当前位置:首页 >> 临床医学 >>

Tumorigenicity and genetic profiling of circulating tumor


articles

Tumorigenicity and genetic profiling of circulating tumor cells in small-cell lung cancer
Cassandra L Hodgkinson1,7, Christopher J Morrow1,7, Yaoyong Li2, Robert L Metcalf1, Dominic G Rothwell1, Francesca Trapani1, Radoslaw Polanski1, Deborah J Burt1, Kathryn L Simpson1, Karen Morris1, Stuart D Pepper3, Daisuke Nonaka4, Alastair Greystoke1,4,5, Paul Kelly1, Becky Bola1, Matthew G Krebs1, Jenny Antonello1, Mahmood Ayub1, Suzanne Faulkner1, Lynsey Priest1, Louise Carter1, Catriona Tate1, Crispin J Miller2,6, Fiona Blackhall4,5,8, Ged Brady1,8 & Caroline Dive1,8
? 2014 Nature America, Inc. All rights reserved.

Small-cell lung cancer (SCLC), an aggressive neuroendocrine tumor with early dissemination and dismal prognosis, accounts for 15–20% of lung cancer cases and ~200,000 deaths each year. Most cases are inoperable, and biopsies to investigate SCLC biology are rarely obtainable. Circulating tumor cells (CTCs), which are prevalent in SCLC, present a readily accessible ‘liquid biopsy’. Here we show that CTCs from patients with either chemosensitive or chemorefractory SCLC are tumorigenic in immune-compromised mice, and the resultant CTC-derived explants (CDXs) mirror the donor patient’s response to platinum and etoposide chemotherapy. Genomic analysis of isolated CTCs revealed considerable similarity to the corresponding CDX. Most marked differences were observed between CDXs from patients with different clinical outcomes. These data demonstrate that CTC molecular analysis via serial blood sampling could facilitate delivery of personalized medicine for SCLC. CDXs are readily passaged, and these unique mouse models provide tractable systems for therapy testing and understanding drug resistance mechanisms.
Improved treatment outcomes for patients with SCLC require new approaches to interrogate the biology and genetics of this disease, appropriate methods to investigate resistance to current chemotherapy and tractable, patient-derived, clinically relevant models to test new therapies. Moreover, minimally invasive monitoring of patients with SCLC is needed to optimize therapy selection. We sought to address these issues by developing unique patient-derived mouse models exploiting the abundant CTCs in patients with SCLC and testing their response to standard platinum and etoposide chemotherapy. In parallel, we validated CTC profiling for patient monitoring. The high response rates, including complete responses, to platinumbased chemotherapy regimens for SCLC in the 1970s and 1980s1–3 led to the belief that cures might soon follow. Four decades later, 5-year survival rates for patients with SCLC remain at 5% owing to inherent or, most commonly, acquired treatment resistance. SCLC cell lines were amongst the first cancer cell lines developed and used for drug testing4,5, and the NCI-H209 line was amongst the first cancer cell lines to be deep sequenced6, accelerating knowledge of SCLC biology. However, although many hypotheses were generated using cell lines, they were not upheld in the clinic7. Trials of targeted therapies in SCLC have proved universally disappointing, with no major advances
1Clinical

since the advent of cisplatin and etoposide treatment8. The frequent, rapid and marked biological transition from chemotherapy-sensitive to chemotherapy-resistant disease suggests that much is unknown regarding drivers of acquired chemotherapy resistance in SCLC. A genetically engineered mouse model of SCLC developed via conditional deletion of the tumor suppressor genes Trp53 and Rb1 allowed new insights into SCLC progression9, but it has not proven amenable to pharmacology-based studies10. A major barrier to comprehensive understanding of human SCLC biology and discovery of ‘druggable’ targets is that access to fresh, sufficient tumor tissue for research is rare. This most aggressive neuroendocrine tumor has a short doubling time and high growth fraction and disseminates early such that surgery is rarely performed. Explant models derived from patients with SCLC (patient derived xenografts, PDXs) exist11,12, but the poor take rate and low frequency with which tumor biopsies are obtained, along with their typically small size and high necrosis content13, make this approach challenging. In the absence of sequential biopsies, the molecular basis of acquired drug resistance in SCLC has yet to be interrogated comprehensively. Comparison of SCLC PDXs and derived cell lines indicates that swift and irreversible changes in gene expression occur in the latter11. Next-generation

npg

and Experimental Pharmacology Group, Cancer Research UK Manchester Institute, University of Manchester, Manchester, UK. 2Computational Biology Support Group, Cancer Research UK Manchester Institute, University of Manchester, Manchester, UK. 3Molecular Biology Core Facility, Cancer Research UK Manchester Institute, University of Manchester, Manchester, UK. 4The Christie NHS Foundation Trust, Manchester, UK. 5Institute of Cancer Sciences, University of Manchester, Manchester, UK. 6RNA Biology Group, Cancer Research UK Manchester Institute, University of Manchester, Manchester, UK. 7These authors contributed equally to this work. 8These authors jointly supervised this work. Correspondence should be addressed to C.D. (caroline.dive@cruk.manchester.ac.uk). Received 17 March; accepted 16 May; published online 1 June 2014; doi:10.1038/nm.3600

nature medicine advance online publication

1

articles
Table 1 Clinical characteristics of donors with SCLC and subsequent generation of CTC tumors in immunocompromised mice
Patient 1 2 3 4 5 6
aCTC

CTC count per 7.5 mla 458 1,625 507 1,376 222 20

Metastatic sites Bone, lung, lymph node Bone, brain, meningeal Lymph node Liver, lung, lymph node Liver, lymph node, pancreas Lymph node, pleura

Chemosensitive/refractory Sensitive Refractory Sensitive Refractory Refractory Sensitive

Patient survival (months)b 7.3 3.5 9.7 0.9 1.7 13.4c

Time from CTC implantation to palpable tumor (months) 4.4 2.4 4.3 3.2 No tumor at 13.7 (mouse culled) No tumor at 12.3 (mouse culled)

count performed on CellSearch platform. bFrom date of CTC sample blood draw. cPatient alive at last follow-up.

? 2014 Nature America, Inc. All rights reserved.

sequencing (NGS) technology recently expanded the genomic landscape of SCLCs beyond the nearly ubiquitous inactivation of the tumor suppressor genes TP53 and RB1, revealing high mutation rates and frequent alterations in regulators of histone modification that underpin genomic instability and tumor heterogeneity6,14,15. With a new approach, we sought to develop patient-derived in vivo models of SCLC that would inform our understanding of metastatic disease and that would be sufficiently tractable for therapy testing. We demonstrated CTCs (expressing epithelial cell adhesion molecule (EpCAM) and cytokeratins) detected using the CellSearch platform were highly prevalent in patients with SCLC compared to other cancers and that CTC number was of prognostic significance16–21. We reasoned that tumor-initiating cells must be present within the CTC population. Here, we present the first formal demonstration, to our knowledge, that CTCs from chemotherapy-naive patients with extensivestage metastatic SCLC are tumorigenic in immunocompromised mice and show that CDXs faithfully recapitulate response to cisplatin and etoposide treatment of donor patients. We also report the first direct genomic comparison of single CTCs directly isolated from patient blood and the resultant matched CDXs obtained following transplant into mice. RESULTS CTCs from patients with SCLC are tumorigenic Blood samples were obtained from six patients with chemotherapynaive, extensive-stage SCLC (two males, four females) who presented between August 2012 and February 2013. Details on patient selection are in the Online Methods. All patients had a tobacco smoking history (mean 47 pack-years; s.d. 24). Median age was 69 years (range 56–78 years), and patients were performance status 1–3. Three patients (1, 3 and 6) were subsequently chemotherapy sensitive, and three patients (2, 4 and 5) had progressive disease within 3 months of completion of chemotherapy (Table 1), defined as refractory disease22. To establish whether patients’ CTCs could form tumors in immunocompromised mice, we enriched the blood from each patient (10 ml) for CTCs and injected it into one or both flanks of a non-obese diabetic (NOD) severe combined immunodeficient (SCID) interleukin-2 receptor g–deficient (NSG) mouse. The number of epithelial CTCs (EpCAM+cytokeratin+) implanted was estimated in a paired 7.5-ml blood sample by CellSearch (Table 1). CTCs from patients 1–4 generated tumors in mice (termed CDXs 1–4, respectively). We detected palpable tumors within 4 months of implantation with doubling times ranging from 5 to 21 d (Fig. 1a–c). CTC number in the paired blood sample correlated with time to palpable tumor (Supplementary Fig. 1). CTC numbers were higher in chemorefractory as compared to chemosensitive patients whose samples gave rise to CDX, and the resulting CDX grew faster (Supplementary Fig. 1). CellSearch CTC number in patients whose blood samples gave rise to CDX were all >400 CTCs per 7.5 ml (Table 1). In contrast, the CellSearch CTC counts for patients 5
2

and 6, whose blood samples have not generated CDXs, were 222 and 20 in 7.5 ml, respectively. CDXs represent clinical SCLC We assessed the histopathology and immunohistochemistry (IHC) of CDXs in comparison to their corresponding clinical specimens (patient 1: fine needle aspirate from the subcarinal lymph node; patient 2: pleural fluid cytology; patient 3: tracheal biopsy; patient 4: bronchial biopsy). We observed typical SCLC morphology23 in both diagnostic specimens and CDXs, with clusters and sheets of densely packed small round or oval cells with scant cytoplasm, enlarged hyperchromatic

a
CDX1

b

c1,200
1,000 800 600 400 200 0 1,200 1,000

CDX2

800

Tumour volume (mm3)

600 400 200 0

npg

1,200 1,000 800 600 400 200 0 1,200 1,000

CDX3

CDX4

800 600 400 200 0 0 40 80 120 160 200

Time after implant (d)

Figure 1 SCLC CTCs are tumorigenic. CTCs enriched from patients with SCLC were injected into mice. Mice carrying CDX1 and CDX3 were injected on both flanks and mice carrying CDX2 and CDX4 on the right flank. (a) Tumor-bearing mice. Scale bar, 1 cm. (b) Tumors at death. Scale bar, 2 mm. (c) Tumor volume over time after implant. Black circles, right tumor; white squares, left tumor; solid line, exponential growth line of best fit. Data show the four passage-1 mice and resultant six CDX.

advance online publication nature medicine

articles a
CK

Human mito

Patient 1

Patient 2

Patient 3

Patient 4

c

human origin and SCLC histology by staining with an anti-human mitochondrial antibody and expression of neuroendocrine markers (Fig. 2c). Response of CDXs to cisplatin and etoposide Mice bearing passage-4 CDX3, CDX2 and CDX4 were treated with cisplatin and etoposide. CDX3 exhibited the greatest response to therapy, where no treated tumor reached four times initial tumor volume over the experimental time course and median maximum tumor regression was 95% (range 72–96%), which is significantly greater than CDX2 or CDX4 (P < 0.0001; Fig. 3a–c and Supplementary Fig. 3). CDX2 exhibited an intermediate response to cisplatin and etoposide, with a significant increase, compared to vehicle-treated group, in time to four times initial tumor volume (P < 0.0001) and median maximal tumor regression of 51% (range 0–67%; P < 0.0001). CDX4 did not respond to therapy. The response of CDXs to therapy closely mirrored overall survival of the corresponding patients (9.7, 3.5 and 0.9 months for patients 3, 2 and 4, respectively). Doubling time analysis of tumors that regrew after regression (or that did not regress) revealed no significant difference in tumor growth of treated as compared to control tumors for any CDX (Fig. 3d). These data imply that degree of tumor regression is consistent with a resistant subpopulation of preexisting cells, the proportion of which dictates therapy response. Genomic analysis of CDXs Next-generation sequencing (NGS) of matching left and right flank tumors (CDX1 and CDX3) and single tumors (CDX2 and CDX4) confirmed that genomic profiles of CDXs maintained previously described characteristics of SCLC13,14. Copy number aberration (CNA) analysis showed clear patient-specific patterns of gains and losses, with CDX1 and CDX3 showing prominent CNA losses and gains in contrast to CDX2 and CDX4, which were characterized by CNA losses but far fewer gains (Fig. 4a). Left and right flank tumors from CDX1 and CDX3 were broadly similar to each other but with some differences that may reflect tumor evolution23,24 either before or after CTC implantation. For example, CDX1L but not CDX1R exhibited loss of chromosome 2p, harboring MYCN, as well as additional copies of BCL2 and SOX2. We detected an additional copy of BCL2 in CDX3R but not CDX3L (Fig. 4a and Supplementary Table 1). For all six tumors, there were deletions affecting 13q (containing RB1), 17p (containing TP53) and 10q (containing PTEN). CNA analysis of 13 individual genes frequently altered in SCLC14,15 confirmed allelic loss of RB1, TP53 and PTEN in all six tumors and loss of RASSF1 and FHIT in all CDXs except CDX4 (Fig. 4b and Supplementary Table 1). However, characteristic SCLC-associated CNA increases of genes including SOX2 (refs. 14,15) were detected only in CDX1 and CDX3 (Fig. 4b and Supplementary Table 1), a pattern also seen with extended analysis of 6,341 cancer-related genes (see Online Methods and below). Sequence analysis of CDX1L, CDX1R and CDX2 revealed large numbers of single nucleotide variants (SNVs) and smaller numbers of insertions and deletions (indels) (Supplementary Table 2). We performed targeted Sanger sequencing on 18 of the SNV loci identified by NGS whole-genome sequencing (WGS), where in 16/18 cases we identified the identical SNV (Supplementary Table 3). The overall confirmation rate of 89% is consistent with previously reported false discovery rates14. Although the absence of patient germline DNA limited unambiguous identification of somatically acquired mutations in CDXs, a large number of identified SNVs were previously identified or predicted as deleterious, associated with disease or
3

Chrom

Syn

CD56

b
CK

Left

CDX1

Right

CDX2

Left

CDX3

CD56 Right CDX4

? 2014 Nature America, Inc. All rights reserved.

npg

cPARP

Ki67

CD56

Syn

Chrom

Figure 2 CDXs and mouse micrometastases are representative of patient specimens. (a,b) Patient specimens (a) and CDXs (b) stained for cytokeratins (CK), chromogranin A (Chrom), synaptophysin (Syn) and CD56. CDXs were also stained for Ki67 and cleaved PARP (cPARP). (c) Lungs from a mouse bearing CDX2 stained for human mitochondria (Human mito), chromogranin A and CD56. Asterisks indicate SCLC micrometastases. Scale bars, 50 ?m. Data show the six passage-1 CDXs and images of lungs from mice bearing passage-1 CDX2.

nuclei, inconspicuous nucleoli, speckled chromatin and focal nuclear molding (Fig. 2a,b). CDXs demonstrated minimal stroma, expressed at least one cytokeratin (detected using a pan-cytokeratin antibody) and neuroendocrine markers synaptophysin, chromogranin A and CD56. We frequently observed mitotic and apoptotic cells (Ki67 and cleaved poly ADP ribose polymerase (PARP) indices ~75% and 3%, respectively), which are typical of SCLC. Crush artifact13 and necrotic foci were also frequent. To determine whether metastases were present in mice bearing CDXs, we harvested internal organs from mice bearing CDX1 and CDX2 but did not note macrometastases on visual inspection. However, we detected micrometastases (indicated by human DNA detection using quantitative PCR (qPCR)) in the lungs of both mice and brain of the mouse bearing CDX1 (Supplementary Fig. 2). Subsequent detailed histological examination of serial lung sections revealed small clusters of tumor cells in the alveolar wall (Fig. 2c) and scattered single cells infiltrating pulmonary parenchyma (data not shown). Metastatic foci were composed of <20 cells (three times larger than lymphocytes) with scant cytoplasm, dispersed chromatin and irregular nuclei, consistent with SCLC. We confirmed their
nature medicine advance online publication

Chrom

articles
CDX3 CDX2
patient 2 OS 3.5 months
100 80 60 40 P = 0.0004 0 20 40 60 20 0 0 P < 0.0001 20 40 60 100 80 60 40 20 0 0 P = 0.20 20

Tumor < 4×ITV (%)

100 80 60 40 20 0

Maximum regression (%)

a

CDX4
patient 4 OS 0.9 months

patient 3 OS 9.7 months

c
100 80 60 40 20 0

***

***

*** ***

***

NS

Vehicle Cis/etop
40 60

b
Tumor volume (mm3)

1,200 1,000 800 600 400 200 0 0 20 40

1,200 1,000 800 600 400 200 60 0 0 20 40

1,200 1,000 800 600 400 200 60 0 0 20

d
Doubling time after regression (d)
40 30 20 10 0

Ve C/E Ve C/E Ve C/E CDX3 CDX2 CDX4 NS NS NS

Vehicle Cis/etop
40 60

Time after randomization (d)

Time after randomization (d)

Time after randomization (d)

Ve C/E Ve C/E Ve C/E CDX3 CDX2 CDX4

? 2014 Nature America, Inc. All rights reserved.

Figure 3 CDXs mirror patient response to therapy. Mice bearing passage-4 CDX3, CDX2 or CDX4 were treated with cisplatin and etoposide or vehicle control, and the tumor volume monitored. CDX3, n = 11 per group; CDX2 n = 14 in cisplatin-and-etoposide group and n = 13 in vehicle group; CDX4 n = 15 per group. (a) Kaplan-Meier survival curves comparing vehicle and cisplatin-and-etoposide–treated groups from randomization until the tumor reaches 4× initial tumor volume (4×ITV). P calculated by log-rank test. (b) Tumor volume of vehicle and cisplatin-and-etoposide–treated groups over time after randomization. Data represent mean ± s.e.m. (c,d) The maximum regression observed for each tumor relative to initial tumor volume (c) and the doubling time, calculated after growth had recommenced if regression was observed (d), for each CDX and treatment group (Ve, vehicle; C/E, cisplatin and etoposide). Line and error bars represent mean and s.e.m. NS, not significant; ***P < 0.001 according to unpaired two-tailed t-test. Patient overall survival is time from blood draw until death.

both (Supplementary Table 2). Furthermore, although CDX1L and CDX1R share a large proportion of genetic lesions, we detected >25% SNVs in only one of the tumors (Supplementary Table 2), indicating tumor heterogeneity and evolution23,25. As expected from the high

mutation frequency in SCLC25–27, we identified patient-specific mutations in both RB1 and TP53 (Fig. 4c,d and Supplementary Table 4). Short tandem repeat DNA fingerprint profiles of CDX samples (Supplementary Table 5) failed to match cell lines in the American

a
CDX1L CDX1R CDX2 CDX3L CDX3R CDX4

b
MYCL1 BCL2 CCNE1 MYCN SOX2 EGFR MYC FGFR1 PTEN RB1 TP53 FHIT RASSF1 Increased

Chemosensitive CDX1L CDX1R CDX3L CDX3R 2 3 3 2 3 3 4 2 2 2 6 6 1 2 2 2 4 3 4 4 2 2 2 2 4 2 2 4 2 2 2 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

Refractory CDX2 CDX4 2 2 2 2 2 2 1 2 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 2 1 2

npg

c

Decreased

N-terminus WT RB1 CDX1 RB1 CDX2 RB1 Rb51 RB1 M5714 RB1 373

Domain A

Spacer Domain B C-terminus

580 640

771

928

d

248 273

Frequency of TP53 missense mutations in human cancers TAD 1/2 PRR

175 245 249 282 220

DNA-binding domain
CDX2 p.V174G CDX3 CDX4 p.y220C p.E298* CDX1 p.G245C

OD CTD

Figure 4 Genomic analysis of CDX. The left (L) and right (R) flank tumors from CDX1 and CDX3 and the single-flank CDX2 and CDX4 tumors were subjected to whole-genome and focused NGS as described in Online Methods. (a) Circos plots summarizing CNA data from all CDXs. Gains are shown in red, losses in blue and no change in gray. (b) CNA analysis of the genes showing frequent gains or losses in SCLC. (c) Position of predicted truncations in RB1 from CDX1 and CDX2 (no RB1 changes were detected in CDX3 or CDX4) alongside two reported RB1 mutations 43,44. (d) Position of TP53 mutations from CDXs 1–4 alongside the frequency of reported TP53 mutations45. TAD 1/2, transactivation domain 1 and 2; PRR, proline-rich region; OD, oligomerization domain; CTD, C terminal domain. p.E298* indicates a nonsense mutation. All indicated mutations were verified independently by a single round of NGS and Sanger sequencing in duplicate.

4

advance online publication nature medicine

articles
Type Culture Collection or our internal database, effectively ruling out the possibility of cell line contamination. Comparison of patient CTCs and CDXs To determine whether CDXs were derived from the same CTC pool enriched by CellSearch, we compared genomic profiles of CTCs isolated from the parallel enumeration of blood samples from patients 2 and 4 to their corresponding CDXs. We isolated CellSearch-enriched CTCs by DEPArray followed by whole-genome amplification (WGA) and WGS-based CNA analysis. We isolated six single CTCs, two pools of ten CTCs and ten white blood cells (WBCs) (germline samples) from patient 2 and two single CTCs, a pool of ten CTCs and ten WBCs from patient 4. Both principal component analysis (PCA) of genomewide CNAs (Fig. 5a) and hierarchical clustering of copy number values for 6,341 cancer-related genes (Fig. 5b) of isolated CTCs strongly correlated with those of their corresponding CDXs and were distinct from those of unrelated CDXs and WBCs. CTC5 from patient 2 exhibited substantial CNA differences from the other patient 2 CTCs and CDX2, suggesting a degree of CTC heterogeneity. Indeed, CTC5 from patient 2 also exhibited an increased size and larger nucleus compared to the other five single CTCs from patient 2 (Fig. 5c). We performed targeted Sanger sequencing on patient 2 CTCs at TP53 and RB1 loci (shown to be mutated in CDX2) (Fig. 5c). The TP53 c.440T>G transversion was present in all CTC samples, with wild-type

a

b
P2CTC5 P2CTC6 P2CTC1 P2CTC4 P2CTC2 CDX2A CDX2 P2CTC3 P2CTCP1 P2CTCP2 P4CTC1 CDX4 CDX4A P4CTC2 P4CTCP CDX1LA CDX1L CDX1R CDX1RA CDX3L CDX3LA CDX3R CDX3RA P2WBC P4WBC

? 2014 Nature America, Inc. All rights reserved.

1

MYCN FHIT RASSF1

Tumor (no WGA)

WGA Tumor

WBC

CTC

2
CDX/patient 1

EGFR MYCL1 SOX2 MYC BCL2 PTEN RB1 TP53 FGFR1 CCNE1

3

CDX/patient 2 CDX/patient 3 CDX/patient 4

c
TP53F

C

C

C

C

C

C

C

C

A

npg

G

G

G

G

G

G

G

G

T

TP53R
TT TT TT TT TT TT T

RB1

No PCR band

No PCR band

CTC1

CTC2

CTC3

CTC4

CTC5 Patient 2

CTC6

CTCP1

CTCP2

WBC

Figure 5 Molecular comparison of CDXs and patient CTCs. Single CTCs, pools of ten CTCs and pools of ten WBCs isolated from patients 2 and 4 were wholegenome–amplified along with 1 ng of DNA from CDXs 1–4. CNA analysis was carried out on the amplified material and unamplified tumor DNA. (a ) PCA of genome-wide CNA data. (b) Hierarchical clustering of copy number values of 6,341 selected cancer-related genes. The positions of 13 genes showing frequent gains or losses in SCLC are indicated to the right of the heatmap. (c) Sanger sequencing of the RB1 and TP53 mutations (detected in CDX2) in six single CTCs, two pools of ten CTCs and a pool of ten WBCs isolated from patient 2, with images of a cytokeratin-stained single CTC from DEPArray. Red arrows indicate somatic mutations, and blue arrows indicate the corresponding unmutated regions. Sanger sequencing of all indicated samples was carried out in duplicate with representative traces presented.

nature medicine advance online publication

5

articles
sequence in corresponding WBC samples. Sanger sequencing revealed the presence of the RB1 c.1963_1963insT in all CTC samples for which locus-specific PCR was obtained, but it was absent in the WBC sample. These data support the hypothesis that CellSearch-enriched CTCs are genetically highly related to tumorigenic SCLC CTCs. DISCUSSION We previously demonstrated that the number of CTCs detected by CellSearch (expressing EpCAM and cytokeratins) has independent prognostic significance in SCLC19, suggesting their biological and functional importance. Prior to the current study, the viability and tumor-initiating capacity of SCLC CTCs was assumed but unknown. Here, we demonstrate for the first time, to our knowledge, that CTCs from patients with SCLC can form tumors in immunocompromised mice with preserved morphological and genetic characteristics. CDXs also faithfully recapitulate responses of donor patients to platinum and etoposide, the standard-of-care chemotherapy for SCLC, enabling clinically relevant studies of SCLC biology and a readily generated and sustainable patient-derived model to test targeted therapeutics. These data demonstrate formally that CTCs are tumorigenic and that the tumors they form recapitulate donor patients’ tumor biology, which we believe confirms the assumed importance of CTCs in disease progression. Moreover, the combined WGS data from isolated CTCs, CDXs and WBCs confirm that CellSearch-enriched CTCs (EpCAM+cytokeratin+) are representative of, or closely related to, the tumor initiating cells present in the blood of patients with SCLC. These new findings open up the possibility of developing and routinely implementing personalized medicine strategies for patients with SCLC based on simple blood collection with subsequent and rapidly reported molecular analysis of CTCs. This is particularly relevant in a disease where repeat tumor biopsies are rarely obtained. In our current study, only blood samples with >400 EpCAM+cytokeratin+ CTCs per 7.5 ml blood gave rise to CDXs. In ongoing investigations, we attempted engraftment of CTCs in mice from 19 chemotherapy-naive, extensive-stage patients with SCLC (including the six presented) with at least 5 months of follow-up to detect tumor formation (C.L.H., C.J.M. and C.D., unpublished data). Tumor engraftment occurred for 9/19 patients (including the four CDXs reported in the current study: 47% take rate). The EpCAM+cytokeratin+ CTC number was 160–7,687 per 7.5 ml blood (median 901, mean 1,974) for the patients whose CTCs generated CDXs and 0–2,048 per 7.5 ml blood (median 31, mean 274) in patients whose CTCs did not. Eight patients had a CTC number >400 per 7.5 ml blood with CDX generation from seven of these eight patients’ samples. Another recent study reported tumor formation from CTCs directly explanted from patients with metastatic breast cancer28; three hundred and fifty patients were recruited and CTCs injected into femurs of 118 immunocompromised mice to generate CTC tumors from three patients. We speculate that the vast discrepancy in take rate compared to our study is due to significantly higher CTC burden in SCLC (2,915 ± 8,115 CTCs per 7.5 ml blood, mean ± s.d.21) compared to breast cancer (84 ± 885 CTCs per 7.5 ml blood, mean ± s.d. 29). A previous study demonstrated that buffy coat preparations from patients with prostate or colorectal cancer formed tumors30, though neither the presence of CTCs in the buffy coat nor the demonstration that the mouse tumors were of human prostate or colorectal origin were reported. The presence of matched somatic TP53 mutation in all CTCs examined in our study and the high degree of overall similarity in CNA patterns of CTCs and CDXs suggests that CTCs from a patient with extensive-stage SCLC are largely homogeneous. This is consistent with
6

recent CTC analysis from patients with lung cancer that also showed a high degree of similarity amongst CTCs from the same patient31. Indeed, a study of disseminated cancer cells (DCCs) in metastatic breast and prostate cancer revealed relatively homogeneous genomes suggesting expansion of a dominant clone32, whereas heterogeneity of CTCs and DCCs is primarily seen in early-stage breast cancer DCCs32 and colorectal CTCs33,34. The current study builds upon and extends previous CTC molecular analysis33,35–37 by providing a means of functionally testing molecular findings via selective interventions in corresponding CDXs. The rapid progression of SCLC prevents ‘avatar trials’38 using CDXs, but our study has implications for treatment of inherent and acquired drug-resistant disease. Although previous studies found no significant differences in overall genomic architecture between resected (likely to be chemosensitive) and autopsy (likely to be chemoresistant) SCLC cases14, this does not exclude the existence of a predictive genomic signature for inherent resistance to conventional chemotherapy. In the present study, we observed gene copy number losses in chemorefractory and chemosensitive samples, but although CNA gains were frequent in chemosensitive CDXs and CTCs, they were far rarer in chemorefractory samples. The chemorefractory CNA ratio seen here was seldom observed in previous SCLC CNA studies14. One possible explanation is that CNA gains (seen in the majority of patients with SCLC14) are responsible for conferring initial chemosensitivity, which is also observed in most patients with SCLC. As our biobank of SCLC CTCs isolated from patients with known clinical outcomes grows, the hypothesis that lack of CNA gain is associated with inherent chemotherapy resistance can be tested. Another notable difference between CDXs generated from chemosensitive and chemorefractory patients is the faster growth rate of CDXs from the chemorefractory patients (Fig. 1c and Supplementary Fig. 1b). This may reflect the more aggressive disease in the chemorefractory patients who had shorter overall survival. However, this hypothesis needs to be tested in a larger cohort. Detection of circulating tumor DNA in in the plasma of patients with cancer shows great potential as a ‘liquid biopsy’ and has been applied to SCLC39. Serial monitoring of circulating tumor DNA with altered prevalence of mutations in patients with emerging resistance to targeted therapies40,41 could be used for future treatment decision making. We consider the molecular analysis of CTCs as a complementary approach42. Although more technically challenging, CTC analysis offers advantages, including detection of co-expressed genetic defects within tumor cells, which is of likely importance in understanding drug resistance mechanisms, and comparison of single CTCs with CDXs to model drug-imposed selection and therapy responses. Our CDX models complement previously reported PDX models, which recapitulated patient responses to chemotherapy11,12. The main advantage of our CDX approach is the potential to examine mechanisms that underpin the acquired drug resistance commonly observed in SCLC. A patient’s blood sample acquired before and after drug-resistant relapse can now be used to generate CDX models for comparison. In summary, these unique CDX models, generated from sequentially available, minimally invasive clinical samples now provide an unprecedented opportunity to study SCLC biology from diagnosis through treatment to progression. CDX models will also facilitate the search for new druggable targets in SCLC and enable routine in vivo testing of targeted therapies for a disease with clear unmet medical need. METHODS Methods and any associated references are available in the online version of the paper.
advance online publication nature medicine

npg

? 2014 Nature America, Inc. All rights reserved.

articles
Accession codes. Next-generation sequencing data have been deposited in the NCBI Sequence Read Archive with BioSample accession codes SAMN02803803, SAMN02803804, SAMN02803805, SAMN02803806, SAMN02803807 and SAMN02803808.
Note: Any Supplementary Information and Source Data files are available in the online version of the paper. ACKNOWLEDGMENTS We are indebted to the patients who agreed to donate their blood samples for this study. We thank R. Marais, N. Jones and D. Ogilvie for their constructive comments on the manuscript. We thank M. Dawson, M. Lancashire, S. Bramley, J. Halstead and J. Castle, who enumerated CTCs using CellSearch. We thank A. Jardine for administrative support and M. Greaves, our laboratory manager. This research was supported by Cancer Research UK via core funding to the Cancer Research UK Manchester Institute (C5759/A12328), the Manchester Experimental Cancer Medicine Centre (C1467/A15578), the Manchester Cancer Research Centre (A12197) and their Translational Research Award for 2012. Funding to support this work was also provided via the European Union CHEMORES FP6 (contract number LSHG-CT-2007-037665). R.L.M. and L.C. were supported by education grants from Cancer Research UK and AstraZeneca. AUTHOR CONTRIBUTIONS C.L.H., P.K. and B.B. performed in vivo studies, F.T., R.P., K.L.S. and D.N. conducted histopathological examinations, D.G.R., D.J.B., S.D.P., A.G., J.A., M.G.K., M.A., L.C. and S.F. conducted the genomic analyses, Y.L., C.T., C.J. Miller and G.B. performed the bioinformatic analysis, K.M. oversaw CTC enumeration by CellSearch, R.L.M., L.C., L.P. and F.B. recruited and consented patients and collected blood samples, C.J. Morrow, C.J. Miller, G.B., F.B. and C.D. conceived and directed the study, interpreted the data and wrote the manuscript. All authors discussed the results and commented on the manuscript. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests.
Reprints and permissions information is available online at http://www.nature.com/ reprints/index.html.
1. Einhorn, L.H., Fee, W.H., Farber, M.O., Livingston, R.B. & Gottlieb, J.A. Improved chemotherapy for small-cell undifferentiated lung cancer. J. Am. Med. Assoc. 235, 1225–1229 (1976). 2. Evans, W.K. et al. VP-16 and cisplatin as first-line therapy for small-cell lung cancer. J. Clin. Oncol. 3, 1471–1477 (1985). 3. Sierocki, J.S. et al. cis-Dichlorodiammineplatinum(ii) and VP-16–213: an active induction regimen for small cell carcinoma of the lung. Cancer Treat. Rep. 63, 1593–1597 (1979). 4. Gazdar, A.F. et al. Establishment of continuous, clonable cultures of small-cell carcinoma of lung which have amine precursor uptake and decarboxylation cell properties. Cancer Res. 40, 3502–3507 (1980). 5. Oboshi, S., Tsugawa, S., Seido, T., Shimosato, Y. & Koide, T. A new floating cell line derived from human pulmonary carcinoma of oat cell type. Gann 62, 505–514 (1971). 6. Pleasance, E.D. et al. A small-cell lung cancer genome with complex signatures of tobacco exposure. Nature 463, 184–190 (2010). 7. Joshi, M., Ayoola, A. & Belani, C.P. Small-cell lung cancer: an update on targeted therapies. Adv. Exp. Med. Biol. 779, 385–404 (2013). 8. William, W.N. Jr. & Glisson, B.S. Novel strategies for the treatment of small-cell lung carcinoma. Nat. Rev. Clin. Oncol. 8, 611–619 (2011). 9. Meuwissen, R. et al. Induction of small cell lung cancer by somatic inactivation of both Trp53 and Rb1 in a conditional mouse model. Cancer Cell 4, 181–189 (2003). 10. Kwon, M.C. & Berns, A. Mouse models for lung cancer. Mol. Oncol. 7, 165–177 (2013). 11. Daniel, V.C. et al. A primary xenograft model of small-cell lung cancer reveals irreversible changes in gene expression imposed by culture in vitro. Cancer Res. 69, 3364–3373 (2009). 12. Poupon, M.F. et al. Response of small-cell lung cancer xenografts to chemotherapy: multidrug resistance and direct clinical correlates. J. Natl. Cancer Inst. 85, 2023–2029 (1993). 13. Davenport, R.D. Diagnostic value of crush artifact in cytologic specimens. Occurrence in small cell carcinoma of the lung. Acta Cytol. 34, 502–504 (1990). 14. Peifer, M. et al. Integrative genome analyses identify key somatic driver mutations of small-cell lung cancer. Nat. Genet. 44, 1104–1110 (2012). 15. Rudin, C.M. et al. Comprehensive genomic analysis identifies SOX2 as a frequently amplified gene in small-cell lung cancer. Nat. Genet. 44, 1111–1116 (2012). 16. Cristofanilli, M. et al. Circulating tumor cells, disease progression, and survival in metastatic breast cancer. N. Engl. J. Med. 351, 781–791 (2004). 17. de Bono, J.S. et al. Circulating tumor cells predict survival benefit from treatment in metastatic castration-resistant prostate cancer. Clin. Cancer Res. 14, 6302–6309 (2008). 18. Hayes, D.F. et al. Circulating tumor cells at each follow-up time point during therapy of metastatic breast cancer patients predict progression-free and overall survival. Clin. Cancer Res. 12, 4218–4224 (2006). 19. Hou, J.M. et al. Clinical significance and molecular characteristics of circulating tumor cells and circulating tumor microemboli in patients with small-cell lung cancer. J. Clin. Oncol. 30, 525–532 (2012). 20. Krebs, M.G. et al. Evaluation and prognostic significance of circulating tumor cells in patients with non-small-cell lung cancer. J. Clin. Oncol. 29, 1556–1563 (2011). 21. Hou, J.M. et al. Evaluation of circulating tumor cells and serological cell death biomarkers in small cell lung cancer patients undergoing chemotherapy. Am. J. Pathol. 175, 808–816 (2009). 22. Evans, W.K. et al. VP-16 alone and in combination with cisplatin in previously treated patients with small cell lung cancer. Cancer 53, 1461–1466 (1984). 23. Swanton, C. Intratumor heterogeneity: evolution through space and time. Cancer Res. 72, 4875–4882 (2012). 24. Martinez, P. et al. Parallel evolution of tumour subclones mimics diversity between tumours. J. Pathol. 230, 356–364 (2013). 25. Arriola, E. et al. Genetic changes in small cell lung carcinoma. Clin. Transl. Oncol. 10, 189–197 (2008). 26. Mori, N. et al. Variable mutations of the RB gene in small-cell lung carcinoma. Oncogene 5, 1713–1717 (1990). 27. Wistuba, I.I. & Gazdar, A.F. & Minna, J.D. Molecular genetics of small cell lung carcinoma. Semin. Oncol. 28, 3–13 (2001). 28. Baccelli, I. et al. Identification of a population of blood circulating tumor cells from breast cancer patients that initiates metastasis in a xenograft assay. Nat. Biotechnol. 31, 539–544 (2013). 29. Allard, W.J. et al. Tumor cells circulate in the peripheral blood of all major carcinomas but not in healthy subjects or patients with nonmalignant diseases. Clin. Cancer Res. 10, 6897–6904 (2004). 30. Pretlow, T.G. et al. Prostate cancer and other xenografts from cells in peripheral blood of patients. Cancer Res. 60, 4033–4036 (2000). 31. Ni, X. et al. Reproducible copy number variation patterns among single circulating tumor cells of lung cancer patients. Proc. Natl. Acad. Sci. USA 110, 21083–21088 (2013). 32. Klein, C.A. Selection and adaptation during metastatic cancer progression. Nature 501, 365–372 (2013). 33. Gasch, C. et al. Heterogeneity of epidermal growth factor receptor status and mutations of KRAS/PIK3CA in circulating tumor cells of patients with colorectal cancer. Clin. Chem. 59, 252–260 (2013). 34. Fabbri, F. et al. Detection and recovery of circulating colon cancer cells using a dielectrophoresis-based device: KRAS mutation status in pure CTCs. Cancer Lett. 335, 225–231 (2013). 35. Heitzer, E. et al. Complex tumor genomes inferred from single circulating tumor cells by array-CGH and next-generation sequencing. Cancer Res. 73, 2965–2975 (2013). 36. Klein, C.A. et al. Genetic heterogeneity of single disseminated tumour cells in minimal residual cancer. Lancet 360, 683–689 (2002). 37. Klein, C.A. et al. Combined transcriptome and genome analysis of single micrometastatic cells. Nat. Biotechnol. 20, 387–392 (2002). 38. Morelli, M.P. et al. Prioritizing phase I treatment options through preclinical testing on personalized tumorgraft. J. Clin. Oncol. 30, e45–e48 (2012). 39. Board, R.E. et al. Isolation and extraction of circulating tumor DNA from patients with small cell lung cancer. Ann. NY Acad. Sci. 1137, 98–107 (2008). 40 Bettegowda, C. et al. Detection of circulating tumor DNA in early- and late-stage human malignancies. Sci .Transl. Med. 6, 224ra224 (2014). 41. Dawson, S.J. et al. Analysis of circulating tumor DNA to monitor metastatic breast cancer. N. Engl. J. Med. 368, 1199–1209 (2013). 42. Krebs, M.G. et al. Molecular analysis of circulating tumour cells—biology and biomarkers. Nat. Rev. Clin. Oncol. 11, 129–144 (2014). 43. Lohmann, D.R. et al. Constitutional RB1-gene mutations in patients with isolated unilateral retinoblastoma. Am. J. Hum. Genet. 61, 282–294 (1997). 44. Szijan, I., Lohmann, D.R., Parma, D.L., Brandt, B. & Horsthemke, B. Identification of RB1 germline mutations in Argentinian families with sporadic bilateral retinoblastoma. J. Med. Genet. 32, 475–479 (1995). 45. Joerger, A.C. & Fersht, A.R. The tumor suppressor p53: from structures to drug discovery. Cold Spring Harb. Perspect. Biol. 2, a000919 (2010).

npg

? 2014 Nature America, Inc. All rights reserved.

nature medicine advance online publication

7

ONLINE METHODS
Patient selection and blood collection. From August 2012 to February 2013, 55 patients were recruited to our broader program of SCLC biomarker research. Patients had histologically or cytologically confirmed chemotherapy-naive SCLC and were referred to a tertiary cancer center, The Christie Hospital NHS Trust. The study was prospectively approved by the NHS NorthWest 9 Research Ethical Committee. Clinical and demographic data were collected. During this period, we initiated our CDX study, and 11 patients provided additional informed consent that specified their samples could be used for in vivo studies and genetic analysis in accordance with UK regulatory requirements. The 11 patients approached were selected for the CDX study as their clinic appointments coincided with the capacity within the in vivo research team for blood processing and enriched CTC implantation in mice. Seven of the 11 patients had the required clinical features of extensive (metastatic stage) disease and were chemotherapy naive. One of these patients was excluded from the study because the recipient mouse showed signs of ill health (confirmed to be non–cancer related), was culled 62 days following CTC implantation and was therefore uninterpretable with respect to CDX formation. Blood was drawn at CDX study entry before administration of chemotherapy and immediately transferred to the laboratory for processing. Blood (10 ml) was drawn into CellSave tubes (Janssen Diagnostics) for CTC enumeration using the CellSearch platform19. CTCs thus defined expressed EpCAM and cytokeratins (cytokeratins, 8, 18 and 19), were >4 ?m in diameter and had an intact DAPIstained nucleus. A paired blood sample (10 ml) was drawn into EDTA vacutainers (Becton Dickinson). Patients’ subsequent response to treatment was evaluated by computed tomography (CT) imaging performed before and following 4 cycles of chemotherapy, or earlier if clinically indicated. Patients who had a radiological response to chemotherapy that was sustained for greater than 3 months following completion of therapy were classified as chemotherapy sensitive. Patients with no evidence of response to therapy or progression within 3 months following completion of therapy were classified as chemotherapy refractory as previously described22. CTC enrichment before implantation into mice. An EDTA blood sample from a patient with SCLC was mixed with 500 ?l RosetteSep Human Circulating Epithelial Tumor Cell Cocktail (Stem Cell Technology) and incubated for 20 min at room temperature with constant mixing. Blood was diluted with 10 ml 9:1 Hank’s Balanced Saline Solution (HBSS) (Life Technologies): HITES medium (RPMI 1640 (Life Technologies), 5 ?g ml-1 insulin, 10 ?g ml-1 transferrin, 10 nM b-estradiol, 30 nM sodium selenite, 10 nM hydrocortisone (Sigma)), layered over 15 ml Ficoll-Plaque Plus (GE Heathcare) and centrifuged at 1,200 × g for 20 min. Cells at the medium-Ficoll boundary were collected, diluted with 30 ml 9:1 HBSS:HITES and centrifuged at 250 × g for 5 min. The cell pellet was resuspended in 100 ?l ice cold HITES and mixed with 100 ?l Matrigel (BD Biosciences) and kept on ice. Growth of CTC tumors in immunocompromised mice. All procedures were carried out in accordance with Home Office Regulations (UK) and the UK Coordinating Committee on Cancer Research guidelines and by approved protocols (Home Office Project license no. 40-3306 and Cancer Research UK Manchester Institute Animal Welfare and Ethical Review Advisory Board). 100–200 ?l of CTCs/HITES/Matrigel was injected subcutaneously into one or both flanks of 8-16 week old female NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (Jackson Laboratories). Mice were housed in individually vented caging systems in a 12-h light/12-h dark environment and maintained at uniform temperature and humidity. Mice were monitored twice weekly for signs of tumor growth, and once a palpable tumor was present this was measured twice a week by calipers and tumor volume calculated as tumor length × tumor width2/2. When the total tumor burden reached 1,000 mm3 or there were demonstrable signs of ill health, the animal was killed, tumor fragments were passaged into NSG mice and the remainder of the tumor was harvested for IHC analysis or DNA or RNA extraction. The internal organs were also harvested for further analysis. No statistical method was used to predetermine sample size as no previous data were available. Cisplatin and etoposide treatment. Thirty female NSG mice were implanted with passage 4 CDX2, CDX3 or CDX4 with the expectation that 10 tumors would not grow successfully, leaving at least 10 animals per treatment group (drugs vs. vehicle). When tumors reached 200–250 mm3, they were randomized by sequential

assignment to cisplatin and etoposide or vehicle treatment groups. Animals were treated by intraperitoneal injection with 5 mg kg-1 cisplatin (Sigma) dissolved in 0.9% saline solution on day 1 and 8 mg kg-1 etoposide (Sigma) dissolved in 12.5:1 0.9% saline solution:0.1% citric acid in 1-methyl-2-pyrrolidinone on days 1, 2 and 3, or corresponding vehicle only. Tumor volume was monitored blinded to treatment group every three days until tumor reached four times initial tumor volume (4xITV) or until animal health deteriorated (censored in survival analysis). Survival analysis was performed in Graphpad Prism with comparison of survival curves by log-rank (Mantle Cox) test. Doubling time was calculated by nonlinear curve fitting of an exponential growth equation (Graphpad Prism). If the tumor exhibited regression, the doubling time was calculated from the point tumor growth recommenced. For comparison of maximum regression and tumor doubling time, normal distribution was assessed by D’Agostino and Pearson omnibus normality test and groups compared by unpaired two-tailed t-test. Human- and mouse-specific qPCR. Tissue was collected from autopsied animals and gDNA isolated using the Ambion RecoverAll kit (Life Technologies). Quantitative PCR was then performed using Bioline SensiFAST qPCR reagents (Bioline) and 10 ng of gDNA from each sample with murine- and human-specific primers targeting the prostaglandin E receptor 2 (Ptger2/PTGER2) and phosphoserine aminotransferase 1 (Psat1/PSAT) genes as described previously46,47. Dissociation curve analysis was used to distinguish between human- and murinespecific amplification products. Normal murine lung gDNA and HNV gDNA were included in all experiments as controls. Immunohistochemistry. IHC was performed on formalin-fixed, paraffinembedded 4-mm tumor and normal tissue sections using antibodies to cytokeratins (pan-cytokeratin antibody to cytokeratins 1–8, 10, 13–16 and 19, mouse AE1/AE3, M3515, 1:60, Dako), CD56 (mouse, 1B6, NCL-CD56-1B6, 1:100, Novocastra), chromogranin A (mouse, LK2H10 + PHE5, MP-010-CM1, 1:600, Menapath), synaptophysin (mouse, 27G12, NCL-L-SYNAP-299, 1:200, Novocastra), cleaved PARP (mouse, Asp214, 51-9000017 1?:?100, BD Pharmingen) and Ki67 (mouse, MIB-1, M7240, 1:600, Dako). A human-specific antimitochondria antibody (rabbit, 1113-1, 1:500, Abcam) was used to detect micrometastases in mouse tissues. Antibody incubations and detection were carried out at room temperature on a Menarini IntelliPATH FLX (A. Menarini Diagnostics) using Menarini’s reagent buffer and detection kits unless otherwise noted. Antigen retrieval was performed in a pressure cooker using access super retrieval fluid (MP-606-PG1) for all antisera except AE1/AE3, which was 10-min incubation with protease (MP-960-K15) on IntelliPATH, and anti-mitochondria, which was performed in citrate buffer pH 6 and microwaved for 15 min at 98 °C. Isotype controls used were rabbit immunoglobulin fraction and mouse IgG1 from Dako. Digital images of whole tissue sections were acquired using a Leica SCN400 histology scanner (Leica Microsystems). Ki67 and cPARP positive index were evaluated using Definiens Developer XD version 2.0.4 and the Tissue Studio Portal version 3.51 (Definiens AG). Regions of interest (ROIs) within the tissue sections were first identified using Definiens Tissue Studio via machine learning technology across pathological samples and tissue control, so that the full range of contrast was defined. Within these ROIs, nuclei were detected and classified as positive or negative based on IHC staining thresholds. Tumor DNA extraction and WGS. CTC tumors were disaggregated using a sterile scalpel and gDNA isolated using the QiaAmp DNA Mini kit (Qiagen, Hilden, Germany). DNA libraries were generated from 50 ng gDNA in the NEBNext Ultra DNA Library kit (NEB) and sequenced on an Illumina HiSeq2500 instrument using the TruSeq PE Cluster Kit V3 and TruSeq SBSv3 chemistry. CTC isolation and WGA. CTCs and WBCs (pre-stained with antibody to CD45, pan-CK and DAPI) were aspirated from the CellSearch cartridge used for the CTC enumeration, and single and groups of cells were isolated using the DEPArray system (Silicon Biosystems) as per manufacturer’s instructions. WGA of single CTCs, pools of 10 CTCs or WBCs and genomic DNA (1 ng input) was performed using the Ampli1 WGA kit (Silicon Biosystems) according to manufacturer’s instructions48. NGS of WGA samples. To remove the Ampli1 amplification primer, all WGA products (250ng) were digested with MSE1 (New England Biolabs) following

npg

? 2014 Nature America, Inc. All rights reserved.

doi:10.1038/nm.3600

nature medicine

supplier’s instructions. Digested samples were quantified using Qubit (Life Technologies) and sonicated with a Bioruptor UCD-200 (Diagenode) for 10 cycles (T1 = 30 s, T2 = 30 s) to produce fragments of about 300–350 bp that were checked by a Bioanalyzer (2100 Bioanalyzer, Agilent Life Sciences and Chemical Analysis). DNA libraries were prepared using the NEBNext Ultra DNA Library Prep Kit for Illumina (New England Biolabs) according to manufacturer’s instructions. Final library PCR products were quantified using KAPPA Library Quantification Kits for Illumina (KAPABiosystems), Bioanalyzer and the Quant-iT assay using Qubit Quantitation Platform according to the manufacturer’s instructions. Sequencing was carried out on an Illumina MiSeq System with paired-end 150-bp runs, and the resultant reads were base called, filtered by quality metrics and aligned to the human reference sequence as recommended by the manufacturer. GeneRead of CDX3 and 4 tumor and CTCs. Genomic DNA from CDX3, CDX4 and WGA product from patient 4 WBCs were analyzed using the GeneRead DNAseq Human Lung Cancer panel (Qiagen) as described in the manufacturer’s protocol. Briefly, 80 ng of DNA was PCR amplified using targeted multiplexed amplicons, and the resulting material cloned into a NEBNext Ultra DNA Library. The libraries were then run on an Illumina MiSeq and analyzed using the Qiagen NGS portal (http://ngsdataanalysis.sabiosciences.com/NGS/).
? 2014 Nature America, Inc. All rights reserved.

FREEC web site (http://bioinfo-out.curie.fr/projects/freec/). For each sample, an estimated copy number was assigned to every cytoband of the human genome (version hg19) by calculating the weighted mean of the overlapping copy number estimates (as computed by FREEC) that map to the given cytoband and passed to Circos53 (Fig. 4a). FREEC predicted copy number data was averaged across cytobands, as before, and imported into MeV to generate the PCA data (median centering mode with recommended MeV algorithm; Fig. 5a). These copy numbers were also mapped to genome coordinates using the Bioconductor package annmap to provide ENSEMBL version 70 annotation54 and clustered in MeV using Pearson correlation average linkage (Fig. 5b). Cancer-related genes. The geneRIF database was downloaded from the NCBI web site (http://www.ncbi.nlm.nih.gov/gene/about-generif/) on 3 February 2014. A human gene was regarded as a cancer-related gene if its RIF text contains at least one of the 10 key words or word stems (carcinogen, cancer, carcinoma, tumor, leukemia, tumour, oncogen, leukaemia, oncolog, malignan). A list of 6,682 cancer-related protein-coding genes were compiled by mapping ENSEMBL gene ID and gene symbol in the geneRIF database to ENSEMBL (v70). Only autosomal chromosomes were used in CNA analysis, so 268 genes on the X, Y or mitochondrial chromosome were removed from the list. Verification of the single-cell CNA approach. As part our evaluation of the WGA/CNA approach, we examined matched single and pooled WBCs and CTCs. Six WBC samples (two pools of 10 WBCs; four single WBCs) and six CTC samples (two pools of 10 CTCs; four single CTCs) were subjected to WGA and NGS (number of mapped reads reported in Supplementary Table 6), with resultant Illumina MiSeq data analyzed for CNA at the cytoband and cancer-related gene level. From this analysis, we detected the expected clear separation of CTC and WBC samples, and within each group the single-cell and 10-cell samples gave comparable results (Supplementary Fig. 4). Based on this evaluation, we also identified a small numbers of potentially unreliable loci (0.8% for cytobands and 1.1% for cancer genes) with reported loss or gain in at least three WBC samples. These loci (Supplementary Table 7) were subsequently removed for CNA analysis of the CDX samples, which reduced the number of cancer-related genes to 6,341. Single tandem repeat DNA fingerprinting. Purified DNA from CDXs was processed using the Powerplex 21 kit (Promega) according to manufacturer’s instructions. STR profiles were compared to the ATCC cell line database and to our internal database of all cell lines in use at the CRUK Manchester Institute.
46. Alcoser, S.Y. et al. Real-time PCR-based assay to quantify the relative amount of human and mouse tissue present in tumor xenografts. BMC Biotechnol. 11, 124 (2011). 47. Thierry, A.R. et al. Origin and quantification of circulating DNA in mice with human colorectal cancer xenografts. Nucleic Acids Res. 38, 6159–6175 (2010). 48. Peeters, D.J. et al. Semiautomated isolation and molecular characterisation of single or highly purified tumour cells from CellSearch enriched blood samples using dielectrophoretic cell sorting. Br. J. Cancer 108, 1358–1367 (2013). 49. McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297–1303 (2010). 50. DePristo, M.A. et al. A framework for variation discovery and genotyping using nextgeneration DNA sequencing data. Nat. Genet. 43, 491–498 (2011). 51. Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res. 38, e164 (2010). 52. Boeva, V. et al. Control-free calling of copy number alterations in deep-sequencing data using GC-content normalization. Bioinformatics 27, 268–269 (2011). 53. Krzywinski, M. et al. Circos: an information aesthetic for comparative genomics. Genome Res. 19, 1639–1645 (2009). 54. Yates, T., Okoniewski, M.J. & Miller, C.J. X:Map: annotation and visualization of genome structure for Affymetrix exon array analysis. Nucleic Acids Res. 36, D780–D786 (2008).

Sanger sequencing (confirmation of NGS). Target amplicons (Supplementary Table 3) of tumor-associated genes were amplified by PCR from 10 ng of genomic DNA from the original xenograft tumor samples and from HNV blood. Each amplicon was purified with a PCR cleanup kit (Qiagen) and subjected to Sanger sequencing on an ABI 3130 Genetic Analyzer using the same primers used for PCR. ABI sequencing files were analyzed using 4Peaks software (http://nucleobytes.com/index.php/4peaks/) and publically available web based BLAST (http:// blast.ncbi.nlm.nih.gov/Blast.cgi) and alignment tools (http://www.ebi.ac.uk/ Tools/msa/clustalo/). WGS analysis of tumors CDX1 and CDX2. Illumina HiSeq data were aligned to the human genome (version hg19) using SMALT (http://smalt.sourceforge. net/ with default strategies: http://www.sanger.ac.uk/resources/software/smalt/). To identify potential contaminating reads of murine origin, the same data was aligned to the mouse reference genome (version mm9). Reads aligning to both human and mouse genomes were discarded. In excess of 1.45 billion reads mapped uniquely to the human genome for CDX1L and CDX1R samples and >1.28 billion reads for the CDX2 samples. Aligned paired-end reads were used to identify SNVs and short indels for each sample using GATK49. Duplicate read removal, realignment around known indels and base and variant quality score recalibration50 were performed as pre- and post-processing. Variant calling was performed using unifiedGenotyper with default settings. The putative SNVs and indels identified by GATK were annotated using ANNOVAR51. CNA analysis of WGA products. Illumina MiSeq whole-genome data from nineteen WGA samples (6 single cell CTCs, two 10-cell CTC pools and one WBC 10-cell pool from CDX2, 2 single-cell CTCs, one 10-cell CTC pools and one WBC 10-cell pool from CDX4, plus WGA products generated from 1 ng CDX1L, CDX1R, CDX2, CDX3L, CDX3R and CDR4 and three WGS samples (WGS for CDX3L, CDX3R and CDX4) were aligned to the human genome using SMALT (number of mapped reads reported in Supplementary Table 6). FREEC52 was used to identify copy number variations (window size: 50 kb; step size 10 kb for WGS samples CDX1L, CDX1R and CDX2; the adaptive window size was used for all other samples) of both the genomic (HiSeq) and the WGA (MiSeq) data. Mappability data for HG19 with an edit distance of 1 were downloaded from the

npg

nature medicine

doi:10.1038/nm.3600


相关文章:
更多相关标签: