Variant Ciz1 is a circulating biomarker for early-stagelung cancerGillian 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

aCizzle Biotech, University of York, Yorkshire YO10 5DD, United Kingdom; bDepartment of Biology, University of York, Yorkshire YO10 5YW, UnitedKingdom; cPaterson Institute for Cancer Research, University of Manchester, Lancashire M20 4BX, United Kingdom; dDivision of Pulmonary, Critical Careand 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 detectearly-stage lung cancer. Here we show that a variant form of thenuclear matrix-associated DNA replication factor Ciz1 is present in34/35 lung tumors but not in adjacent tissue, giving rise to stableprotein quantifiable by Western blot in less than a microliter ofplasma from lung cancer patients. In two independent sets, with170 and 160 samples, respectively, variant Ciz1 correctly identifiedpatients who had stage 1 lung cancer with clinically useful accuracy.For set 1, mean variant Ciz1 level in individuals without diagnosedtumors established a threshold that correctly classified 98% ofsmall cell lung cancers (SCLC) and non-SCLC patients [receiveroperator characteristic area under the curve (AUC) 0.958]. Withinset 2, comparison of patients with stage 1 non-SCLC with asymp-tomatic age-matched smokers or individuals with benign lungnodules correctly classified 95% of patients (AUCs 0.913 and 0.905),with overall specificity of 76% and 71%, respectively. Moreover,using the mean of controls in set 1, we achieved 95% sensitivityamong patients with stage 1 non-SCLC patients in set 2 with 74%specificity, demonstrating the robustness of the classification. RNAi-mediated selective depletion of variant Ciz1 is sufficient to restrainthe growth of tumor cells that express it, identifying variant Ciz1as a functionally relevant driver of cell proliferation in vitro and invivo. The data show that variant Ciz1 is a strong candidate for acancer-specific single marker capable of identifying early-stage lungcancer within at-risk groups without resort to invasive procedures.

Lung cancer is the leading cause of cancer death worldwide.Approximately 80% of lung tumors are classified as nonsmall

cell lung cancer (NSCLC), including squamous cell carcinomaand adenocarcinoma, and the remainder as small cell type(SCLC). SCLCs are primarily neuroendocrine in origin, rangingfrom low-grade typical carcinoid to high-grade neuroendocrinetumors (HGNTs), although some HGNTs are classified as largecell type (1, 2). The risk of lung cancer is increased dramaticallyby smoking, and genetic factors appear to play a role in ourability to deal with smoking-related damage. However, clearlyheritable 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 inhigh-risk groups. Although the impact of early detection on sur-vival has been questioned, several studies have looked at thepotential benefit (3), and it recently became clear that screeningwith low-dose spiral CT can achieve a significant reduction inmortality among heavy smokers (4). However, because around aquarter of individuals require follow-up procedures to investigatesuspicious imaging results, the cost of this approach is extremelyhigh, highlighting the need for a second-line noninvasive test thatcan confirm malignancy. Here we present evidence that protein-level detection of a variant form of the nuclear matrix proteinCiz1 has the potential to meet this need. Ciz1 promotes initiationof mammalian DNA replication, where it helps coordinate thesequential functions of cyclin E- and A-dependent proteinkinases (5). It interacts directly with cyclins E and A (6), with

CDK2, and with the cyclin-dependent kinase inhibitor p21 (7)and also plays an indirect role in DNA replication by modulatingthe expression of genes, including cyclin D, that influence cellproliferation (8). Normally, Ciz1 is attached to the salt- andnuclease-resistant protein component of the nucleus referredto as the “nuclear matrix” and resides within foci that partiallycolocalize with sites of DNA replication (9), implicating Ciz1 inthe spatial organization of DNA replication. Here we describe aCiz1 variant that lacks part of a C-terminal domain involved innuclear matrix attachment (Fig. 1A). Expression of this stablevariant is apparently restricted to tumor cells, making clinicalexploitation as a cancer biomarker highly tractable.

ResultsAs part of a gene-focused analysis of function, we cloned humanCiz1 from a SCLC cell line and recovered multiple variants, in-cluding a prevalent transcript in which 24 nucleotides from the 3′end of exon 14 (2475_2498del) is excluded, leading to in-framedeletion of eight amino acids (VEEELCKQ). Analysis of thesequence surrounding exons 14 and 15 revealed a second splicedonor site within exon 14 (2475/6) that could support alternativesplicing. Location identifiers refer to Ciz1 reference sequenceNM_012127.2. We refer to the whole of predicted exon 14 as“14a,” the shorter alternative as “14b,” and Ciz1 transcriptsharboring 14b as “b-variant.” Transcript frequencies among ESTsthat map to the Ciz1 Unigene cluster Hs. 212395 suggested thatb-variant is prevalent in neuroendocrine lung tumors, and thisprevalence was confirmed by analysis of SCLC cell lines usingindependent detection methods (Fig. S1 and below).As a nuclear matrix protein characterized by resistance to

harsh extraction conditions, b-variant could offer a robust bio-marker with potential to remain stable and detectable in bodyfluids. Consistent with this idea, an affinity-purified polyclonalantibody directed against the unique peptide encoded at thejunction of exon 14b/exon15 (Fig. S2) detected b-variant proteinby Western blot in 1 μL of plasma from patients with SCLC andNSCLC but not from healthy individuals (Fig. 1B). A diffuse butspecific band of 50–60 kDa (Fig. S2) is reproducibly seen andremained 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. andD.C. analyzed data; and J.F.X.A. and D.C. wrote the paper.

Conflict of interest statement: The findings reported in this paper arise directly out ofbasic research at the University of York (York, Yorkshire, United Kingdom) on the Ciz1protein and its role in the spatial and temporal organization of DNA replication. Thefindings are the result of an academic and commercial collaboration, funded in part byCizzle Biotech, which is a spin-out company of the University of York. G.H. and D.C. arepartially funded by Cizzle Biotech. D.C. and J.F.X.A hold shares in Cizzle Biotech.

This article is a PNAS Direct Submission.1To 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

MED

ICALSC

IENCE

SPN

ASPL

US

Quantification of b-variant protein in 119 SCLC and NSCLCsamples (summary information in Table 1 with sample-specificinformation in Dataset S1) gave a mean signal intensity of 95.8densitometry units with an SD of 51.8, indicating considerableheterogeneity among patients. In contrast, the mean signal in-tensity for 51 individuals without diagnosed malignancy (who hadchronic obstructive pulmonary disease, asthma, anemia, or nodisease) was just 14.1 (SD 15.2). Unexpectedly, limited- andextensive-stage SCLC and stages 1, 2, 3, and 4 NSCLC were allsignificantly different from the control groups when analyzedseparately (Fig. 1C), with a trend toward expression increasingwith stage for stages 1, 2, and 3 NSCLC. Receiver operating char-acteristic (ROC) analysis for all 170 samples returned an areaunder the curve (AUC) of 0.958 (Fig. 1D), demonstrating con-siderable discriminatory power in this context. To our knowledge,and despite considerable effort and progress with circulatingbiomarkers for other cancers (10, 11), no other single biomarkeris capable of achieving this level of discrimination between

patients with early-stage lung cancer and those without lungcancer. However, for maximum clinical utility a blood test mustbe able to identify individuals with early-stage lung cancer withinpopulations 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 identified by CT. A test that canpositively identify those individuals with lung cancer could helpreduce the frequency of surgical intervention and favorably affectboth cost and outcome. To look specifically at the discriminatorypower of b-variant Ciz1 in comparison with (i) age-matchedsmokers with ≥10 y of cigarette smoking history but withoutdiagnosed cancer and (ii) individuals with nonmalignant lungnodules or inflammatory lung disease, we analyzed a second,independent, archived set of case and control plasma samples(set 2, summary information in Table 1 with sample-specific in-formation in Dataset S2), using high-throughput Western blotsand just 0.5 μL of each sample. Comparison of the smokers(median age 64.5 y) with 20 individuals with stage 1 squamous

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 exon14/15 junction which gives rise to b-variant Ciz1. Exons that encode DNA replication domain (5) and nuclear matrix anchor domain (9) are indicated by dottedlines above. Exons that are commonly excluded from known Ciz1 variants (23) are shaded in dark gray. A complete representation of Ciz1 alternative splicevariants was assembled previously (24), and transcript diversity was discussed recently (25, 26). (B) Western blot showing b-variant protein detected withantibody 2B in 1 μL of plasma from patients with SCLC or NSCLC plus five samples from individuals with no diagnosed disease. Endogenous Ig is used tonormalize for loading (control). (C) Box-and-whisker plots showing the median, upper, and lower quartile, range, and outliers for data derived from Westernblots by densitometry. Results for a total of 119 pretreatment lung cancer patients (Left) with the indicated type and stage of disease (Right), plus 51 samplesfrom individuals with no disease or from patients with COPD, asthma, or anemia are shown. Individual sample values are given in Dataset S1. Usinga threshold set at the mean of the noncancer samples, the test correctly classified 98% of all 119 lung cancer patients, with specificity of 85%. (D) ROC curve,with 95% confidence interval, generated for all 170 samples (AUC is 0.958). Student’s two-tailed t test with unequal variance returned a P value of <0.0001 forthe noncancer samples compared with all sample sets from individuals with lung cancer.

2 of 8 | www.pnas.org/cgi/doi/10.1073/pnas.1210107109 Higgins et al.

cell carcinoma (median age 67 y) and 20 patients with stage1 adenocarcinoma (median age 71.5 y) revealed clear discrimi-natory power in both contexts and ROC AUC values in excess of0.9 (Fig. 2). Western blot images for all samples are shown inFig. S3, and raw and normalized results are given in Dataset S2.These data show that b-variant Ciz1 has significant potential fordetecting 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 indi-viduals with benign lung nodules or inflammatory diseases of thelung (set 2), b-variant Ciz1 correctly classified 95% of patientswith stage 1 lung cancer, with an overall specificity of 71% (set 2).Moreover, when the threshold is set at the mean of all of thenoncancers in the discovery set (set 1), 95% of the patients withstage 1 lung cancer in the validation set (set 2) are correctly clas-sified, with an overall specificity of 75%. Under both circum-stances the false-positive rate was ∼50%, suggesting that, if it

were applied as a secondary screen to high-risk groups withsuspicious CT results (4), b-variant Ciz1 could halve the numberof 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% ofpatients with stage 1 lung cancer (in set 2) falling above the meanof the smokers and only 34% of smokers being placed in thesuspicious category. Thus, positioning the test either before orafter CT could offer significant advantages over the use ofCT alone.Consistent with the presence of b-variant Ciz1 in patient plasma,

immunohistochemical analysis of primary tumors from patientswith neuroendocrine lung cancer revealed b-variant–positivecells in 34 of 35 patients (Fig. 3A). Staining was heterogeneous inboth distribution and level, with positive cells evident throughoutthe tumor in some individuals and limited to isolated cells inothers. These results confirm that b-variant Ciz1 is a tumor antigen

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

Diagnosis/classification No. of patients Median age Sex (%F) Median pack years* Mean b-variant (±SD)

Set 1 170Stage 1 NSCLC 52 62 40 N/d 65.7 ± 40.0Stage 2 NSCLC 6 66.5 50 N/d 94.8 ± 29.8Stage 3 NSCLC 16 66 38 N/d 103.6 ±− 26.3Stage 4 NSCLC 14 69 43 N/d 85.0 ± 52.1Limited stage SCLC 21 67 48 N/d 112.0 ± 67.1Extensive stage SCLC 10 64 70 N/d 118.0 ± 70.1All cancer 119 65 44 N/d 95.8 ± 51.8COPD 5 59 20 N/d −0.3 ± 4.3Asthma 12 42.5 58 N/d 17.6 ± 16.9Anemia 20 47.5 80 N/d 23.7 ± 8.0No diagnosed disease 14 36 29 N/d 5.9 ± 12.3All noncancer 51 46 55 N/d 14.1 ±15.2

Set 2 160Stage 1 squamous cell carcinoma 20 67 20 45 64.3 ± 32.0Stage 1 adenocarcinoma 20 71.5 60 41 35.2 ± 14.4Benign lung nodules 20 60 85 20 8.6 ± 22.0Inflammatory lung disease 20 67.5 60 30 6.3 ± 0.4Smokers 80 64.5 55 34 6.35 ± 12.2

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

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

Thresholds used False negatives Sensitivity (%) False positivesFalse positive

(%)Total in wrong

groupSpecificity

(%)

Set 1Mean of all noncancers

in set 114.07 3/119 (all cancer) 98 23/51 45 26/170 85

Mean of all noncancersin set 1

14.07 0/52 (stage 1 NSCLC) 100 23/51 45 23/103 78

Set 2Mean of all noncancers

in set 26.27 2/40 95 48/120 40 50/160 69

Mean of all smokersin set 2

6.35 2/40 95 27/80 34 29/120 76

Mean of benign andinflammatory in set 2

7.45 2/40 95 21/40 53 23/80 71

Mean of all noncancersin set 1

14.07 2/40 (stage 1) 95 40/120(all non cancer)

33 42/160 74

Mean of all noncancersin set 1

14.07 2/40 (stage 1) 95 23/80 (smokers) 29 25/120 79

Mean of all noncancersin set 1

14.07 2/40 (stage 1) 95 18/40 (benign andinflammatory)

45 20/80 75

Higgins et al. PNAS Early Edition | 3 of 8

MED

ICALSC

IENCE

SPN

ASPL

US

with the potential for exploitation as a lung cancer biomarker.Complementary transcript analysis reported at least 40-fold el-evation in bulk tumor RNA from neuroendocrine tumors com-pared with matched adjacent tissue for all three patients tested(Fig. 3B). In addition, b-variant transcript was elevated in 13 SCLCcell lines compared with control lines (Fig. 3C and Fig. S1E),

confirming 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 immunofluorescence revealed that the b-

variant protein resides as subnuclear foci that are somewhatlarger than WT Ciz1 in the nucleus in SCLC cell line SBC5 but

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 40patients with stage 1 NSCLC with similar smoking history (Left) and for 20 individuals diagnosed with stage 1 adenocarcinoma, inflammatory lung disease(granuloma), benign lung nodules (carcinoid, hamartoma), or stage 1 squamous cell carcinoma (Right), showing lower, median, and upper quartiles andoutliers (circles). (B) ROC curve with 95% confidence intervals for the indicated comparisons. AUCs are 0.913 when samples from 80 smokers are comparedwith samples from 40 patients with stage1 lung cancer, 0.905 when samples from 40 patients with benign nodules or inflammatory disease are compared withsamples from 40 patients with stage1 lung cancer, and 0.909 when all samples from smokers and patients with benign nodules or inflammatory disease arecompared with all samples from patients with stage1 lung cancer, but are only 0.503 when samples from smokers are compared with samples from patientswith benign nodules or inflammatory disease. Western blots showing b-variant protein in the 0.5 μL of plasma used for each of the 160 samples are shown inFig. S3; individual b-variant levels and clinical parameters are given in Dataset S2. Student’s two-tailed t test with unequal variance returned a P value of <0.0001for the samples from patients with cancer compared with samples either from smokers or from patients with benign nodules or inflammatory disease.

4 of 8 | www.pnas.org/cgi/doi/10.1073/pnas.1210107109 Higgins et al.

not in the normal lung cell line Wi38 (Fig. 3D) (9). Whentransfected into NIH 3T3 cells, GFP-tagged Ciz1 constructs withexon 14a or 14b produced protein that was exclusively nuclearand punctate and which resisted extraction (Fig. 3E). Thus, wesaw no gross differences in subcellular localization or attachmentto nonchromatin nuclear structures. However subtle differencesin subnuclear distribution were evident, with the b-variantforming larger aggregates and localizing to the perimeter of thenucleus as well as to foci. Therefore, the loss of eight amino acidsfrom the anchor domain (AD) of Ciz1 does subtly affect the spatialorganization of Ciz1, leading us to suggest that it may influencethe functional compartment into which the DNA replicationactivity of Ciz1 is targeted and possibly, therefore, the executionof replication-coupled events. Notably, normal cells do not tol-erate 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 overexpressionultimately results in the loss of nuclear integrity in NIH 3T3 cellsand loss of adherence within 33 h of transfection via a mecha-nism that is not apoptosis (Fig. S4).The increasingly well-characterized role of Ciz1 in DNA rep-

lication makes it a candidate therapeutic target with the poten-tial to modulate cell proliferation. Inhibition of Ciz1 expressionusing siRNA restrains cell proliferation in murine NIH 3T3 cells

(5) and human SBC5 cells (Fig. S5). Therefore, if tumor cells thatexpress the b-variant rely on it to support S phase, its inhibitioncould selectively restrain their proliferation. From a panel ofjunction-spanning siRNAs that suppress the expression of theb-variant transcript and protein but not Ciz1 isoforms that con-tain exon 14a (Fig. S6 and Table S1), we selected the most potentsequence (Fig. S5A) and tested its effect on the proliferation ofSBC5 SCLC cells (which express endogenous b-variant; Fig. 3 Cand D) and also H727 cells (which do not express endogenousb-variant; Fig. 3C and Fig. S5B) when delivered from an inducibleshRNA vector. Like RNAi targeted against a constant region ofCiz1, b-variant shRNA effectively restrained cell proliferation inculture (Fig. 4A). As expected, b-variant transcript, but not otherforms of Ciz1, was suppressed selectively (Fig. 4B), and thissuppression had a significant impact on b-variant protein (Fig.4C). Importantly, in a xenograft model of human SCLC in whichs.c. tumors were derived from SBC5 cells (12), b-variant shRNAdramatically restrained tumor growth. Similar results were obtainedwhen expression was induced at the time of tumor cell in-oculation or after tumors had formed (Fig. 4D). Together, thesein vitro and in vivo experiments show that the unique junctiongenerated by alternative splicing of Ciz1 exon 14 can be exploitedto suppress variant expression specifically and that this suppres-sion is sufficient to inhibit tumor cells that express it, identifying

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 (darkbrown), 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 oftumor sections did not contain b-variant–positive cells. (B) Relative quantification (RQ) of b-variant transcript (primers P1/P2, probe T2) and Ciz1 AD transcriptin SCLC tumors of the indicated disease stages plus adjacent tissue from three patients. Results are normalized to actin and calibrated to the first adjacentsample. (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 technicalreplicates 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. DNAis 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 fociare larger for the b-variant and are accompanied by localization around the periphery (white arrowhead).

Higgins et al. PNAS Early Edition | 5 of 8

MED

ICALSC

IENCE

SPN

ASPL

US

b-variant Ciz1 as a functionally relevant driver of tumor cellproliferation.

DiscussionThe evidence presented here shows that a cancer-associated var-iant isoform of the nuclear matrix protein Ciz1 holds consider-able promise as the basis of a blood test for early-stage lungcancer. As biomarkers, nuclear matrix proteins offer severaladvantages. They are at the heart of a dysregulated nucleus andare likely to play causative roles in epigenetic control of geneexpression. Perhaps more importantly, nuclear matrix proteinsare inherently stable, offering practical advantages that makeexploitation feasible. As a group they are biochemically definedas the fraction that remains after extraction to remove lipids,soluble proteins, cytoskeleton, chromatin, and associated proteins,which when observed microscopically forms a proteinaceous net-work (13) made up of constitutive and cell-type–specific com-ponents (14). Increasingly nuclear matrix proteins such as NMP22,BCLA4, PC1, and NM179 are gaining credibility as tumor markersfor bladder (15), cervix (16), and prostate cancer (17), respectively.

However, despite their promise, the need remains for markerswith tumor-specific profiles that can be used to complementexamination of the structure of nuclei. B-variant Ciz1 is one suchmolecule, robustly quantifiable in the blood of patients who havelung cancer, that is both a driver of DNA replication and tumorcell proliferation and a stable derivative of circulating tumorcells. Notably, this biomarker was identified through intensivegene-focused analysis of expression and function rather than bymutation- or gene-expression–profiling approaches.Analysis by high-throughput Western blot offers a direct, visual

output clarified by sample denaturation that eliminates epitopemasking in native samples. Furthermore, fractionation on thebasis of molecular weight effectively isolates signal from that ofpotentially reactive contaminants, giving a clear picture of thepotential of the biomarker. Western blot analysis, however, maynot report the full dynamic range of biomarker concentrationsand is not easily applied outside a research laboratory. There-fore, routine application of a b-variant–based test in a clinicalcontext will require the development of a more streamlinedmethod, such as ELISA, with a simplified quantitative output.

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-variantjunction-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 SBC5b-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 isblue. (Scale bar, 10 microns.) (D) Xenograft tumor formation in mice. Control mice were inoculated with SBC5 b-variant shRNA 4 cells and maintained in theabsence of doxycycline (open circles, group 1). Additional cohorts (filled 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 withlow variation. Graphs show mean tumor weight from 10–15 animals, with SEM. (Data are given in Dataset S3.)

6 of 8 | www.pnas.org/cgi/doi/10.1073/pnas.1210107109 Higgins et al.

Thus, we acknowledge that, although the data presented showthat the biomarker holds considerable discriminatory power, theassay ultimately will need to switch platforms and be validatedfurther in that context.Alteration in the structure of nuclei, detected by light micros-

copy of tumor biopsies, has remained the most reliable methodof cancer diagnosis for more than 100 y (18), but the underlyingabnormalities have been difficult to define at the molecular level.Detailed and focused analysis of the relationship between nu-clear architecture and essential nuclear events is beginning toidentify new players involved in establishing order (19). More-over, disease-associated molecular changes in these players arenow becoming apparent, offering exciting opportunities for thedevelopment of new clinically relevant tools.

Materials and MethodsAdditional methods are described in SI Materials and Methods.

Cells. NIH 3T3 cells were grown as described (20). SBC5 cells, obtained fromthe Japanese Collection of Research Bioresources Cell Bank (JCRB) weregrown in Eagle’s minimal essential medium (Sigma), 1% penicillin, strepto-mycin, glutamine (Gibco), and 10% (vol/vol) FBS (Biosera). RNAi with siRNAand inducible shRNA is described in SI Materials and Methods. All other cellswere from the European Collection of Cell Cultures or JCRB and were cul-tured as recommended.

Quantitative RT-PCR. RNA from cell lines (Table S2) was isolated with TRIzol,and cDNA was synthesized using SuperScript III (Invitrogen). RNA fromtumors and adjacent tissue (Cytomyx) was reverse transcribed using M-MLV(Promega) and random primers (Promega) plus oligo-dT primer 12–18(Invitrogen). Primers (Table S3) P6/P7 or P1/P2 (Sigma Aldrich) were coupledwith junction probe T2 (MWG) for b-variant Ciz1 and were verified usingplasmid templates (Fig. S1). Relative expression was calculated using thecomparative cycle threshold (Ct) method (2−ΔΔCt), and results were expressedrelative to normal cells or tissue.

Antibodies. B-variant–specific rabbit polyclonal antibodies (2B) were gener-ated by Covalab using the splice-junction sequence (C-DEEEIEVRSRDIS) (Fig.S2A). Sera first were subtracted for antibodies reactive against flankingsequences in a negative control peptide that includes the amino acids absentfrom b-variant (DEDEEEIEVEEELCKQVRSRDISR) and then were adsorbed topeptide containing the b-variant junction to generate affinity-purified anti-body 2B. Purified antibodies were characterized in a range of assays includingimmunofluorescence and Western blot on cells expressing WT or b-variantforms of human GFP-Ciz1 (Fig. S2 B and C) and lung cancer cells expressingendogenous b-variant Ciz1 with and without specific suppression by b-variantshRNA (Figs. 3D and 4C).

Immunodetection and Analysis. For immunofluorescence, the b-variant wasdetected after paraformaldehyde fixation using affinity-purified polyclonalantibody 2B (Fig. S2) and then anti-rabbit Alexa Fluor 488 (Invitrogen, FolioBioscience). Subcellular fractionation was carried out as described (21). Forimmunohistochemistry, 2B was applied to SCLC tissue microarray ARY-HH0117 (Folio Biosciences) and visualized using Ultraview Universal DAB(Ventana) with hematoxylin counterstain. Duplicate cores representing 35SCLC (grade 4) and tissue from five normal or cancer-adjacent lungs wereanalyzed by a certified pathologist at Phylogeny, Inc. Cores were classified asPP (containing mostly b-variant–positive cancer cells), P (containing b-var-

iant–positive cancer cells with variable intensity and number, exceeding10% of cells), PS (containing positive stromal cells), PPS (containing stronglypositive stromal cells), or N (containing no positive cells). Images shown inFig. 3A are derived from (i) core H8, (ii) G7. (iii) D10, (iv) G6, and (v) F8.

Plasma Analysis. For sample set 1, 1 μL of plasma was resolved by 8% SDS/PAGE through 12 cm × 12 cm maxi-gels (Atto) and was transferred to ni-trocellulose by semidry blotting. For samples in set 2, 0.5 μL of plasma wasdenatured using Epage buffer, resolved by electrophoresis through precastE-PAGE 8% 48-well gels, and was transferred to nitrocellulose using iBlot dryblotting system (all Invitrogen). A more detailed description is given in SIMaterials and Methods. Membranes were probed with anti-Ciz1 b-variant–specific affinity purified polyclonal antibody 2B, diluted 1/500 in PBS, 10%(wt/vol) milk powder, and 0.1% Tween 20. Antibody 2B was detected withperoxidase-IgG light chain-specific, highly cross-adsorbed monoclonal anti-rabbit antibody (211-032-171; Jackson Immunological Research). Western blotswere quantified using Image J (National Institutes of Health), and sampleswere assigned a value after normalization to loading control and calibration toa constant normal sample (given a value of 0), and to an NSCLC sample (givena value of 100).

Statistics. Groups were compared using Student’s two-tailed t-tests in whicha P value <0.05 is considered significant. Results are displayed in box-and-whisker plots showing sample minimum and maximum, lower, median, andupper 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 SDswere 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 protocolsat the Paterson Institute for Cancer Research, University of Manchester, andwere processed as described previously (22). Additional plasma samples listedin Dataset S1 were obtained from biobanks and were collected according toInstitutional Review Board-approved protocols. All cases and control samplesin set 2 were collected according to National Cancer Institute Early DetectionResearch Network standard operating procedures. Samples from patientsdiagnosed with lung cancer or other nonmalignant lung disorders werecollected at the Thoracic Oncology Laboratories of the New York UniversityLangone Medical Center (New York) at the time of thoracic surgery, and alldiagnoses were confirmed histologically (as adenocarcinoma, inflammatorylung disease, benign lung nodules, or squamous cell carcinoma) after exci-sion of the abnormal lesion. For the smoker cohort, samples were collectedfrom age-matched smokers at the Division of Pulmonary, and Critical Care,and Sleep Medicine, New York University School of Medicine. Study-specificapproval was granted by the Department of Biology, University of York,Research Ethics Committee.

ACKNOWLEDGMENTS. We thank Rose Wilson, Jian Mei Hou, Faisal AbdelRahman, andMark Thornquist; Ellen Eylers, Ting-An Yie, and Nihir Patel (NewYork University School of Medicine) for sample collection and Alissa Green-berg, MD, for clinical follow-up; Lynsey Priest for sample collection at TheChristie National Health Service Foundation Trust; Said El Alaou (Covalab) forgeneration of variant-specific antibodies; and Yulia Maxuitenko (SouthernResearch Institute) for in vivo RNAi analysis. Sample collection at The ChristieNational Health Service Foundation Trust was supported by CHEMORES SixthFramework Programme Contract LSHC-CT-2007-037665 (to F.H.B), and at NewYork University School of Medicine by National Cancer Institute Early De-tection Research Network Grant U01CA086137 (to W.N.R. and H.I.P.) This workwas funded by Cizzle Biotech, the University of York proof-of-concept fund,and Yorkshire Cancer Research.

1. Brambilla E, Travis WD, Colby TV, Corrin B, Shimosato Y (2001) The new World HealthOrganization classification of lung tumours. Eur Respir J 18(6):1059–1068.

2. Jones MH, et al. (2004) Two prognostically significant subtypes of high-grade lungneuroendocrine tumours independent of small-cell and large-cell neuroendocrinecarcinomas identified by gene expression profiles. Lancet 363(9411):775–781.

3. Baldwin DR (2011) Imaging in lung cancer: Recent advances in PET-CT and screening.Thorax 66(4):275–277.

4. Aberle DR, et al.National Lung Screening Trial Research Team (2011) Reduced lung-cancer mortality with low-dose computed tomographic screening. N Engl J Med 365(5):395–409.

5. Coverley D, Marr J, Ainscough JF-X (2005) Ciz1 promotes mammalian DNA replication.J Cell Sci 118(Pt 1):101–112.

6. Copeland NA, Sercombe HE, Ainscough JF, Coverley D (2010) Ciz1 cooperates withcyclin-A-CDK2 to activate mammalian DNA replication in vitro. J Cell Sci 123(Pt 7):1108–1115.

7. Mitsui K, Matsumoto A, Ohtsuka S, Ohtsubo M, Yoshimura A (1999) Cloning andcharacterization of a novel p21(Cip1/Waf1)-interacting zinc finger protein, ciz1.Biochem Biophys Res Commun 264(2):457–464.

8. den Hollander P, Rayala SK, Coverley D, Kumar R (2006) Ciz1, a Novel DNA-bindingcoactivator of the estrogen receptor α, confers hypersensitivity to estrogen action.Cancer Res 66(22):11021–11029.

9. Ainscough JF, et al. (2007) C-terminal domains deliver the DNA replication factor Ciz1to the nuclear matrix. J Cell Sci 120(Pt 1):115–124.

10. EhmannM, et al. (2007) Identification of potential markers for the detection of pancreaticcancer through comparative serum protein expression profiling. Pancreas 34(2):205–214.

11. Leman ES, et al. (2008) Evaluation of colon cancer-specific antigen 2 as a potentialserum marker for colorectal cancer. Clin Cancer Res 14(5):1349–1354.

12. Miki T, Yano S, Hanibuchi M, Sone S (2000) Bone metastasis model with multiorgandissemination of human small-cell lung cancer (SBC-5) cells in natural killer cell-depleted SCID mice. Oncol Res 12(5):209–217.

Higgins et al. PNAS Early Edition | 7 of 8

MED

ICALSC

IENCE

SPN

ASPL

US

13. Wan KM, Nickerson JA, Krockmalnic G, Penman S (1999) The nuclear matrix prepared

by amine modification. Proc Natl Acad Sci USA 96(3):933–938.14. Albrethsen J, Knol JC, Jimenez CR (2009) Unravelling the nuclear matrix proteome. J

Proteomics 72(1):71–81.15. 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):

1384–1391.16. Keesee SK, et al. (1999) Preclinical feasibility study of NMP179, a nuclear matrix

protein marker for cervical dysplasia. Acta Cytol 43(6):1015–1022.17. Subong EN, et al. (1999) Monoclonal antibody to prostate cancer nuclear matrix

protein (PRO:4-216) recognizes nucleophosmin/B23. Prostate 39(4):298–304.18. Zink D, Fischer AH, Nickerson JA (2004) Nuclear structure in cancer cells. Nat Rev

Cancer 4(9):677–687.19. Rajapakse I, Groudine M (2011) On emerging nuclear order. J Cell Biol 192(5):711–721.20. 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):523–528.

21. Munkley J, et al. (2011) Cyclin E is recruited to the nuclear matrix duringdifferentiation, but is not recruited in cancer cells. Nucleic Acids Res 39(7):2671–2677.

22. Hou JM, et al. (2009) Evaluation of circulating tumor cells and serological cell deathbiomarkers in small cell lung cancer patients undergoing chemotherapy. Am J Pathol175(2):808–816.

23. Rahman FA, Ainscough JF-X, Copeland N, Coverley D (2007) Cancer-associatedmissplicing of exon 4 influences the subnuclear distribution of the DNA replicationfactor CIZ1. Hum Mutat 28(10):993–1004.

24. Rahman FA, Aziz N, Coverley D (2010) Differential detection of alternatively splicedvariants 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. AnnNeurol 71(4):458–469.

26. Greaves EA, Copeland NA, Coverley D, Ainscough JF (2012) Cancer-associated variantexpression and interaction of CIZ1 with cyclin A1 in differentiating male germ cells.J Cell Sci 125(Pt 10):2466–2477.

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