Variant Ciz1 is a circulating biomarker for early-stage lung cancer

Gillian Higginsa,b, Katherine M. Ropera,b, Irene J. Watsona,b, Fiona H. Blackhallc, William N. Romd, Harvey I. Passe, Justin F. X. Ainscoughf, and Dawn Coverleya,b,1
a Cizzle Biotech, University of York, Yorkshire YO10 5DD, United Kingdom; bDepartment of Biology, University of York, Yorkshire YO10 5YW, United Kingdom; cPaterson Institute for Cancer Research, University of Manchester, Lancashire M20 4BX, United Kingdom; dDivision of Pulmonary, Critical Care and Sleep Medicine, New York University School of Medicine, New York, NY 10016; eNew York University Langone Medical Center, New York, NY 10016; and fSchool of Medicine, Leeds University, Yorkshire LS2 9JT, United Kingdom

Edited by Dennis A. Carson, University of California at San Diego, La Jolla, CA, and approved September 20, 2012 (received for review June 15, 2012)

There is an unmet need for circulating biomarkers that can detect early-stage lung cancer. Here we show that a variant form of the nuclear matrix-associated DNA replication factor Ciz1 is present in 34/35 lung tumors but not in adjacent tissue, giving rise to stable protein quantiable by Western blot in less than a microliter of plasma from lung cancer patients. In two independent sets, with 170 and 160 samples, respectively, variant Ciz1 correctly identied patients who had stage 1 lung cancer with clinically useful accuracy. For set 1, mean variant Ciz1 level in individuals without diagnosed tumors established a threshold that correctly classied 98% of small cell lung cancers (SCLC) and non-SCLC patients [receiver operator characteristic area under the curve (AUC) 0.958]. Within set 2, comparison of patients with stage 1 non-SCLC with asymptomatic age-matched smokers or individuals with benign lung nodules correctly classied 95% of patients (AUCs 0.913 and 0.905), with overall specicity of 76% and 71%, respectively. Moreover, using the mean of controls in set 1, we achieved 95% sensitivity among patients with stage 1 non-SCLC patients in set 2 with 74% specicity, demonstrating the robustness of the classication. RNAimediated selective depletion of variant Ciz1 is sufcient to restrain the growth of tumor cells that express it, identifying variant Ciz1 as a functionally relevant driver of cell proliferation in vitro and in vivo. The data show that variant Ciz1 is a strong candidate for a cancer-specic single marker capable of identifying early-stage lung cancer within at-risk groups without resort to invasive procedures.

ung cancer is the leading cause of cancer death worldwide. Approximately 80% of lung tumors are classied as nonsmall cell lung cancer (NSCLC), including squamous cell carcinoma and adenocarcinoma, and the remainder as small cell type (SCLC). SCLCs are primarily neuroendocrine in origin, ranging from low-grade typical carcinoid to high-grade neuroendocrine tumors (HGNTs), although some HGNTs are classied as large cell type (1, 2). The risk of lung cancer is increased dramatically by smoking, and genetic factors appear to play a role in our ability to deal with smoking-related damage. However, clearly heritable forms of increased lung cancer risk are not common. Lung cancer diagnosis and staging relies heavily on imaging, suggesting that imaging may offer a route to early detection in high-risk groups. Although the impact of early detection on survival has been questioned, several studies have looked at the potential benet (3), and it recently became clear that screening with low-dose spiral CT can achieve a signicant reduction in mortality among heavy smokers (4). However, because around a quarter of individuals require follow-up procedures to investigate suspicious imaging results, the cost of this approach is extremely high, highlighting the need for a second-line noninvasive test that can conrm malignancy. Here we present evidence that proteinlevel detection of a variant form of the nuclear matrix protein Ciz1 has the potential to meet this need. Ciz1 promotes initiation of mammalian DNA replication, where it helps coordinate the sequential functions of cyclin E- and A-dependent protein kinases (5). It interacts directly with cyclins E and A (6), with

Results As part of a gene-focused analysis of function, we cloned human Ciz1 from a SCLC cell line and recovered multiple variants, including a prevalent transcript in which 24 nucleotides from the 3 end of exon 14 (2475_2498del) is excluded, leading to in-frame deletion of eight amino acids (VEEELCKQ). Analysis of the sequence surrounding exons 14 and 15 revealed a second splice donor site within exon 14 (2475/6) that could support alternative splicing. Location identiers refer to Ciz1 reference sequence NM_012127.2. We refer to the whole of predicted exon 14 as 14a, the shorter alternative as 14b, and Ciz1 transcripts harboring 14b as b-variant. Transcript frequencies among ESTs that map to the Ciz1 Unigene cluster Hs. 212395 suggested that b-variant is prevalent in neuroendocrine lung tumors, and this prevalence was conrmed by analysis of SCLC cell lines using independent detection methods (Fig. S1 and below). As a nuclear matrix protein characterized by resistance to harsh extraction conditions, b-variant could offer a robust biomarker with potential to remain stable and detectable in body uids. Consistent with this idea, an afnity-puried polyclonal antibody directed against the unique peptide encoded at the junction of exon 14b/exon15 (Fig. S2) detected b-variant protein by Western blot in 1 L of plasma from patients with SCLC and NSCLC but not from healthy individuals (Fig. 1B). A diffuse but specic band of 5060 kDa (Fig. S2) is reproducibly seen and remained stable even after extended periods of storage at 4 C.

Author contributions: J.F.X.A. and D.C. designed research; G.H., K.M.R., I.J.W., and D.C. performed research; F.H.B., W.N.R., and H.I.P. contributed new patient samples; G.H. and D.C. analyzed data; and J.F.X.A. and D.C. wrote the paper. Conict of interest statement: The ndings reported in this paper arise directly out of basic research at the University of York (York, Yorkshire, United Kingdom) on the Ciz1 protein and its role in the spatial and temporal organization of DNA replication. The ndings are the result of an academic and commercial collaboration, funded in part by Cizzle Biotech, which is a spin-out company of the University of York. G.H. and D.C. are partially funded by Cizzle Biotech. D.C. and J.F.X.A hold shares in Cizzle Biotech. This article is a PNAS Direct Submission.
1

To whom correspondence should be addressed. E-mail: dawn.coverley@york.ac.uk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1210107109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1210107109

PNAS Early Edition | 1 of 8

MEDICAL SCIENCES

CDK2, and with the cyclin-dependent kinase inhibitor p21 (7) and also plays an indirect role in DNA replication by modulating the expression of genes, including cyclin D, that inuence cell proliferation (8). Normally, Ciz1 is attached to the salt- and nuclease-resistant protein component of the nucleus referred to as the nuclear matrix and resides within foci that partially colocalize with sites of DNA replication (9), implicating Ciz1 in the spatial organization of DNA replication. Here we describe a Ciz1 variant that lacks part of a C-terminal domain involved in nuclear matrix attachment (Fig. 1A). Expression of this stable variant is apparently restricted to tumor cells, making clinical exploitation as a cancer biomarker highly tractable.

PNAS PLUS

Fig. 1. Variant Ciz1 protein in 170 samples in plasma set 1. (A) Ciz1 gene showing translated exons (numbered) and the alternative-splicing event at the exon 14/15 junction which gives rise to b-variant Ciz1. Exons that encode DNA replication domain (5) and nuclear matrix anchor domain (9) are indicated by dotted lines above. Exons that are commonly excluded from known Ciz1 variants (23) are shaded in dark gray. A complete representation of Ciz1 alternative splice variants was assembled previously (24), and transcript diversity was discussed recently (25, 26). (B) Western blot showing b-variant protein detected with antibody 2B in 1 L of plasma from patients with SCLC or NSCLC plus ve samples from individuals with no diagnosed disease. Endogenous Ig is used to normalize for loading (control). (C ) Box-and-whisker plots showing the median, upper, and lower quartile, range, and outliers for data derived from Western blots by densitometry. Results for a total of 119 pretreatment lung cancer patients (Left) with the indicated type and stage of disease (Right), plus 51 samples from individuals with no disease or from patients with COPD, asthma, or anemia are shown. Individual sample values are given in Dataset S1. Using a threshold set at the mean of the noncancer samples, the test correctly classied 98% of all 119 lung cancer patients, with specicity of 85%. (D) ROC curve, with 95% condence interval, generated for all 170 samples (AUC is 0.958). Students two-tailed t test with unequal variance returned a P value of <0.0001 for the noncancer samples compared with all sample sets from individuals with lung cancer.

Quantication of b-variant protein in 119 SCLC and NSCLC samples (summary information in Table 1 with sample-specic information in Dataset S1) gave a mean signal intensity of 95.8 densitometry units with an SD of 51.8, indicating considerable heterogeneity among patients. In contrast, the mean signal intensity for 51 individuals without diagnosed malignancy (who had chronic obstructive pulmonary disease, asthma, anemia, or no disease) was just 14.1 (SD 15.2). Unexpectedly, limited- and extensive-stage SCLC and stages 1, 2, 3, and 4 NSCLC were all signicantly different from the control groups when analyzed separately (Fig. 1C), with a trend toward expression increasing with stage for stages 1, 2, and 3 NSCLC. Receiver operating characteristic (ROC) analysis for all 170 samples returned an area under the curve (AUC) of 0.958 (Fig. 1D), demonstrating considerable discriminatory power in this context. To our knowledge, and despite considerable effort and progress with circulating biomarkers for other cancers (10, 11), no other single biomarker is capable of achieving this level of discrimination between
2 of 8 | www.pnas.org/cgi/doi/10.1073/pnas.1210107109

patients with early-stage lung cancer and those without lung cancer. However, for maximum clinical utility a blood test must be able to identify individuals with early-stage lung cancer within populations that are at high risk of developing the disease. An important potential application is as a test for lung cancer in individuals with lung nodules identied by CT. A test that can positively identify those individuals with lung cancer could help reduce the frequency of surgical intervention and favorably affect both cost and outcome. To look specically at the discriminatory power of b-variant Ciz1 in comparison with (i ) age-matched smokers with 10 y of cigarette smoking history but without diagnosed cancer and (ii ) individuals with nonmalignant lung nodules or inammatory lung disease, we analyzed a second, independent, archived set of case and control plasma samples (set 2, summary information in Table 1 with sample-specic information in Dataset S2), using high-throughput Western blots and just 0.5 L of each sample. Comparison of the smokers (median age 64.5 y) with 20 individuals with stage 1 squamous
Higgins et al.

Table 1. Summary characteristics of patients in plasma sets 1 and 2

Diagnosis/classication Set 1 Stage 1 NSCLC Stage 2 NSCLC Stage 3 NSCLC Stage 4 NSCLC Limited stage SCLC Extensive stage SCLC All cancer COPD Asthma Anemia No diagnosed disease All noncancer Set 2 Stage 1 squamous cell carcinoma Stage 1 adenocarcinoma Benign lung nodules Inammatory lung disease Smokers No. of patients 170 52 6 16 14 21 10 119 5 12 20 14 51 160 20 20 20 20 80 Median age 62 66.5 66 69 67 64 65 59 42.5 47.5 36 46 67 71.5 60 67.5 64.5 Sex (%F) 40 50 38 43 48 70 44 20 58 80 29 55 20 60 85 60 55 Median pack years* N/d N/d N/d N/d N/d N/d N/d N/d N/d N/d N/d N/d 45 41 20 30 34 Mean b-variant (SD) 65.7 94.8 103.6 85.0 112.0 118.0 95.8 0.3 17.6 23.7 5.9 14.1 64.3 35.2 8.6 6.3 6.35 40.0 29.8 26.3 52.1 67.1 70.1 51.8 4.3 16.9 8.0 12.3 15.2 32.0 14.4 22.0 0.4 12.2

N/d, no data. *One pack year is dened as 20 cigarettes smoked every day for a year.

cell carcinoma (median age 67 y) and 20 patients with stage 1 adenocarcinoma (median age 71.5 y) revealed clear discriminatory power in both contexts and ROC AUC values in excess of 0.9 (Fig. 2). Western blot images for all samples are shown in Fig. S3, and raw and normalized results are given in Dataset S2. These data show that b-variant Ciz1 has signicant potential for detecting limited-stage lung cancer among at-risk groups. The overall performance with set 1 and set 2 is shown in Table 2. Encouragingly, using a threshold set at the mean of 40 individuals with benign lung nodules or inammatory diseases of the lung (set 2), b-variant Ciz1 correctly classied 95% of patients with stage 1 lung cancer, with an overall specicity of 71% (set 2). Moreover, when the threshold is set at the mean of all of the noncancers in the discovery set (set 1), 95% of the patients with stage 1 lung cancer in the validation set (set 2) are correctly classied, with an overall specicity of 75%. Under both circumstances the false-positive rate was 50%, suggesting that, if it

were applied as a secondary screen to high-risk groups with suspicious CT results (4), b-variant Ciz1 could halve the number of individuals referred for costly and invasive follow-up procedures. In the context of high-risk smokers without CT-based diagnosis, the biomarker also would appear to have utility, with 95% of patients with stage 1 lung cancer (in set 2) falling above the mean of the smokers and only 34% of smokers being placed in the suspicious category. Thus, positioning the test either before or after CT could offer signicant advantages over the use of CT alone. Consistent with the presence of b-variant Ciz1 in patient plasma, immunohistochemical analysis of primary tumors from patients with neuroendocrine lung cancer revealed b-variantpositive cells in 34 of 35 patients (Fig. 3A). Staining was heterogeneous in both distribution and level, with positive cells evident throughout the tumor in some individuals and limited to isolated cells in others. These results conrm that b-variant Ciz1 is a tumor antigen

Table 2. Sensitivity and specicity estimates for sample sets 1 and 2

Thresholds used Set 1 Mean of all noncancers in set 1 Mean of all noncancers in set 1 Set 2 Mean of all noncancers in set 2 Mean of all smokers in set 2 Mean of benign and inammatory in set 2 Mean of all noncancers in set 1 Mean of all noncancers in set 1 Mean of all noncancers in set 1 14.07 14.07 False negatives 3/119 (all cancer) 0/52 (stage 1 NSCLC) Sensitivity (%) 98 100 False positives 23/51 23/51 False positive (%) 45 45 Total in wrong group 26/170 23/103 Specicity (%) 85 78

6.27 6.35 7.45 14.07 14.07 14.07

2/40 2/40 2/40 2/40 (stage 1) 2/40 (stage 1) 2/40 (stage 1)

95 95 95 95 95 95

48/120 27/80 21/40 40/120 (all non cancer) 23/80 (smokers) 18/40 (benign and inammatory)

40 34 53 33 29 45

50/160 29/120 23/80 42/160 25/120 20/80

69 76 71 74 79 75

Higgins et al.

PNAS Early Edition | 3 of 8

MEDICAL SCIENCES

PNAS PLUS

Fig. 2. Variant Ciz1 protein in 160 samples in plasma set 2. (A) Box plot showing results for 80 smokers with more than 10 y of smoking history and for 40 patients with stage 1 NSCLC with similar smoking history (Left) and for 20 individuals diagnosed with stage 1 adenocarcinoma, inammatory lung disease (granuloma), benign lung nodules (carcinoid, hamartoma), or stage 1 squamous cell carcinoma (Right), showing lower, median, and upper quartiles and outliers (circles). (B) ROC curve with 95% condence intervals for the indicated comparisons. AUCs are 0.913 when samples from 80 smokers are compared with samples from 40 patients with stage1 lung cancer, 0.905 when samples from 40 patients with benign nodules or inammatory disease are compared with samples from 40 patients with stage1 lung cancer, and 0.909 when all samples from smokers and patients with benign nodules or inammatory disease are compared with all samples from patients with stage1 lung cancer, but are only 0.503 when samples from smokers are compared with samples from patients with benign nodules or inammatory disease. Western blots showing b-variant protein in the 0.5 L of plasma used for each of the 160 samples are shown in Fig. S3; individual b-variant levels and clinical parameters are given in Dataset S2. Students two-tailed t test with unequal variance returned a P value of <0.0001 for the samples from patients with cancer compared with samples either from smokers or from patients with benign nodules or inammatory disease.

with the potential for exploitation as a lung cancer biomarker. Complementary transcript analysis reported at least 40-fold elevation in bulk tumor RNA from neuroendocrine tumors compared with matched adjacent tissue for all three patients tested (Fig. 3B). In addition, b-variant transcript was elevated in 13 SCLC cell lines compared with control lines (Fig. 3C and Fig. S1E),
4 of 8 | www.pnas.org/cgi/doi/10.1073/pnas.1210107109

conrming that expression is common in lung cancer cells. Notably, for both the cell lines and the primary tumors, Ciz1 also is elevated (Fig. 3 B and C), but to a lesser extent. Visualization by immunouorescence revealed that the bvariant protein resides as subnuclear foci that are somewhat larger than WT Ciz1 in the nucleus in SCLC cell line SBC5 but
Higgins et al.

Fig. 3. Expression of Ciz1 b-variant in tumors and cell lines. (A) Sections from SCLC tissue array ARY-HH0117, probed with b-variant antibody 2B (dark brown), showing (i ) normal lung negative for b-variant; (ii and iii ) b-variant in representative SCLC; (iv) SCLC with positive tumor cells and stromal cells; (v) SCLC with positive stromal cells ; (vi ) SCLC negative for b-variant. B-variant was recorded in 29/35 tumors and in stromal cells in 5/35 tumors. Only one pair of tumor sections did not contain b-variantpositive cells. (B) Relative quantication (RQ) of b-variant transcript (primers P1/P2, probe T2) and Ciz1 AD transcript in SCLC tumors of the indicated disease stages plus adjacent tissue from three patients. Results are normalized to actin and calibrated to the rst adjacent sample. (C ) Ciz1 b-variant AD (primers 6/7, probe 16) and replication domain (13/14, probe 7) in lung-derived cell lines, showing average of three technical replicates after normalization to actin and calibration to WI38 cells, with SEM. Sequences are given in Table S3. All three SCLC lines, plus 10 others (Fig. S1), have elevated b-variant. (D) Detergent-resistant b-variant protein (red) in nuclei of SCLC cells, but not normal lung cells, was detected with antibody 2B. DNA is blue. (Scale bar, 5 microns.) (E ) GFP-hCiz1 WT (full length with exon 14a) and GFP-hCiz1 b-variant (with exon 14b). Both proteins form nuclear foci, but foci are larger for the b-variant and are accompanied by localization around the periphery (white arrowhead).

not in the normal lung cell line Wi38 (Fig. 3D) (9). When transfected into NIH 3T3 cells, GFP-tagged Ciz1 constructs with exon 14a or 14b produced protein that was exclusively nuclear and punctate and which resisted extraction (Fig. 3E). Thus, we saw no gross differences in subcellular localization or attachment to nonchromatin nuclear structures. However subtle differences in subnuclear distribution were evident, with the b-variant forming larger aggregates and localizing to the perimeter of the nucleus as well as to foci. Therefore, the loss of eight amino acids from the anchor domain (AD) of Ciz1 does subtly affect the spatial organization of Ciz1, leading us to suggest that it may inuence the functional compartment into which the DNA replication activity of Ciz1 is targeted and possibly, therefore, the execution of replication-coupled events. Notably, normal cells do not tolerate sustained overexpression of ectopic Ciz1 or b-variant Ciz1 [despite an early positive effect on initiation of DNA replication (5), shown previously for WT Ciz1], because this overexpression ultimately results in the loss of nuclear integrity in NIH 3T3 cells and loss of adherence within 33 h of transfection via a mechanism that is not apoptosis (Fig. S4). The increasingly well-characterized role of Ciz1 in DNA replication makes it a candidate therapeutic target with the potential to modulate cell proliferation. Inhibition of Ciz1 expression using siRNA restrains cell proliferation in murine NIH 3T3 cells
Higgins et al.

(5) and human SBC5 cells (Fig. S5). Therefore, if tumor cells that express the b-variant rely on it to support S phase, its inhibition could selectively restrain their proliferation. From a panel of junction-spanning siRNAs that suppress the expression of the b-variant transcript and protein but not Ciz1 isoforms that contain exon 14a (Fig. S6 and Table S1), we selected the most potent sequence (Fig. S5A) and tested its effect on the proliferation of SBC5 SCLC cells (which express endogenous b-variant; Fig. 3 C and D) and also H727 cells (which do not express endogenous b-variant; Fig. 3C and Fig. S5B) when delivered from an inducible shRNA vector. Like RNAi targeted against a constant region of Ciz1, b-variant shRNA effectively restrained cell proliferation in culture (Fig. 4A). As expected, b-variant transcript, but not other forms of Ciz1, was suppressed selectively (Fig. 4B), and this suppression had a signicant impact on b-variant protein (Fig. 4C). Importantly, in a xenograft model of human SCLC in which s.c. tumors were derived from SBC5 cells (12), b-variant shRNA dramatically restrained tumor growth. Similar results were obtained when expression was induced at the time of tumor cell inoculation or after tumors had formed (Fig. 4D). Together, these in vitro and in vivo experiments show that the unique junction generated by alternative splicing of Ciz1 exon 14 can be exploited to suppress variant expression specically and that this suppression is sufcient to inhibit tumor cells that express it, identifying
PNAS Early Edition | 5 of 8

MEDICAL SCIENCES

PNAS PLUS

Fig. 4. Suppression of the expression of the Ciz1 b-variant restrains proliferation of lung cancer cells. (A) Effect of luciferase control shRNA (luc) and b-variant junction-selective RNAi sequence 4 on SBC5 cells, expressed as fold increase in cell number over 4 d after induction of shRNA with doxycycline at days 1 and 3 (gray arrowheads). RNAi sequences are described in Fig. S6. (B) RT-PCR analysis of Ciz1 AD (primers P1/P2), b-variant (P4/P3), and actin (P11/P12) in SBC5 b-variant shRNA 4 cells, without ( dox) and with induction of expression at 26 h (+) or as two doses (++) administered at 26 h and 1 h before isolation. (C ) Detergent-resistant endogenous b-variant protein (antibody 2B, green) in the nucleus of SBC5 b-variant shRNA 4 cells 2 d after induction of shRNA. DNA is blue. (Scale bar, 10 microns.) (D) Xenograft tumor formation in mice. Control mice were inoculated with SBC5 b-variant shRNA 4 cells and maintained in the absence of doxycycline (open circles, group 1). Additional cohorts (lled circles) were given doxycycline continuously from 3 d before inoculation (group 3, Right) or after 21 d (group 2, Left). For the comparison of groups 1 and 2, mice with tumors less than 100 mg at 21 d were discounted, creating cohorts with low variation. Graphs show mean tumor weight from 1015 animals, with SEM. (Data are given in Dataset S3.)

b-variant Ciz1 as a functionally relevant driver of tumor cell proliferation. Discussion The evidence presented here shows that a cancer-associated variant isoform of the nuclear matrix protein Ciz1 holds considerable promise as the basis of a blood test for early-stage lung cancer. As biomarkers, nuclear matrix proteins offer several advantages. They are at the heart of a dysregulated nucleus and are likely to play causative roles in epigenetic control of gene expression. Perhaps more importantly, nuclear matrix proteins are inherently stable, offering practical advantages that make exploitation feasible. As a group they are biochemically dened as the fraction that remains after extraction to remove lipids, soluble proteins, cytoskeleton, chromatin, and associated proteins, which when observed microscopically forms a proteinaceous network (13) made up of constitutive and cell-typespecic components (14). Increasingly nuclear matrix proteins such as NMP22, BCLA4, PC1, and NM179 are gaining credibility as tumor markers for bladder (15), cervix (16), and prostate cancer (17), respectively.
6 of 8 | www.pnas.org/cgi/doi/10.1073/pnas.1210107109

However, despite their promise, the need remains for markers with tumor-specic proles that can be used to complement examination of the structure of nuclei. B-variant Ciz1 is one such molecule, robustly quantiable in the blood of patients who have lung cancer, that is both a driver of DNA replication and tumor cell proliferation and a stable derivative of circulating tumor cells. Notably, this biomarker was identied through intensive gene-focused analysis of expression and function rather than by mutation- or gene-expressionproling approaches. Analysis by high-throughput Western blot offers a direct, visual output claried by sample denaturation that eliminates epitope masking in native samples. Furthermore, fractionation on the basis of molecular weight effectively isolates signal from that of potentially reactive contaminants, giving a clear picture of the potential of the biomarker. Western blot analysis, however, may not report the full dynamic range of biomarker concentrations and is not easily applied outside a research laboratory. Therefore, routine application of a b-variantbased test in a clinical context will require the development of a more streamlined method, such as ELISA, with a simplied quantitative output.
Higgins et al.

Thus, we acknowledge that, although the data presented show that the biomarker holds considerable discriminatory power, the assay ultimately will need to switch platforms and be validated further in that context. Alteration in the structure of nuclei, detected by light microscopy of tumor biopsies, has remained the most reliable method of cancer diagnosis for more than 100 y (18), but the underlying abnormalities have been difcult to dene at the molecular level. Detailed and focused analysis of the relationship between nuclear architecture and essential nuclear events is beginning to identify new players involved in establishing order (19). Moreover, disease-associated molecular changes in these players are now becoming apparent, offering exciting opportunities for the development of new clinically relevant tools. Materials and Methods
Additional methods are described in SI Materials and Methods. Cells. NIH 3T3 cells were grown as described (20). SBC5 cells, obtained from the Japanese Collection of Research Bioresources Cell Bank (JCRB) were grown in Eagles minimal essential medium (Sigma), 1% penicillin, streptomycin, glutamine (Gibco), and 10% (vol/vol) FBS (Biosera). RNAi with siRNA and inducible shRNA is described in SI Materials and Methods. All other cells were from the European Collection of Cell Cultures or JCRB and were cultured as recommended. Quantitative RT-PCR. RNA from cell lines (Table S2) was isolated with TRIzol, and cDNA was synthesized using SuperScript III (Invitrogen). RNA from tumors and adjacent tissue (Cytomyx) was reverse transcribed using M-MLV (Promega) and random primers (Promega) plus oligo-dT primer 1218 (Invitrogen). Primers (Table S3) P6/P7 or P1/P2 (Sigma Aldrich) were coupled with junction probe T2 (MWG) for b-variant Ciz1 and were veried using plasmid templates (Fig. S1). Relative expression was calculated using the comparative cycle threshold (Ct) method (2Ct), and results were expressed relative to normal cells or tissue. Antibodies. B-variantspecic rabbit polyclonal antibodies (2B) were generated by Covalab using the splice-junction sequence (C-DEEEIEVRSRDIS) (Fig. S2A). Sera rst were subtracted for antibodies reactive against anking sequences in a negative control peptide that includes the amino acids absent from b-variant (DEDEEEIEVEEELCKQVRSRDISR) and then were adsorbed to peptide containing the b-variant junction to generate afnity-puried antibody 2B. Puried antibodies were characterized in a range of assays including immunouorescence and Western blot on cells expressing WT or b-variant forms of human GFP-Ciz1 (Fig. S2 B and C ) and lung cancer cells expressing endogenous b-variant Ciz1 with and without specic suppression by b-variant shRNA (Figs. 3D and 4C). Immunodetection and Analysis. For immunouorescence, the b-variant was detected after paraformaldehyde xation using afnity-puried polyclonal antibody 2B (Fig. S2) and then anti-rabbit Alexa Fluor 488 (Invitrogen, Folio Bioscience). Subcellular fractionation was carried out as described (21). For immunohistochemistry, 2B was applied to SCLC tissue microarray ARYHH0117 (Folio Biosciences) and visualized using Ultraview Universal DAB (Ventana) with hematoxylin counterstain. Duplicate cores representing 35 SCLC (grade 4) and tissue from ve normal or cancer-adjacent lungs were analyzed by a certied pathologist at Phylogeny, Inc. Cores were classied as PP (containing mostly b-variantpositive cancer cells), P (containing b-var1. Brambilla E, Travis WD, Colby TV, Corrin B, Shimosato Y (2001) The new World Health Organization classication of lung tumours. Eur Respir J 18(6):10591068. 2. Jones MH, et al. (2004) Two prognostically signicant subtypes of high-grade lung neuroendocrine tumours independent of small-cell and large-cell neuroendocrine carcinomas identied by gene expression proles. Lancet 363(9411):775781. 3. Baldwin DR (2011) Imaging in lung cancer: Recent advances in PET-CT and screening. Thorax 66(4):275277. 4. Aberle DR, et al.National Lung Screening Trial Research Team (2011) Reduced lungcancer mortality with low-dose computed tomographic screening. N Engl J Med 365 (5):395409. 5. Coverley D, Marr J, Ainscough JF-X (2005) Ciz1 promotes mammalian DNA replication. J Cell Sci 118(Pt 1):101112. 6. Copeland NA, Sercombe HE, Ainscough JF, Coverley D (2010) Ciz1 cooperates with cyclin-A-CDK2 to activate mammalian DNA replication in vitro. J Cell Sci 123(Pt 7): 11081115.

iantpositive cancer cells with variable intensity and number, exceeding 10% of cells), PS (containing positive stromal cells), PPS (containing strongly positive stromal cells), or N (containing no positive cells). Images shown in Fig. 3A are derived from (i ) core H8, (ii ) G7. (iii ) D10, (iv) G6, and (v) F8. Plasma Analysis. For sample set 1, 1 L of plasma was resolved by 8% SDS/ PAGE through 12 cm 12 cm maxi-gels (Atto) and was transferred to nitrocellulose by semidry blotting. For samples in set 2, 0.5 L of plasma was denatured using Epage buffer, resolved by electrophoresis through precast E-PAGE 8% 48-well gels, and was transferred to nitrocellulose using iBlot dry blotting system (all Invitrogen). A more detailed description is given in SI Materials and Methods. Membranes were probed with anti-Ciz1 b-variant specic afnity puried polyclonal antibody 2B, diluted 1/500 in PBS, 10% (wt/vol) milk powder, and 0.1% Tween 20. Antibody 2B was detected with peroxidase-IgG light chain-specic, highly cross-adsorbed monoclonal antirabbit antibody (211-032-171; Jackson Immunological Research). Western blots were quantied using Image J (National Institutes of Health), and samples were assigned a value after normalization to loading control and calibration to a constant normal sample (given a value of 0), and to an NSCLC sample (given a value of 100). Statistics. Groups were compared using Students two-tailed t-tests in which a P value <0.05 is considered signicant. Results are displayed in box-andwhisker plots showing sample minimum and maximum, lower, median, and upper quartile, and outliers, calculated using www.physics.csbsju.edu/stats/ (accessed October 21, 2011) and as ROC curves using www.rad.jhmi.edu/ jeng/javarad/roc/JROCFITi.html (accessed February 7, 2011). Means and SDs were calculated using Excel 2008 for Mac (Microsoft). Study Approval. For sample set 1, blood samples were collected with written, informed consent, according to National Health Service-approved protocols at the Paterson Institute for Cancer Research, University of Manchester, and were processed as described previously (22). Additional plasma samples listed in Dataset S1 were obtained from biobanks and were collected according to Institutional Review Board-approved protocols. All cases and control samples in set 2 were collected according to National Cancer Institute Early Detection Research Network standard operating procedures. Samples from patients diagnosed with lung cancer or other nonmalignant lung disorders were collected at the Thoracic Oncology Laboratories of the New York University Langone Medical Center (New York) at the time of thoracic surgery, and all diagnoses were conrmed histologically (as adenocarcinoma, inammatory lung disease, benign lung nodules, or squamous cell carcinoma) after excision of the abnormal lesion. For the smoker cohort, samples were collected from age-matched smokers at the Division of Pulmonary, and Critical Care, and Sleep Medicine, New York University School of Medicine. Study-specic approval was granted by the Department of Biology, University of York, Research Ethics Committee. ACKNOWLEDGMENTS. We thank Rose Wilson, Jian Mei Hou, Faisal Abdel Rahman, and Mark Thornquist; Ellen Eylers, Ting-An Yie, and Nihir Patel (New York University School of Medicine) for sample collection and Alissa Greenberg, MD, for clinical follow-up; Lynsey Priest for sample collection at The Christie National Health Service Foundation Trust; Said El Alaou (Covalab) for generation of variant-specic antibodies; and Yulia Maxuitenko (Southern Research Institute) for in vivo RNAi analysis. Sample collection at The Christie National Health Service Foundation Trust was supported by CHEMORES Sixth Framework Programme Contract LSHC-CT-2007-037665 (to F.H.B), and at New York University School of Medicine by National Cancer Institute Early Detection Research Network Grant U01CA086137 (to W.N.R. and H.I.P.) This work was funded by Cizzle Biotech, the University of York proof-of-concept fund, and Yorkshire Cancer Research.

7. Mitsui K, Matsumoto A, Ohtsuka S, Ohtsubo M, Yoshimura A (1999) Cloning and characterization of a novel p21(Cip1/Waf1)-interacting zinc nger protein, ciz1. Biochem Biophys Res Commun 264(2):457464. 8. den Hollander P, Rayala SK, Coverley D, Kumar R (2006) Ciz1, a Novel DNA-binding coactivator of the estrogen receptor , confers hypersensitivity to estrogen action. Cancer Res 66(22):1102111029. 9. Ainscough JF, et al. (2007) C-terminal domains deliver the DNA replication factor Ciz1 to the nuclear matrix. J Cell Sci 120(Pt 1):115124. 10. Ehmann M, et al. (2007) Identication of potential markers for the detection of pancreatic cancer through comparative serum protein expression proling. Pancreas 34(2):205214. 11. Leman ES, et al. (2008) Evaluation of colon cancer-specic antigen 2 as a potential serum marker for colorectal cancer. Clin Cancer Res 14(5):13491354. 12. Miki T, Yano S, Hanibuchi M, Sone S (2000) Bone metastasis model with multiorgan dissemination of human small-cell lung cancer (SBC-5) cells in natural killer celldepleted SCID mice. Oncol Res 12(5):209217.

Higgins et al.

PNAS Early Edition | 7 of 8

MEDICAL SCIENCES

PNAS PLUS

13. Wan KM, Nickerson JA, Krockmalnic G, Penman S (1999) The nuclear matrix prepared by amine modication. Proc Natl Acad Sci USA 96(3):933938. 14. Albrethsen J, Knol JC, Jimenez CR (2009) Unravelling the nuclear matrix proteome. J Proteomics 72(1):7181. 15. Van Le TS, Myers J, Konety BR, Barder T, Getzenberg RH (2004) Functional characterization of the bladder cancer marker, BLCA-4. Clin Cancer Res 10(4): 13841391. 16. Keesee SK, et al. (1999) Preclinical feasibility study of NMP179, a nuclear matrix protein marker for cervical dysplasia. Acta Cytol 43(6):10151022. 17. Subong EN, et al. (1999) Monoclonal antibody to prostate cancer nuclear matrix protein (PRO:4-216) recognizes nucleophosmin/B23. Prostate 39(4):298304. 18. Zink D, Fischer AH, Nickerson JA (2004) Nuclear structure in cancer cells. Nat Rev Cancer 4(9):677687. 19. Rajapakse I, Groudine M (2011) On emerging nuclear order. J Cell Biol 192(5):711721. 20. Coverley D, Laman H, Laskey RA (2002) Distinct roles for cyclins E and A during DNA replication complex assembly and activation. Nat Cell Biol 4(7):523528.

21. Munkley J, et al. (2011) Cyclin E is recruited to the nuclear matrix during differentiation, but is not recruited in cancer cells. Nucleic Acids Res 39(7): 26712677. 22. Hou JM, et al. (2009) Evaluation of circulating tumor cells and serological cell death biomarkers in small cell lung cancer patients undergoing chemotherapy. Am J Pathol 175(2):808816. 23. Rahman FA, Ainscough JF-X, Copeland N, Coverley D (2007) Cancer-associated missplicing of exon 4 inuences the subnuclear distribution of the DNA replication factor CIZ1. Hum Mutat 28(10):9931004. 24. Rahman FA, Aziz N, Coverley D (2010) Differential detection of alternatively spliced variants of Ciz1 in normal and cancer cells using a custom exon-junction microarray. BMC Cancer 10:482. 25. Xiao J, et al. (2012) Mutations in CIZ1 cause adult onset primary cervical dystonia. Ann Neurol 71(4):458469. 26. Greaves EA, Copeland NA, Coverley D, Ainscough JF (2012) Cancer-associated variant expression and interaction of CIZ1 with cyclin A1 in differentiating male germ cells. J Cell Sci 125(Pt 10):24662477.

8 of 8 | www.pnas.org/cgi/doi/10.1073/pnas.1210107109

Higgins et al.