Supporting Information

Higgins et al. 10.1073/pnas.1210107109 SI Results mL. In vivo analysis was carried out at Southern Research In- At the time of analysis more than 600 ESTs derived from normal stitute using 6-wk-old female NOD/SCID mice from the Na- and diseased tissues and cell lines mapped to the National Center tional Cancer Institute National Laboratory for Cancer for Biotechnology Information unigene cluster Hs. 212395 for Research. Forty mice were assigned randomly to three groups on human Ciz1 (www.ncbi.nlm.nih.gov/sites/entrez?db=unigene). In- day −3. On day −3, 2% sucrose was added to the drinking water dividual ESTs were translated, and the predicted amino acid se- of groups 1 and 2 (15 mice each), and 2 mg/mL doxycycline in quences were compared with full-length Ciz1 with the aim of 2% sucrose was added to the water of group 3 (10 mice). On day 7 verifying the existence of b-variant Ciz1. Ciz1 transcripts that span 0, mice were inoculated s.c. on the right flank with 1.0 × 10 exons 14 and 15 were detected in a total of 39 EST libraries, in- SBC5 ptTS-Neo b-variant shRNA 4 cells (90–95% confluence, cluding 28 that were derived from tumors. Three of five libraries 98.7% viability), in 0.1 mL medium without G418 and hygrom- derived from neuroendocrine lung cancers, 4 of 22 carcinoma li- ycin. On day 21, mice in group 2 were switched to water containing braries, 0 of 3 normal lung libraries, and 1 of 9 libraries derived sucrose and doxycycline. Tumor size was measured twice weekly from other normal tissues or cell lines were represented by b-variant and used to estimate tumor weight. transcripts. Among the lung carcinoma-derived transcripts that span the region, 4 of 10 contained exon 14b, compared with only Imaging. Fluorescence images were collected at 600× magnifi- 1 of 36 noncancer reads. These results show that Ciz1 exon 14 exists cation with a Zeiss Axiovert 200 M microscope fitted with an as two variants (Fig. S1A) and suggest that exon 14b is over- AxioCam camera (Carl Zeiss Vision) and Openlab image ac- represented in lung tumors. Prevalence of b-variant transcripts quisition software (Improvision), using a 63/1.40 oil immersion in cells was confirmed using a panel of lung-derived objective and Zeiss filter sets 10, 15, and 2 for Hoescht 33258 cell lines using PCR tools that span the unique exon-junction and (DNA), Alexa Fluor 568, and FITC/GFP, respectively. Constant support selective amplification of transcripts containing exon exposure parameters were used where images give a representa- 14b. Selectivity of junction primer P4 was established using a pair tion of the relative amount of antigen. Images were exported to of sequence-verified Ciz1 clones as pure templates (Fig. S1B). Adobe Photoshop CS2 to allow manipulation of brightness for When applied to cell line-derived cDNAs, this tool generated a reproduction purposes. Identical manipulations were applied so product from neuroendocrine lung cancers but not from a nor- that image sets reflect the relative amount of antigen present. mal fetal lung cell line. In contrast, exon 14a (anchor domain, AD) was detected in all lines, albeit at a lower level in normal cells (Fig. Additional Expression. For cloning, RNA extracted with TRIzol S1C). The majority of the analysis described in this paper exploits (Invitrogen) was reverse transcribed with M-MLV (Promega) and a junction-selective probe (T2, Table S3), rather than a primer set. a 1:1 mix of oligo(dT)/random primers. Ciz1 was amplified using Probe selectivity was established in reactions with two different primers P8/P2, products were subcloned into pGEMTeasy, and primer sets that span exons 14 and 15, using mixtures of Ciz1 DNA from random colonies was digested with EcoR1. Fragment plasmids 20 (exon 14b) and 19 (exon 14a) as template (Fig. S1D), size was visualized by electrophoresis, and insert identity was and was verified in the cellular context by application to a panel of verified by sequencing. Examples of results are shown in Fig. S1F. neuroendocrine lung cancer cell lines. B-variant Ciz1 was elevated For junction-selective PCR, Ciz1 was amplified with junction- dramatically in all small cell lung cancer (SCLC) lines as com- spanning primer P4 coupled with P3. Selectivity was verified pared with three normal lung lines and one neuroendocrine car- using plasmid templates that were isolated from SCLC cell line cinoid (Fig. 3C), and this result was supported by analysis of an SBC2 and cloned in frame into EGFPC3. Sequencing across additional 10 SCLC lines (Fig. S1E). Ciz1 replication domain exon 14 confirmed that clone 20 has exon 14b and that clone 19 (RD) and AD also were elevated in the three tumor lines, but to has exon 14a (Fig. S1B). For quantitative PCR, data were gen- a much lesser extent (Fig. 3C). Similar results also were obtained erated using an ABI 7000, SDS v1.2.3 (Applied Biosystems). usingacloningandsequencing-basedmethodforsixneuroen- Quantitative Western blot analysis was done routinely using docrine lung cancer cell lines, two nonsmall cell lung cancer cell high-throughput precast E-PAGE 8% 48-well gels, Epage sample lines, and three lines derived from normal fetal lung cells (Fig. buffer, and batch-checked, preassembled iBlot dry blotting transfer S1F). In summary, we conclude that exon 14b is overrepresented sandwiches (all Invitrogen). Although only 0.5 μL of plasma is in lung cancer cell lines, that three independent detection meth- required for quantitative detection of b-variant Ciz1, we routinely ods lead to consistent conclusions, and that junction probe T2 is make up 5 μL in a 95-μL loading buffer to minimize possible a selective tool suitable for application to sample sets from pri- inaccuracies in sample loading (10-μL volume). To minimize gel- mary tumors. to-gel variation, complete sample sets are run and transferred on the same day, and all membranes are probed and visualized SI Materials and Methods throughout the process as one, using low exposure of high-per- RNAi. Ten 21-nt siRNAs that span exon14b/exon15 were produced formance chemiluminescence film (GE Healthcare). Residual in vitro (Ambion) or were chemically synthesized (Invitrogen). variation between gels is corrected using a pair of positive and NIH 3T3 cells were transfected with siRNA and/or GFP-hCiz1 negative standards, which are included at both ends of each gel using Mirus 3T3 (Mirus Bio). For inducible delivery of shRNA, to allow assessment of the evenness of transfer. Densi- clonal populations of SBC5 cells containing ptTS-Neo and pSIREN- tometry values for variant Ciz1 typically are normalized to en- RetroQ-TetP were generated as described (Clontech). Strains dogenous Ig to control for possible inaccuracies in sample harboring chosen shRNAs were cultured using tetracycline free loading or plasma concentration, as shown in supporting Datasets FBS (Clontech) and were maintained in 100 μg/mL G418, 100 μg/ S1 and S2. This normalization impacts the final outcome, re- mL hygromycin. ShRNAs were induced by doxycycline at 2.5 μg/ ducing the spread of the data and improving ROC AUC values.

Higgins et al. www.pnas.org/cgi/content/short/1210107109 1of9 Fig. S1. Validation of b-variant–selective transcript detection tools. (A) Schematic representation of alternative splicing at Ciz1 exon 14 to yield exon 14a or exon 14b. The conditionally excluded region encoding the peptide sequence VEEELCKQ is shown as a red box. Location of primers and probes used in the analysis of Ciz1 expression are shown below, including exon 14b junction-spanning primer P4 and probe T2 (purple). Primer and probe sequences are given in Table S3. (B) Method for analysis of exon 14b expression using exon junction-spanning primer P4 coupled with reverse primer P3. Primer selectivity was verified using plasmid templates. No product is detected when clone 19 is used as template (Lower Left), whereas both plasmids are amplified efficiently by primers P1 and P2 which flank the alternatively spliced region (Lower Right). (C) RT-PCR products generated from cDNA prepared from two neuroendocrine lung cancer cell lines (L95, SBC5) and one normal fetal lung cell line (HFL1), using the b-variant primer set P4/P3 (Top). Total Ciz1 anchor domain (Ciz1 AD) detected with primer set P1/P2 (Middle) and actin detected with P11/P12 (Bottom) are shown also. Products were sequence verified. “No T” is a no-template control. (D) Alternative method for detecting the b-variant transcript using the exon junction-spanning probe T2, illustrated by amplification of mixtures of plasmid clones 19 and 20, at ratios of 100:0, 75:25, 50:50, 25:75, or 0:100. Primers spanning exon 14 (P1/P2, closed symbols, P6/P7, open symbols) that yield a product from both plasmids were coupled with a probe that spans the unique b-variant junction (T2) or which recognizes a region of the AD that is not alternatively spliced (T3 or T4, as indicated). The graph shows that the cycle number required to reach the threshold is constant for nonselective detection tools but is affected bythe composition of the plasmid mixture for the b-variant–selective probe T2. (E) Quantification of b-variant transcript in 10 SCLC cell lines as indicated, detected with primers P1/P2 coupled with probe T2. Data are normalized to actin and calibrated to the result for IMR90 cells, which is given an arbitrary value of one, and are shown with SEM. All SCLC cell lines have elevated b-variant transcript. (F) Example of mobility shift seen in PCR products from clones with exon 14a or 14b. Representative data for normal embryonic lung cell line IMR90 and SCLC line L95 are shown.

Higgins et al. www.pnas.org/cgi/content/short/1210107109 2of9 Fig. S2. Tools for detecting Ciz1 b-variant–selective protein. (A) The shorter peptide was used to immunize rabbits. Sera were depleted of antibodies reactive to the longer peptide and then were affinity purified against immunogen to generate pool 2B, which is selective for the series of amino acids that define the exon 14b/exon 15 junction in b-variant Ciz1. (B) Selectivity for b-variant Ciz1 over isoforms with the full-length exon 14 was tested by Western blot on whole- cell extracts from NIH 3T3 cells transiently transfected with full-length GFP-hCiz1 (WT) or GFP-hCiz1 b-variant (b), which encode fusion that migrate with a relative molecular mass of ∼125 kDa. Their reactivity with preimmune serum and with anti-Ciz1 polyclonal antibody 1793 is shown for comparison. (C) By immunofluorescence, pool 2B detects nuclei expressing GFP-b-variant but not GFP-Ciz1. Recombinant protein is green, primary antibody is detected with anti-rabbit Alexa Fluor 488 (red), and DNA is counterstained with Hoechst 33258 (blue). (Scale bar, 10 microns.) (D) Western blot strips showing representative plasma from a patient with SCLC and from a normal individual, probed with pool 2B after separation of proteins by 8% SDS/PAGE. Pool 2B detects a strong but diffuse band in the lung cancer plasma only. This band is smaller than the predicted mass for endogenous Ciz1 of 98kDa and appears to reflect proteolytic degradation to a stable C-terminal domain. Endogenous Ig also is detected in both samples by the secondary antibody used to reveal 2B.

Higgins et al. www.pnas.org/cgi/content/short/1210107109 3of9 Fig. S3. Western blots showing b-variant Ciz1 (upper band), and endogenous Ig (lower band) in 0.5 μL of plasma from 160 individuals in sample set 2. At the beginning and end of each gel the same positive and negative calibrator samples are included (as for set 1) to allow verification of quantitative recovery across the gel and normalization between gels. Variance between the calibrator lanes is given in Dataset S2. Band intensities were quantified using Image J (National Institutes of Health), and results for the b-variant were normalized to those for endogenous Ig. For each gel, the value for the first negative calibrator lane (at start of the gel) was subtracted from all samples to set an arbitrary zero; then the first positive calibrator lane was adjusted to 100 so that all results are expressed in relative terms. Raw values and workings are shown in Dataset S2.

Higgins et al. www.pnas.org/cgi/content/short/1210107109 4of9 Fig. S4. (A)(Left) Effect of ectopic expression of GFP-Ciz1 (WT, with exon 14a), GFP-b-variant Ciz1 (with exon 14b), or GFP alone on the normal cell line NIH 3T3. Results are expressed as the percent of green cells at the indicated hours after transfection. For all constructs the proportion of cells expressing GFP reaches ∼50% by 4.5 h (indicating ∼50% transfection efficiency) and continues to remain at a similar level in the population transfected with empty vector. For cells transfected with either WT or b-variant Ciz1 the proportion of green cells falls dramatically after ∼17 h. The black dotted line indicates the proportions that would be expected if expression of Ciz1 or b-variant Ciz1 had a cytostatic effect, with the untransfected cells in the population continuing to divide every 24 h. Results suggest that ectopic expression is eventually cytotoxic, killing transfected cells within the time frame of the experiment. Error bars show SEM for three technical replicates. (Right) Images show representative cells transfected with WT or b-variant Ciz1 (green) at the indicated times. Between 7 and 22 h the majority of green cells showed poor nuclear integrity, with lumpy chromatin and an irregular periphery. By 30 h these cells had largely disappeared from the population, leaving a few apparently healthy Ciz1-expressing cells (shown). DNA is blue. (Scale bar, 5 microns.) (B and C) At 24 h after transfection, WT or b-variant Ciz1 expression does not cause stabilization of p53 (B) or apoptotic DNA fragmentation, as revealed by TUNEL assay (Invitrogen) (C). Control images show p53 (red) in transformed mouse fibroblasts that constitutively express p53 and TUNEL (red) on cells transfected with empty GFP vector (green) and treated with DNase1 (as recommended; Epicentre). Within each experiment red image acquisition parameters are constant.

Higgins et al. www.pnas.org/cgi/content/short/1210107109 5of9 Fig. S5. Additional analyses of function and specificity. (A) Fold increase in cell number for SBC5 cells (which naturally express exon 14b) 4 d after transient transfection with the indicated WT (A, B, and C) and exon 14b-selective (14b2 and1–8) siRNAs, using Amaxa Nucelofector (80% transfection efficiency). The schematic below indicates the location of Ciz1 RNAi sequences relative to Ciz1 exons. (B) The indicated selected siRNAs were transfected into neuroendocrine carcinoid cell line H727 (which does not express exon 14b; Fig. 3C), using Amaxa Nucleofector (90% transfection efficiency), and the fold increase in cell number was calculated after 4 d. Comparisons of the data in A and B show that suppression of all forms of Ciz1 using WT siRNA restrains proliferation of both cell types, whereas exon 14b junction-selective siRNA impinges only on the growth of SBC5 cells. (C) Effect of luciferase control shRNA (luc) and b-variant junction-selective RNAi sequence 4 (14b 4) or WT Ciz1 RNAi sequence C (WT C) on SBC5 cells 4 d after induction of shRNA with doxycycline, expressed as fold increase in cell number. (D)AsinC, except that results are shown for H727 cells. RNAi sequences are given in Table S1.

Higgins et al. www.pnas.org/cgi/content/short/1210107109 6of9 Fig. S6. Selection and validation of b-variant–selective siRNA. (A) The set of siRNA sequences that span the unique junction in exon 14b transcripts. 14b1 and 14b2 were synthesized using an in vitro transcription kit (Ambion); 1–8 were chemically synthesized (Invitrogen). Shaded box shows effective and selective sequences (also boxed in C). (B) Effect of control siRNA (con) and siRNAs 1–8 on expression of endogenous Ciz1 AD transcript (detected with P1/P2), b-variant (P4/P3), and actin (P11/P12) 24 h after transient transfection into SBC5 cells. Histograms show quantification of band intensity using Image J after normalization to actin and calibration to cells transfected with control siRNA. (C) Histogram shows ratio of Ciz1 AD expression to Ciz1 b-variant expression for samples shown in B. Sequences 1–4efficiently suppress b-variant with a high degree of specificity. (D) Western blot showing effect of b-variant siRNA (14b2) on the expression of recombinant human Ciz1 protein. EGFP-Ciz1 with exon 14a (clone 19) and EGFP-Ciz1 with exon 14b (clone 20) were cotransfected into mouse NIH 3T3 cells, with junction-selective siRNA or with control siRNA as indicated. (Note that clone 19 also is alternatively spliced at exon 8 and so encodes a smaller product than 20.) Cells were harvested after 24 h, and endogenous mouse Ciz1 and recombinant human Ciz1 were detected with anti-Ciz1 antibody 1793, and actin was detected with AC-40 (Sigma-Aldrich). Results show selective suppression of b-variant Ciz1 by siRNA 14b but not by control siRNA.

Table S1. siRNA sequences Designation Functional sequence Exon

14b1 AAGAAGAGATCGAGGTGAGGT 14b 14b2 AAGAGATCGAGGTGAGGTCCA 14b siRNA 1 AGAAGAGATCGAGGTGAGGTC 14b siRNA 2 GAAGAGATCGAGGTGAGGTCC 14b siRNA 3 AGAGATCGAGGTGAGGTCCAG 14b siRNA 4 GAGATCGAGGTGAGGTCCAGA 14b siRNA 5 AGATCGAGGTGAGGTCCAGAG 14b siRNA 6 GATCGAGGTGAGGTCCAGAGA 14b siRNA 7 ATCGAGGTGAGGTCCAGAGAT 14b siRNA 8 TCGAGGTGAGGTCCAGAGATA 14b A AGAGATTGAGGAGCTCAGA 7/8 B GAGATTCAGAATGAGTCGG 9 C GTTCTATCACAGCAACTCA 16 Luciferase pSIREN-RetroQ-TetH-LUC (Clontech) — Scrambled control AAGAGACTGAGAGGAGTGATG —

Higgins et al. www.pnas.org/cgi/content/short/1210107109 7of9 Table S2. Cell lines and tissue-derived RNA samples Tissue derivation Cell line Source

SCLC L51 (HIGNT1)* ECACC SCLC L95 ECACC SCLC L88 (HIGNT1)* ECACC SCLC SBC5 JCRB SCLC SBC2 (HIGNT1)* JCRB Lung carcinoid H727 ECACC Squamous cell carcinoma VMRC-LCP JCRB Large cell carcinoma LU99 JCRB Human fetal lung IMR90 ECACC Human fetal lung WI38 ECACC Human fetal lung MRC5 JCRB Tissue RNA SCLC stage 1B — Cytomyx Visually normal adjacent tissue — Cytomyx SCLC stage 1A — Cytomyx Emphysematous adjacent tissue — Cytomyx Neuroendocrine carcinoma stage1B — Cytomyx Fibrotic adjacent tissue — Cytomyx

ECACC, European Collection of Cell Cultures; JCRB, Japanese Collection of Research Bioresources Cell Bank. *Lung cancer classification HIGNT1 is outlined in ref. 1.

1. Jones MH, et al. (2004) Two prognostically significant subtypes of high-grade lung neuroendocrine tumours independent of small-cell and large-cell neuroendocrine carcinomas identified by profiles. Lancet 363(9411):775–781.

Table S3. Primers and probes Primer or probe Sequence Exon

Primer p1 CAGGGGCATAAGGACAAAG hCiz1 Exon 13F p2 TCCGAGCCCTTCCACTCCTCTCTGG hCiz1 Exon 15R p3 TCAGGTTTTGAGGCGGGTTGAG hCiz1 Exon 17R p4 GAAGAGATCGAGGTGAGGTC hCiz1 Exon 14bF p6 CGAGGGTGATGAAGAAGAGGA hCiz1 Exon 14F p7 CCCCTGAGTTGCTGTGATA hCiz1 Exon 16R p8 CAGGGGCATAAGGACAAAG hCiz1 Exon 13F p9 CACAACTGGCCACTCCAAAT hCiz1 Exon 5F p10 CCTCTACCACCCCCAATCG hCiz1 Exon 5R p11 CAACCGCGAGAAGATGACC Human Actin F p12 TCCAGGGCGACGTAGCACA Human Actin R p13 ACACACCAGAAGACCAAGATTTACC hCiz1 Exon 6/7 junction F p14 TGCTGGAGTGCGTTTTTCCT hCiz1 Exon 7R Probe T2 TGGACCTCACCTCGATCTCTTCTTCA 14b T3 CACGGGCACCAGGAAGTCCA 16 T4 TGGTCCTCATCTTGGCCAGCA 14 T5 CGCCAGTCCTTGCTGGGACC 5 T6 CCCTGTACGCCTCTGGCCGT Actin T7 CCCTGCCCAGAGGACATCGCC 7

Dataset S1. B-variant protein levels in pretreatment plasma samples (set 1), from the indicated sources, detected with affinity-purified b-variant junction-specific antibody 2B

Dataset S1

Results are given in densitometry units after normalization to endogenous loading control shown in Fig. 1 and calibration to positive and negative standards [Patterson Institute for Cancer Research (PICR) extensive-stage lung cancer sample 15 and Innovative Research normal sample n5, respectively].

Higgins et al. www.pnas.org/cgi/content/short/1210107109 8of9 Dataset S2. B-variant protein levels in pretreatment plasma samples (set 2) detected with affinity-purified b-variant junction-specific antibody 2B

Dataset S2

Results are given in densitometry units after normalization to endogenous loading control shown in Fig. S3 and calibration to positive and negative standards (PICR extensive-stage lung cancer sample 15 and Innovative Research normal sample n5, respectively).

Dataset S3. Raw tumor weights in the mice from the study shown in Fig. 4

Dataset S3

Higgins et al. www.pnas.org/cgi/content/short/1210107109 9of9