA contamination. The quality of the extracted RNA was assessed using a BioAnalyzer 2000 system (Agilent Technologies). The RNA was then Cy3-labeled using a Low Input Quick Amp Labeling Kit (Agilent Technologies). Labeled cRNA samples were hybridized to 44K Whole Human Genome Oligonucleotide Microarrays (Agilent Technologies) at 65°C for 17 hours. After hybridization, the microarrays were washed and scanned with a Scanner C (Agilent Technologies). Signal intensities were evaluated by Feature Extraction 10.7 software and then analyzed using Gene Spring 12.0 software (Agilent Technologies). Unsupervised two-dimensional hierarchical clustering analysis (Pearson correlation) was performed on 10,278 genes selected by one-way ANOVA (P < 0.05). The dataset for these DNA microarrays has been deposited in Gene Expression Omnibus (accession number GSE55851).

The expression levels of the microRNA miR-31 were quantified using a TaqMan-based MicroRNA Assay (Applied Biosystems), as described previously (31), and normalized to RNU48 expression level. Helios mRNA transcript variants were examined using reverse transcription PCR (RT-PCR) with Platinum Taq DNA Polymerase High Fidelity (Invitrogen), as described previously (32). To detect and distinguish alternative splicing variants, PCR analyses were performed with sense and antisense primer sets specific for the first and final exons of the Helios gene. The PCR products were then sequenced to determine the exact type of transcript variant. A mixture of Hel-1, Hel-2, Hel-5, and Hel-6 cDNA fragments was used as a “Helios standard” in the electrophoresis of RT-PCR samples.

The clinical profiles of the 32 cases analyzed are shown in Supplementary Table S1. We first examined CADM1 expression in each subpopulation (H, I, and L) of the CD3 versus CD7 plot. Representative data (for a case of smoldering ATL) are shown in Fig. 1A. The results demonstrate that CADM1 was expressed almost exclusively in the I and L subpopulations. Drawing a CADM1 versus CD7 plot for CD4⁺ cells revealed three distinct subpopulations (P, CADM1^negCD7^pos; D, CADM1^posCD7^dim; and N, CADM1^pos CD7^neg). As shown in Fig. 1B, the P, D, and N subpopulations corresponded to the H, I, and L subpopulations in the CD3 versus CD7 plot. In the previous CD3 versus CD7 plot, the lower case (AC no. 2) showed three distinct subpopulations. However, in the upper case (smoldering no. 12), the H and I subpopulations substantially intermingled with each other and were not clearly separated. In contrast, the CADM1 versus CD7 plot clearly revealed three distinct subpopulations in both cases.

On the basis of previous reports (25, 27), we expected HTLV-I–infected cells to be enriched in the CADM1⁺ subpopulations in our analysis. Figure 2A shows the PVL measurements of the three subpopulations in the CADM1 versus CD7 plot for three representative cases. HTLV-I–infected cells were highly enriched in the CADM1⁺ subpopulations (D and N). The PVL data indicate that most of the cells in the D and N subpopulations were HTLV-I infected. Figure 2B shows the results of semiquantitative PCR of the HBZ gene in representative cases. In the D and N subpopulations, the HBZ gene was amplified to the same degree as in the HTLV-I–positive cell line. To confirm these results, FISH was performed in one asymptomatic carrier. As shown in Supplementary Fig. S2, HTLV-I–infected cells were highly enriched in the D and N subpopulations, which supports the results of the PVL analysis and semiquantitative PCR of the HBZ gene. In the FISH analysis, percentages of HTLV-I–infected cells in D and N did not reach 100%. This may have been due to a technical issue. Because the cells subjected to FISH analysis were sorted by FACS, several fluorochrome-conjugated antibodies may have remained on their surfaces, even after treatment with protease.

Compared with the CD3 versus CD7 plot, the CADM1 versus CD7 plot was revealed to be clear in its distinction of the three subpopulations and efficient in enrichment of HTLV-I–infected cells. On the basis of these findings, we analyzed clinical samples of asymptomatic carriers and three clinical subtypes of ATL: the smoldering, chronic, and acute subtypes. Data for representative cases, presented in Fig. 3A, suggest that the continual changes in the proportions of the three subpopulations are associated with disease progression. In the CADM1 versus CD7 plot, normal control samples showed a P-dominant pattern. With progression of the disease from the asymptomatic carrier state with a low PVL to that with a high PVL, and to indolent-type ATL, the D and N subpopulations increased gradually. As the disease further progressed to acute-type ATL, the N subpopulation showed remarkable expansion. Data for all analyzed samples are presented in Fig. 3B. The results suggest that the CADM1 versus CD7 plot of peripheral blood samples represents progression of the disease in HTLV-I carriers. Data for the normal control cases analyzed are shown in Supplementary Fig. S3. In all normal controls, the percentages of the D and N subpopulations were low. Supplementary Fig. S4 shows temporal data for a patient with chronic-type ATL who progressed from stable disease to a relatively progressive state and the concomitant change in the flow cytometry profile.

To characterize the three subpopulations further, the clonal composition of each subpopulation was analyzed by inverse long PCR, which amplifies part of the provirus long terminal repeat and the flanking genomic sequence of the integration sites. Cells in each subpopulation were sorted by FACS, and subjected to inverse long PCR analysis. Representative results for smoldering-, chronic-, and acute-type ATL samples are presented in Fig. 4A. Major clones, indicated by intense bands, were detected in the D and N subpopulations. The major clones in the D and N subpopulations in each case were considered to be the same based on the sizes of the amplified bands, suggesting that clonal evolution is accompanied by downregulation of CD7 expression. Fig. 4B shows representative results for three cases of asymptomatic carrier. In all cases, weak bands in the P subpopulation were visible, indicating that this population contains only minor clones. In these asymptomatic carriers, the proportion of abnormal lymphocytes and PVL increases from left to right. The consistent increase in the D and N subpopulations, together with growth of major clones as shown in the inverse PCR analysis, were considered to reflect these clinical data.

To determine the molecular basis for the biologic differences among the three subpopulations in the CADM1 versus CD7 plot, we next characterized the gene-expression profiles of the subpopulations of the following clinical subgroups: asymptomatic carriers (n = 2), smoldering-type ATLs (n = 2), chronic-type ATL (n = 1), acute-type ATLs (n = 3), and normal controls (n = 3). The two asymptomatic carriers (nos. 5 and 9) had high PVLs (11.6 and 26.2%, respectively) and relatively high proportions of D and N subpopulations (Supplementary Table S1). Unsupervised hierarchical clustering analysis of the results revealed three clusters (A, B1, and B2) or two major clusters A and B, where A is composed solely of the samples of the acute-type N subpopulation and B is subdivided into two clusters (B1 and B2; Fig. 5A). The B2 cluster is composed of the P subpopulation of all clinical subtypes and of normal controls, whereas the B1 cluster is composed of the D and N subpopulations of asymptomatic carriers and indolent ATLs (smoldering- and chronic-type).

Figure 5B shows a Venn diagram of the upregulated genes in the D subpopulation (left) or the N subpopulation (right) common to asymptomatic carriers (n = 2) and indolent ATLs (n = 3). These diagrams demonstrate that the changes in the gene expression profiles of the D and N subpopulations of asymptomatic carriers were similar to those of indolent ATLs. Furthermore, the gene expression profiles of the D and N subpopulations of asymptomatic carriers and indolent ATLs were similar (Fig. 5C). In contrast, the upregulated genes showed distinct differences between the N subpopulation of acute-type ATL and that of indolent ATLs and asymptomatic carriers, although approximately 70% were common to both (Fig. 5D).

To determine whether the novel subpopulations identified had other properties in common with ATL cells, we examined miR-31 levels and Helios mRNA patterns in sorted subpopulations (31, 32). Expression of miR-31 decreased drastically in the D subpopulation derived from indolent ATLs and asymptomatic carriers, and was even lower in the N subpopulation derived from asymptomatic carriers and indolent/acute ATLs (Fig. 6A). In addition, examination of Helios mRNA transcript variants revealed that expression levels of Hel-2, which lacks part of exon 3, were upregulated in the D and N subpopulations of asymptomatic carriers and indolent ATLs, and it was dominantly expressed in the N subpopulation of acute ATLs (Fig. 6B).

Supplementary Fig. S5 presents a summary of this study. The representative flow-cytometric profile shows how the CADM1 versus CD7 plot reflects disease progression in HTLV-I infection. The plot together with the gene expression profiles clearly distinguished the subpopulations of distinct oncogenic stages. The groups classified according to gene expression profile are shown as blue, yellow, and red and are superimposed on the CADM1 versus CD7 plot. Collectively, our data suggest that CADM1 expression and stepwise downregulation of CD7 were closely associated with clonal expansion of HTLV-I–infected cells in ATL progression.

We showed that the CADM1 versus CD7 plot is capable of discriminating clonally expanding HTLV-I–infected cells in indolent ATLs and even in asymptomatic carriers, as well as in acute-type ATLs. Our analysis demonstrated efficient enrichment of HTLV-I–infected cells in the CADM⁺ subpopulations (D and N in the CADM1 vs. CD7 plot), based on the results of real-time PCR (PVL analysis), semiquantitative PCR analysis of the HBZ gene, and FISH analysis (Fig. 2 and Supplementary Fig. S2). Furthermore, the CADM1 versus CD7 plot was shown to discriminate the three subpopulations more clearly than the CD3 versus CD7 plot (Fig. 1). Clonality analysis of ATLs and asymptomatic carriers (Fig. 4A and B) revealed that CADM1⁺ subpopulations (D and N) contained clonally expanded HTLV-I–infected cells, whereas cells in the P subpopulation (CADM1⁻) did not show clonal expansion in this analysis. Current molecular analyses of ATL cells have been limited to HTLV-I–infected cell lines and primary cells from acute/lymphoma type ATL, because in these cases, the predominant expanding clones are readily available with relatively high purity. However, the separation of clonally expanding ATL cells from indolent ATLs and asymptomatic carriers has not yet been achieved. The CADM1 versus CD7 plot from FACS allows efficient purification of such clones in vitro.

In an unsupervised clustering analysis of the gene expression data, the D and N subpopulations of asymptomatic carriers/indolent ATLs were grouped together, suggesting that the biologic characteristics of these subpopulations are similar (Fig. 5A and B) but distinct from the N subpopulation of acute-type ATLs (Fig. 5D). These results support the notion that in indolent ATLs and even in asymptomatic carriers, the D and N subpopulations are clonally expanding cells representing the intermediate oncogenic stage. Although the D and N subpopulations have similar gene expression profiles (Fig. 5C), there are potentially important differences distinguishing these subpopulations, according to the apparent decrease in the D subpopulation and increase in the N subpopulation that were observed as the disease progressed from indolent to acute-type ATL (Fig. 3). Detailed analysis of the genomic and epigenomic differences between these two subpopulations will provide us with information about the genomic and epigenomic lesions that are involved in disease progression. Another important finding is that the expression profiles of cells in the N subpopulation of indolent and acute-type ATLs showed significant differences, even though the majority of the genes were common to both groups (Fig. 5D). Characterization of the genes that show distinct expression patterns will reveal the molecular events that contribute to the progression from indolent to aggressive ATLs.

To address whether the emerging molecular hallmark of ATL was conserved in the novel subpopulations identified, we examined the miR-31 level and Helios mRNA pattern in sorted subpopulations (Fig. 6). Through integrative analyses of ATL cells, we recently showed that the expression of miR-31, which negatively regulates noncanonical NF-κB signaling by targeting NIK, is genetically and epigenetically suppressed in ATL cells, leading to persistent NF-κB activation, and is thus inversely correlated with the malignancy of the cells (31). The miR-31 levels in the P subpopulations in asymptomatic carriers and indolent ATLs were as high as those in normal P subpopulations, PBMCs, and resting T cells, whereas those in the D subpopulations decreased significantly and those in the N subpopulations were as low as in acute-type N subpopulations (Fig. 6A). Previously, we also identified ATL-specific aberrant splicing of Helios mRNA and demonstrated its functional involvement in ATL (32). As shown in Fig. 6B, the Hel-2 type variant, which lacks part of exon 3 and thus lacks one of the four DNA-binding zinc-finger domains, accumulated in the D and N subpopulations of asymptomatic carriers and indolent ATLs, and was dominantly expressed in the N subpopulation of acute-type ATLs. Collectively, the molecular abnormality of ATL cells became evident in the gradual progression from P to D to N, even in asymptomatic carriers, strongly supporting the notion that the CADM1/CD7 expression pattern correlates with the multistep oncogenesis of ATL.

One of the more remarkable findings in the expression profile analysis was that the D and N subpopulations of asymptomatic carriers clustered within the same group as those of the indolent ATL cases (Fig. 5A and B). The asymptomatic carriers used in this analysis had high PVLs and relatively high proportions of the D and N subpopulations (Supplementary Table S1). In addition, mono- or oligoclonal expansion of the HTLV-I–infected cells was demonstrated in these cases. HTLV-I–infected cells in the D and N subpopulations of these asymptomatic carriers comprise clonally expanding cells that are potentially at the premalignant and intermediate stages according to their clonality, comprehensive gene expression profile, miR31 expression, and aberrant RNA splicing, all indicating that they can be categorized as asymptomatic carriers with high risk of developing into ATL, requiring careful follow-up (15, 30, 33, 34). Our flow-cytometric analysis of PBMCs from asymptomatic carriers using the CADM1 versus CD7 plot may provide a powerful tool for identifying high-risk asymptomatic carriers. Certain indolent ATL cases are difficult to distinguish from asymptomatic carriers, according to Shimoyama's criteria based on the morphologic characteristics determined by microscopic examination. Characterization of peripheral blood T cells by the CADM1 versus CD7 plot may provide useful information for clinical diagnosis.

According to Masuda and colleagues, manipulation of CADM1 gene expression in leukemic cell lines suggested that CADM1 expression confers upon ATL cells tissue invasiveness and a growth advantage (35). The mechanism by which HTLV-I infection regulates CADM1 expression and the significance of CADM1 expression in ATL oncogenesis will require clarification by future studies.

Finally, as summarized in Supplementary Fig. S5, we demonstrated that (1) HTLV-I–infected and clonally expanded cells are efficiently enriched in CADM1⁺ subpopulations; (2) the proportions of the three subpopulations in the CADM1 versus CD7 plot, discriminated by CADM1 expression and stepwise downregulation of CD7, accurately reflect the disease stage in HTLV-I infection; and (3) the CADM1⁺CD7^dim/neg subpopulations are at the intermediate stage of ATL progression and can be identified even in asymptomatic carriers. These findings will help to elucidate the molecular events involved in multistep oncogenesis of ATL.

No potential conflicts of interest were disclosed.

Conception and design: S. Kobayashi, T. Watanabe, K. Uchimaru

Development of methodology: T. Ishigaki, T. Yamochi, N. Watanabe

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Kobayashi, E. Watanabe, K. Yuji, N. Oyaizu, S. Asanuma, A. Tojo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Kobayashi, K. Nakano, T. Ishigaki, N. Oyaizu, M. Yamagishi, T. Watanabe

Writing, review, and/or revision of the manuscript: S. Kobayashi, K. Nakano, A. Tojo, T. Watanabe, K. Uchimaru

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Ishigaki, N. Ohno, N. Watanabe

Study supervision: A. Tojo, T. Watanabe, K. Uchimaru

The authors thank Drs. Kazunari Yamaguchi (National Institute of Infectious Diseases, Tokyo, Japan) and Yoshinori Murakami (the University of Tokyo) for their constructive comments; Yuji Zaike (Clinical Laboratory, Research Hospital, Institute of Medical Science, the University of Tokyo) for his excellent technical advice; Keisuke Takahashi, Sanae Suzuki, and members of our laboratory for assistance; and the hospital staff, which has made a commitment to providing high-quality care to all patients. The English in this document has been checked by at least two professional editors, both native speakers of English.

This work was supported by grants-in-aid for scientific research awarded to K. Uchimaru (no. 22591028) and T. Watanabe (no. 23390250) by the Ministry of Education, Culture, Sports, Science and Technology of Japan.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.