Preserved genetic diversity in organoids cultured from biopsies of human colorectal cancer metastases

Significance

Chemotherapy has been proven in clinical studies to improve overall survival significantly. Unfortunately, there is a significant degree of heterogeneity in tumor chemosensitivity, often resulting in unnecessary treatment and needless exposure to toxic side-effects. A platform is needed that can identify preemptively which patients will or will not benefit from treatment. Tumor organoids, 3D cultures of cancer cells, present such an individualized platform. In this study we demonstrate that organoid cultures can be established from metastatic biopsy specimens with a high success rate and genetically represent the metastasis they were derived from. These data support the translation of this innovative technology to the clinic as an ex vivo screening platform for tailoring treatment.

Abstract

Tumor organoids are 3D cultures of cancer cells. They can be derived from the tumor of each individual patient, thereby providing an attractive ex vivo assay to tailor treatment. Using patient-derived tumor organoids for this purpose requires that organoids derived from biopsies maintain the genetic diversity of the in vivo tumor. In this study tumor biopsies were obtained from 14 patients with metastatic colorectal cancer (i) to test the feasibility of organoid culture from metastatic biopsy specimens and (ii) to compare the genetic diversity of patient-derived tumor organoids and the original tumor biopsy. Genetic analysis was performed using SOLiD sequencing for 1,977 cancer-relevant genes. Copy number profiles were generated from sequencing data using CopywriteR. Here we demonstrate that organoid cultures can be established from tumor biopsies of patients with metastatic colorectal cancer with a success rate of 71%. Genetic analysis showed that organoids reflect the metastasis from which they were derived. Ninety percent of somatic mutations were shared between organoids and biopsies from the same patient, and the DNA copy number profiles of organoids and the corresponding original tumor show a correlation of 0.89. Most importantly, none of the mutations that were found exclusively in either the tumor or organoid culture are in driver genes or genes amenable for drug targeting. These findings support further exploration of patient-derived organoids as an ex vivo platform to personalize anticancer treatment.

The considerable variation among patients in sensitivity to antineoplastic treatment emphasizes the need for more accurate treatment selection. Numerous attempts have been made to develop a personalized in vitro platform to predict treatment response in individual patients, but no method has presently gained clinical acceptance (1). Cell lines are not readily established for every individual patient, and although a higher take rate and early signs of predictive value have been reported for patient-derived xenograft models, the 6–8 mo required to gain sufficient material may prohibit its development as a clinical decision-making tool (2).

A recently developed 3D culture system, enriching for the stem cell population, has enabled the indefinite propagation of normal and tumor epithelial cells of a variety of tissues (3–9). These cultures, referred to as “organoids,” are established in a relatively short time frame, are easy to manipulate, and facilitate high-throughput screens (10). However, their most important feature is that cultures can be derived from the tumor tissue of each individual patient. With adequate amounts of starting material, such as tumor resections, the success rate of derivation for colorectal cancer organoids is greater than 90% (10). However, in the metastatic setting there generally is no access to resection material, and acquisition of fresh tumor tissue often is limited to biopsies. To use patient-derived tumor organoids as a personalized screening tool for tailoring treatment, it is imperative that cultures from biopsy specimens have a high success rate and that the genetic landscape of the tumor be preserved in culture. This article addresses both topics.

Materials and Methods

Patients.

Fourteen patients with metastatic colorectal cancer gave written informed consent to undergo two to four 18-gauge biopsies from accessible metastatic lesions (Fig. 1). This study was approved by the institutional review board of the University Medical Center Utrecht (protocol number NCT01855477).

Fig. 1.

Overview of the study design.

Organoid Cultures.

Organoid culturing was performed according to methods described by Sato et al. (3). Briefly, biopsies were dissected and embedded in Matrigel [Basement Membrane Matrix, growth Factor Reduced, Phenol Red free (MG), (BD; #56231)]. The Matrigel including the embedded cells was overlaid with growth medium optimized for selective outgrowth of tumor cells [Advanced Dulbecco's Modified Eagles Medium with Nutrient Mixture F-12 Hams (Ad-DF) (Invitrogen; #12634), 1% penicillin/streptomycin (Invitrogen; #15140-122), 10 mM Hepes (Invitrogen #15630-056), 1% GlutaMAX (Invitrogen #35050), 20% R-Spondin conditioned medium (3), 10% Noggin conditioned medium (3), 1× B27 supplement (Invitrogen #17504-044), 1.25 mM N-acetyl-cysteine (Sigma-Aldrich; A9165-5G), 10 mM nicotinamide (Sigma-Aldrich; N0636), 50 ng/mL human EGF (Invitrogen Biosource; PMG8043), 10 nM gastrin (Sigma; #G9145), 500 nM TGFb type I Receptor inhibitor A83-01 (Tocris; #2939), 10 μM p38 MAPK inhibitor (p38i) SB202190 (Sigma-Aldrich; S7067), 10 nM prostaglandin E2 (Tocris; #2296), and 1× Primocin (InvivoGen; #ant-pm-1)]. Fresh medium was added every 2 or 3 d. Outgrowing organoids were passaged every 7–10 d after mechanical and enzymatic disruption.

DNA Sequencing.

A separate biopsy was used to compare the mutation and copy number profiles of the tumor biopsy and biopsy-derived cultures. Blood was drawn to obtain control DNA for germline variation. DNA was isolated from organoid cultures after 2–3 mo (depending on organoid growth rate). DNA was isolated using the Tissue and Blood DNA 500 Extraction Kit (Diasorin) according to the manufacturer’s protocol. Matched blood, tumor biopsies, and biopsy-derived cultures were sequenced using the 5500xl SOLiD system (Applied Biosystems) according to the manufacturer’s protocol. The minimum tumor percentage required to enable sequencing was 30%. Tumor cell percentage and percentage of necrosis were determined by a trained pathologist based on H&E staining. Six hundred nanograms of genomic DNA were required per sample to generate barcoded fragment libraries as described by Harakalova et al. (11). Samples were enriched using SureSelect technology (Agilent Technologies) for the actionable cancer genome consisting of 1,977 genes, as previously described by Hoogstraat et al. (12) and Vermaat et al. (13). The actionable cancer genome consists of all exons of genes known to be oncogenes, tumor-suppressor genes, kinases, or functioning in pathways involved in tumorigenesis and comprises ∼16% of the exome. Libraries were sequenced to an average coverage of 150×. Mapping, variant calling, and annotation were done as described by Hoogstraat et al. (12). Somatic mutations were classified as unique for either tumor biopsy or biopsy-derived cultures if the allele frequency in the matching sample was <1%.

DNA Copy Number Analysis.

DNA copy number profiles were generated from the BAM files using CopywriteR as described by Kuilman et al. (14). In short, sequence reads outside the captured genomic regions (off-target reads) were used to generate DNA copy number profiles. A depth-of-coverage method was used for 100-kb bins, and the read count was normalized for GC content and mappability. Log₂ ratios were calculated for all tumor or organoid samples versus reference (blood) samples. The normalized and corrected log₂ ratios from CopywriteR were analyzed further by circular binary segmentation (CBS) and CGHcall (Bioconductor) (15). CBS allows the detection of segments with nearly identical copy number states. CGHcall is used to classify data points as copy number gain, loss, or neutral.

Results

Organoid culture from metastatic biopsy specimens was successful in 10 of 14 cases (71%). For eight patients DNA from the tumor biopsy and its matched biopsy-derived culture were sequenced to assess whether the original genetic diversity of the tumor was preserved in culture (Fig. 1). The remaining two patients from whom organoid cultures were established were excluded for this analysis because the tumor cell percentage of the biopsy was below 30%.

A mean of 19 (±4) somatic mutations were observed per patient (Fig. 2). Ninety percent of somatic mutations were shared between tumor biopsy and organoid cultures. Somatic mutations in driver genes or other genes of interest were invariably detected in both tumor biopsy and biopsy-derived cultures from all patients (Fig. 2). An average correlation of 0.89 (range 0.77–0.96) was detected for DNA copy number profiles between matching tumor biopsy and organoid samples (Fig. 2). The average percentage of data points detected identically as gain, loss, or neutral was 81% (range: 52–99%). Somatic mutations identified are presented in Table S1, sequencing statistics in Table S2, and tumor cell percentage and necrosis percentage of original biopsies in Table S3. Copy number profiles for matched tumor biopsies and biopsy-derived cultures per patient are shown in Fig. S1.

Fig. 2.

Copy number and mutational profiles of biopsy-derived organoid cultures and matched tumor biopsies for patients 01–08. (A) Venn diagrams displaying the number of shared and unique (tumor biopsy or biopsy-derived cultures) mutations per patient as identified by SOLiD sequencing. (B) Mutations in known driver genes or genes of interest for colorectal cancer per patient. Mutations were invariably present in both tumor biopsy and biopsy-derived organoid cultures. (C) Heat map comparing copy number aberrations between matched tumor biopsy and biopsy-derived organoid cultures in a log₂ scale. Red colors indicate gains, and blue colors indicate losses. The band on the left represents chromosomes 1–22 and X. For each patient the correlation between the tumor biopsy and biopsy-derived cultures is shown, as well as the percentage of data points detected identically as gain, loss, or neutral (overlap).

Table S1.

SOLiD data somatic mutations

Subject	Chromosome	Position	Base change	Amino acid change	Mean coverage	Gene	Variant type	Germline coverage	Tumor Coverage	Organoid coverage	Germline allele frequency	Tumor allele frequency	Organoid allele frequency	Shared (S), tumor unique (T), or organoid unique (O)
1	1	209,797,208	T/A	K705I	332	LAMB3	SNV	298	459	239	2	85	99	S
	2	179,407,388	C/A	NA	493,67	TTN	SNV	489	634	358	0	20	19	S
	2	179,486,728	C/A	W6034L	344,33	TTN	SNV	253	520	260	0	39	43	S
	3	25,665,813	G/A	P824S	775,67	TOP2B	SNV	694	1127	506	1	22	24	S
	3	119,582,332	G/A	R357W	401	GSK3B	SNV	316	640	247	1	28	39	S
	3	195,605,915	C/G	NA	45,33	TNK2	SNV	53	71	12	0	20	17	S
	4	65,298,248	A/G	NA	11,33	RP11-204H9.2	SNV	15	17	2	0	76	100	S
	5	112,175,021	C/T	Q1244*	96,67	APC	SNV	96	133	61	1	21	34	S
	6	4,059,028	G/A	V514I	57	PRPF4B	SNV	82	34	55	1	21	7	S
	8	8,999,186	C/T	NA	83,33	PPP1R3B	SNV	113	119	18	0	68	100	S
	11	106,810,698	C/G	E232Q	693,33	GUCY1A2	SNV	508	1145	427	0	27	24	S
	11	117,306,398	C/A	R1673L	301	DSCAML1	SNV	349	390	164	0	31	29	S
	12	57,849,972	C/T	P132S	180	INHBE	SNV	122	323	95	0	21	32	S
	13	32,972,674	G/A	E3342K	846,67	BRCA2	SNV	490	1434	616	0	17	28	S
	15	99,442,705	G/A	NA	351	IGF1R	SNV	241	603	209	0	26	39	S
	16	3,777,892	C/T	G2348R	49,67	CREBBP	SNV	45	76	28	0	53	68	S
	20	2,845,936	C/T	A402V	187,33	VPS16	SNV	105	347	110	0	8	19	S
	X	41,027,389	C/T	R852*	164,67	USP9X	SNV	251	118	125	0	79	93	S
	17	75,775,39	G/A	R248W	205,33	TP53	SNV	235	280	101	0	69	79	S
2	1	10,363,653	G/C	E804Q	163,67	KIF1B	SNV	128	184	179	0	8	17	S
	1	43,814,930	C/G	NA	103	MPL	SNV	86	127	96	0	17	21	S
	2	99,172,162	G/A	NA	737,33	INPP4A	SNV	606	996	610	0	24	30	S
	2	179,486,342	A/G	I6005T	899,67	TTN	SNV	1422	224	1053	0	16	28	S
	2	179,507,003	G/A	R4567C	307	TTN	SNV	714	55	152	0	44	45	S
	3	1,414,000	G/A	V504I	375,67	CNTN6	SNV	530	158	439	0	1	16	S
	3	112,991,517	G/A	A310T	20,67	BOC	SNV	7	34	21	0	41	14	S
	4	23,825,968	T/G	K271T	465	PPARGC1A	SNV	793	153	449	0	16	26	S
	4	146,467,864	C/T	A262V	152	SMAD1	SNV	287	24	145	0	38	42	S
	5	112,155,013	C/-	NA	150,67	APC	Indel	179	155	118	0	60	90	S
	5	170,818,803	G/A	A118T	240,67	NPM1	SNV	328	179	215	0	0	18	O
	6	57,075,028	T/A	I51F	474	RAB23	SNV	621	176	625	0	51	43	S
	11	92,533,992	T/A	Y2605N	783	FAT3	SNV	877	596	876	0	40	41	S
	16	77,353,825	G/A	T818M	106,33	ADAMTS18	SNV	129	122	68	0	53	72	S
	17	7,576,927	C/A	NA	144	TP53	SNV	175	124	133	0	64	99	S
	19	39,996,053	A/C	K352T	60,33	DLL3	SNV	67	70	44	0	13	16	S
	3	46,307,365	G/A	R260Q	656	CCR3	SNV	693	686	589	0	23	20	S
	12	25,398,281	C/T	G13D	207,33	KRAS	SNV	326	109	187	0	85	91	S
3	1	227,300,460	C/T	R601Q	693,67	CDC42BPA	SNV	655	587	839	0	14	35	S
	4	48,073,695	C/G	NA	237	TXK	SNV	247	282	182	0	0	18	O
	5	112,175,480	G/T	E1397*	349	APC	SNV	408	289	350	0	2	36	S
	6	32,184,739	C/A	C615F	133,33	NOTCH4	SNV	104	172	124	1	5	47	S
	10	64,968,097	G/C	S892C	285,33	JMJD1C	SNV	328	204	324	0	24	68	S
	10	114,912,204	G/A	NA	130,67	TCF7L2	SNV	96	187	109	1	7	32	S
	11	117,308,068	C/T	R1287Q	87	DSCAML1	SNV	61	137	63	2	15	94	S
	11	118,342,652	C/G	P293A	446,33	MLL	SNV	390	462	487	0	9	39	S
	12	120,653,800	T/G	K4T	27	PXN	SNV	19	49	13*	0	18	0	T
	12	121,416,813	T/C	F81S	18,33	HNF1A	SNV	8	22*	25	0	0	16	O
	15	44,843,120	T/C	V65A	161,33	EIF3J	SNV	192	145	147	0	0	25	O
	15	88,690,618	G/A	R138W	494,33	NTRK3	SNV	500	490	493	0	26	99	S
	17	7,577,095	G/T	D281E	192,67	TP53	SNV	160	297	121	0	8	95	S
	18	45,371,862	T/C	NA	276	SMAD2	SNV	297	230	301	0	20	99	S
	19	7,976,387	G/T	V377F	39,67	MAP2K7	SNV	42	51	26	5	12	77	S
	12	25,398,284	C/G	G12A	227,33	KRAS	SNV	296	265	121	0	4	26	S
4	1	160,388,795	G/T	D66Y	63,33	VANGL2	SNV	56	21	113	0	19	46	S
	1	226,566,974	T/A	−538-	154,67	PARP1	SNV	170	53	241	4	38	13	S
	1	227,216,861	C/T	R500H	284,67	CDC42BPA	SNV	357	53	444	0	32	39	S
	1	228,402,727	C/A	Q586K	23,33	OBSCN	SNV	26	8*	36	0	0	22	O
	2	179,643,820	C/T	R1330H	160,33	TTN	SNV	301	25	155	0	60	43	S
	3	10,070,375	G/A	D12N	99,67	FANCD2	SNV	235	39	25	0	36	20	S
	5	6,743,939	A/T	I161F	96,67	PAPD7	SNV	233	25	32	0	40	72	S
	5	38,960,005	C/T	E643K	211,33	RICTOR	SNV	537	33	64	0	27	42	S
	5	112,173,917	C/T	R876*	312	APC	SNV	591	69	276	0	43	98	S
	7	19,156,452	T/A	S33C	68	TWIST1	SNV	83	32	89	0	0	17	O
	8	2,887,015	C/T	E2562K	81,67	CSMD1	SNV	140	6	99	0	50	35	S
	8	38,275,833	C/T	R357Q	84,33	FGFR1	SNV	108	13	132	1	85	57	S
	8	103,307,432	G/A	P1347L	628	UBR5	SNV	717	79	1088	0	68	68	S
	9	113,496,548	C/T	R216W	318,67	MUSK	SNV	431	98	427	1	28	22	S
	11	11,373,733	G/A	R312W	205,67	CSNK2A3	SNV	324	46	247	0	13	33	S
	17	40,491,345	C/T	R152Q	228,33	STAT3	SNV	408	47	230	0	70	98	S
	20	2,842,501	G/A	R317H	45,67	VPS16	SNV	58	11	68	0	27	21	S
	X	1,404,820	C/T	NA	99,67	CSF2RA	SNV	101	48	150	0	56	40	S
	X	153,592,746	G/A	NA	32,67	FLNA	SNV	22	9*	67	0	0	19	O
	12	25,398,285	C/A	G12C	140,67	KRAS	SNV	240	56	126	1	41	56	S
	17	7,578,406	C/T	R175H	31	TP53	SNV	33	5	55	0	80	96	S
5	1	242,016,664	C/T	R96*	431	EXO1	SNV	532	385	376	0	17	34	S
	2	99,180,106	C/A	NA	5,33	INPP4A	SNV	1	4	11	0	75	55	S
	4	103,446,721	G/A	NA	74,33	NFKB1	SNV	94	58	71	0	14	21	S
	4	104,070,038	T/A	Q1244L	243,67	CENPE	SNV	388	226	117	0	22	10	S
	5	112,175,621	C/T	Q1444*	280	APC	SNV	436	226	178	0	36	88	S
	6	30,856,718	G/A	R40Q	200	DDR1	SNV	187	206	207	2	33	64	S
	6	162,683,713	C/T	D86N	210,33	PARK2	SNV	220	246	165	0	8	19	S
	7	141,424,934	C/T	R444C	235	WEE2	SNV	186	275	244	0	23	40	S
	10	17,142,212	-/A	NA	200,67	CUBN	Indel	217	204	181	0	7	18	S
	15	64,275,855	C/T	R64H	58,33	DAPK2	SNV	54	65	56	0	3	20	S
	17	27,208,869	C/T	E227K	33	FLOT2	SNV	39	26	34	0	50	94	S
	18	4,891,888	G/T	D351Y	419	SMAD4	SNV	583	353	321	0	52	100	S
	12	25,398,281	C/T	G13D	182	KRAS	SNV	202	164	180	0	49	95	S
	14	21,854,022	G/T	P2499H	35,67	CHD8	SNV	31	38	38	0	5	18	S
	17	7,578,406	C/T	R175H	28,67	TP53	SNV	43	14	29	0	50	97	S
6	1	45,809,000	G/A	D307N	125,67	TOE1	SNV	290	8	79	1	25	47	S
	2	170,009,480	G/A	NA	96	LRP2	SNV	93	32	163	0	38	4	S
	3	14,200,245	T/G	S343R	127,33	XPC	SNV	185	98	99	0	33	17	S
	4	114,186,052	G/A	NA	28,67	ANK2	SNV	55	7*	24	0	0	29	O
	5	112,174,864	AA/-	NA	256,33	APC	Indel	505	100	164	0	36	43	S
	6	32,944,234	T/A	L273H	316	BRD2	SNV	473	124	351	0	81	76	S
	7	76,111,907	G/-	NA	23,33	DTX2	Indel	31	15	24	0	27	12	S
	8	41,166,327	C/T	A118T	11,67	SFRP1	SNV	10	17	8	0	35	12	S
	8	103,307,229	T/G	K1390T	542	UBR5	SNV	661	135	830	0	17	0	T
	9	71,504,018	T/G	F147C	237,33	PIP5K1B	SNV	363	51	298	0	0	23	O
	9	77,359,089	T/G	K1690Q	274,33	TRPM6	SNV	405	96	322	0	0	21	O
	18	9,522,319	G/A	E289K	246,67	RALBP1	SNV	422	31	287	1	23	3	S
	19	1,455,191	C/A	L153I	32,67	APC2	SNV	77	10	11	0	50	82	S
	19	50,374,886	T/G	K182T	104,67	AKT1S1	SNV	91	87	136	0	0	16	O
	19	56,042,661	C/T	G102D	12,33	SBK2	SNV	6	19	12	0	21	25	S
	20	8,639,221	G/C	Q244H	355,33	PLCB1	SNV	440	185	441	1	16	7	S
	X	591,711	G/A	G27R	54,33	SHOX	SNV	45	64	54	0	56	20	S
	17	7,577,114	C/T	C275Y	87,67	TP53	SNV	151	11	101	0	82	68	S
7	1	41,404,534	A/-	NA	13,33	RP11-348A7.1	Indel	8	2*	30	0	0	17	O
	1	44,154,629	G/A	E634K	125,67	KDM4A	SNV	134	91	152	1	18	2	S
	1	198,225,575	G/C	E106Q	232	NEK7	SNV	331	156	209	1	19	42	S
	2	148,657,319	C/A	S127*	442,33	ACVR2A	SNV	760	303	264	0	50	94	S
	3	178,916,725	C/T	R38C	603,67	PIK3CA	SNV	750	459	602	0	54	77	S
	3	184,293,701	G/A	A314T	60,33	EPHB3	SNV	59	49	73	2	16	1	S
	5	112,175,423	C/T	Q1378*	330	APC	SNV	403	228	359	0	68	91	S
	7	98,565,256	T/A	Y2476N	169,67	TRRAP	SNV	260	64	185	0	75	88	S
	8	3,432,515	T/G	E433D	148	CSMD1	SNV	256	67	121	0	25	86	S
	10	412,306	C/A	R726I	548,33	DIP2C	SNV	616	414	615	0	21	55	S
	10	8,115,818	C/G	N388K	353,67	GATA3	SNV	446	314	301	0	6	54	S
	10	64,967,678	G/C	P1032A	454,67	JMJD1C	SNV	659	275	430	0	21	47	S
	11	116,719,906	G/A	S983F	45	SIK3	SNV	41	37	57	0	22	2	S
	11	117,392,003	G/A	T412M	107,33	DSCAML1	SNV	127	117	78	0	11	26	S
	13	110,804,709	C/T	A1634T	141	COL4A1	SNV	122	123	178	1	8	42	S
	16	68,842,430	C/A	P87Q	207,67	CDH1	SNV	275	126	222	0	6	17	S
	4	153,247,288	C/A	R505L	583,33	FBXW7	SNV	783	382	585	0	25	46	S
	12	25,398,284	C/A	G12V	176,33	KRAS	SNV	294	111	124	1	13	28	S
	17	7,578,406	C/T	R175H	25	TP53	SNV	39	27	9	0	26	67	S
8	1	9,793,472	T/C	E795G	153,67	CLSTN1	SNV	186	161	114	0	14	46	S
	1	12,164,486	G/A	E107K	430,67	TNFRSF8	SNV	453	474	365	1	9	38	S
	1	165,876,941	C/T	Q223*	128	UCK2	SNV	157	113	114	0	13	57	S
	2	33,482,484	A/T	K441N	344,67	LTBP1	SNV	389	341	304	0	18	34	S
	2	121,708,928	G/A	V122M	48,33	GLI2	SNV	57	33	55	0	6	29	S
	2	171,884,902	C/T	E347K	267,67	TLK1	SNV	345	323	135	1	6	22	S
	2	179,547,627	A/G	V10647A	110,67	TTN	SNV	168	77	87	0	14	34	S
	2	179,622,344	C/A	G3489W	305	TTN	SNV	344	299	272	4	12	48	S
	2	215,593,455	G/A	S131L	597	BARD1	SNV	639	555	597	1	12	38	S
	3	1,444,098	A/C	I900L	305,67	CNTN6	SNV	429	231	257	0	38	94	S
	3	25,659,978	C/A	L1089F	140,67	TOP2B	SNV	191	136	95	1	3	16	S
	3	30,732,969	C/T	R528C	49	TGFBR2	SNV	77	34	36	0	29	89	S
	3	171,455,712	C/T	V44M	134,33	PLD1	SNV	208	77	118	0	10	26	S
	5	112,175,675	AG/-	NA	288,33	APC	Indel	382	257	226	0	11	54	S
	7	6,038,819	C/T	G209R	590,67	PMS2	SNV	611	612	549	1	12	38	S
	8	48,715,930	G/A	R3286W	173	PRKDC	SNV	224	173	122	1	6	39	S
	8	121,282,371	C/A	S1057R	190,67	COL14A1	SNV	303	120	149	1	8	31	S
	9	94,486,903	C/T	D485N	156,67	ROR2	SNV	196	139	135	1	24	47	S
	10	81,006,67	C/T	S214L	38,67	GATA3	SNV	53	30	33	2	7	39	S
	10	89,692,905	G/T	R130L	234	PTEN	SNV	281	212	209	0	14	46	S
	11	108,128,322	A/C	N789H	147	ATM	SNV	149	125	167	1	28	54	S
	12	52,374,912	A/G	E195G	352	ACVR1B	SNV	352	298	406	1	35	69	S
	12	53,686,729	G/A	V2046M	52	ESPL1	SNV	46	47	63	0	15	38	S
	19	15,276,227	C/T	G110E	125,67	NOTCH3	SNV	164	81	132	1	26	42	S
	X	44,920,568	G/A	NA	55	KDM6A	SNV	63	47	55	3	19	0	T
	X	71,813,093	C/A	G1052V	264,67	PHKA1	SNV	324	265	205	1	15	72	S
	7	140,453,136	A/T	V28E	422	BRAF	SNV	474	395	397	0	6	35	S
	10	89,692,904	C/T	R130*	233	PTEN	SNV	278	212	209	0	16	45	S

Overview of all somatic mutations identified by SOLiD sequencing. Unique mutations are highlighted in boldface. The allele frequency is the percentage of reads in which the mutant allele has been detected. SNV, single-nucleotide variant.

^*Low coverage for samples with a unique mutation (might have resulted in the inability to identify a mutation).

Table S2.

Sequencing statistics depicted separately for all samples from each patient

Sample	Total reads	% mapped	% uniquely mapped	% in target bases	Average coverage	Median coverage	% bases covered	% bases covered >20×
Subject01 - blood	55,725,994	87.65	82.13	55.57	231.69	174.00	94.55	80.71
Subject01 - organoid	40,839,236	87.72	81.78	58.99	180.51	117.00	93.65	76.53
Subject01 - tumor	89,556,002	88.03	82.11	51.25	344.62	230.00	95.82	85.74
Subject02 - blood	78,031,553	88.48	82.46	55.59	327.50	248.00	95.41	84.21
Subject02 - organoid	57,602,210	87.38	81.41	55.31	237.68	155.00	94.80	81.20
Subject02 - tumor	58,474,524	85.93	80.16	54.31	232.86	157.00	96.19	86.96
Subject03 - blood	60,117,496	87.51	82.05	55.64	249.91	190.00	94.68	81.40
Subject03 - organoid	55,752,800	88.87	83.60	55.66	235.65	160.00	94.75	80.97
Subject03 - tumor	72,388,430	86.26	80.84	55.52	296.03	248.00	96.07	87.57
Subject04 - blood	59,027,598	86.57	80.78	55.33	241.00	189.00	95.30	83.18
Subject04 - organoid	42,977,631	89.45	84.03	56.05	184.11	106.00	94.36	78.99
Subject04 - tumor	20,034,396	79.87	73.20	41.30	56.46	31.00	89.01	59.71
Subject05 - blood	58,658,789	87.12	81.22	53.58	233.41	181.00	94.89	81.71
Subject05 - organoid	45,661,474	88.40	82.37	54.46	187.51	144.00	94.87	80.89
Subject05 - tumor	51,527,869	88.07	82.24	56.14	217.18	157.00	94.57	81.14
Subject06 - blood	62,923,615	79.67	74.27	54.87	234.91	195.00	94.68	83.73
Subject06 - organoid	24,908,884	68.34	62.27	45.03	65.48	42.00	94.45	70.95
Subject06 - tumor	61,725,207	63.98	59.68	56.61	190.95	132.00	93.63	79.90
Subject07 - blood	66,097,552	82.22	76.95	58.13	269.70	226.00	95.15	85.49
Subject07 - organoid	65,525,831	79.34	74.36	59.96	266.39	201.00	94.99	84.86
Subject07 - tumor	41,859,372	76.29	71.42	60.34	164.69	132.00	94.54	82.40
Subject08 - blood	57,698,769	78.87	73.78	59.47	231.04	188.00	95.19	84.77
Subject08 - organoid	55,896,268	75.59	70.73	60.17	217.12	162.00	94.67	82.34
Subject08 - tumor	49,018,897	76.00	71.18	60.87	193.64	156.00	94.68	82.79

Table S3.

Percentage of tumor cells and necrosis of original tumor biopsies per patient

Sample	Tumor cells, %	Necrosis, %
01	80	10
02	90	10
03	80	40
04	90	10
05	70	10
06	90	50
07	80	50
08	90	Not available

Fig. S1.

Copy number plots for tumor biopsy and biopsy-derived organoid cultures of patients 01–08.

Discussion

These data show that organoid cultures from colorectal cancer can be established readily from biopsies of metastases. The culture success rate of 71% for colorectal cancer compares favorably with the take rate of 15–20% for organoid cultures of biopsies from prostate cancer described by Gao et al. (6). Sequencing data and copy number analysis were largely concordant between organoid cultures and tumor biopsies. The most important observation from our study, however, is that none of the 15 mutations that were found exclusively in either the tumor or organoid culture are in driver genes or genes amenable for drug targeting. Discordance, when present, could be attributed to biological factors such as sampling variation from a heterogeneous tumor or ongoing normal tumor evolution or to selection resulting from culture conditions. With current knowledge it is not possible to distinguish between these factors. Aside from biological processes, technical limitations can result in a misrepresentation of the actual concordance by impeding factors such as low coverage of a particular gene or sample. For instance, 6 of 15 unique mutations have a coverage of <25× of the gene in the matched sample (where the mutation was not identified), and for patient 04 and patient 06 overall coverage of the tumor sample and biopsy-derived culture, respectively, was very low (median coverage <50×) (Tables S1 and S2). Furthermore, the presence of stromal cells and tumor necrosis in biopsies (resulting in a lower tumor percentage) may have resulted in less accurate detection of copy number aberrations and somatic mutations in the tumor biopsy as compared with the cultures. A lower tumor percentage can result in a misrepresentation of allele frequency in a negative way, so that the frequency can fall under the limit of detection. The actual allele frequency of a mutation in tumor biopsies thus is sometimes higher than depicted in Table S1. A lower tumor percentage also affects copy number profiles by reducing the signal toward the diploid state, resulting in discrepancies in the intensity of red and blue shading between tumor biopsies and corresponding biopsy-derived cultures in Fig. 2. This effect could have contributed to the discordance in patient 3 (Fig. S1 and Table S3).

To conclude, organoid cultures from colorectal cancer can be established readily from biopsies of metastases with preservation of the mutational and copy number landscape. This result supports the translation of patient-derived tumor organoids to the clinic as an ex vivo test platform, with the potential to meet the dire need for a means to make more rational treatment decisions for the individual patient. Clinical trials to validate the predictive value of organoids have been initiated.