Identification of genes that are regulated transcriptionally by Myc in childhood tumors

Abstract

BACKGROUND

Amplification of the N-myc oncogene is associated with adverse outcomes in the common childhood tumor, neuroblastoma. Because the transforming properties of Myc are related to its ability to modulate gene expression, the authors used cDNA microarrays to identify potential Myc target genes.

METHODS

Expression levels of 4608 genes were analyzed in a series of neuroblastoma cell lines. Identical analyses were performed in a panel of medulloblastoma cell lines to identify c-Myc targets and to determine the extent to which N-Myc targets and c-Myc targets were shared. Comparisons were made between cell lines with high levels versus low levels of Myc protein expression.

RESULTS

Array analyses yielded 121 genes with increased expression levels (≥ 1.65-fold) and 9 genes with decreased expression levels in N-Myc-expressing versus nonexpressing cell lines. Many of these were newly identified targets of biologic interest. Fifty percent of the N-Myc targets (60 of 121) were mutual c-Myc targets. A significant correlation between the level of N-myc and selected target gene expression was demonstrated independently in 27 neuroblastoma tumor samples and in an N-myc-inducible cell line system.

CONCLUSIONS

A number of diverse pathways are modulated by N-Myc in neuroblastoma. Although, overall, there was significant correlation between myc and target transcript expression among cohorts of tumors, great variability in levels of target expression was seen among individual tumor samples, and this biologic heterogeneity in the levels of target gene expression may offer insight into differences in the clinical behavior of neuroblastoma and may prove to be of prognostic significance in the future. Cancer 2003;98:841–53. © 2003 American Cancer Society.

DOI 10.1002/cncr.11584

Members of the myc family of oncogenes, including N-myc, c-myc, and L-myc, have been implicated in the development of many human tumors.1, 2 Myc forms a heterodimer with Max and binds to E-box elements in promoter and/or enhancer regions of target genes to activate transcription.3–5 Alternatively, Myc also can repress transcription of genes.6–8 Although myc family members can function in place of one another, these genes have been conserved independently throughout vertebrate evolution, implying unique functions as well.9 The extent to which each myc family member shares target genes still is uncertain.

Amplification of the N-myc oncogene is associated with an adverse outcome in the common childhood tumor neuroblastoma.9–11 The oncogenic potential of the N-myc gene product has been established in the laboratory setting using a number of in vitro and in vivo transformation assays.12–15 These observations indicate that the N-Myc protein plays a direct role in progression of neuroblastoma. Curiously, an association between N-myc expression and clinical outcome has been difficult to establish. Although some studies have correlated outcome with myc expression, other studies have reported a lack of correlation between N-myc mRNA/protein expression and clinical outcome in patients with tumors that lack N-myc gene amplification.16 These observations have suggested that the adverse outcome associated with N-myc amplification may not be explained simply by N-myc expression alone.

The ability of Myc to activate or repress target gene expression is influenced by many factors. The relative levels of binding partners, such as TRRAP, Miz-1, and Bin1, may influence target gene expression.3, 17, 18 Furthermore, the chromatin configuration of a potential Myc target gene and the availability of additional transcription factors also may affect the ability of Myc to modulate gene expression. It is our hypothesis that these factors contribute to the inability to demonstrate consistently a correlation between N-myc RNA/protein expression and clinical outcome. In some tumors, certain target genes may be inaccessible despite high levels of N-Myc, whereas easy activation may occur in others. In addition, the level of N-Myc protein, although relatively high, may not achieve a threshold necessary to activate target gene expression.

In an effort to define new prognostic markers in patients with neuroblastoma and to provide further insight into the biologic mechanisms of Myc-mediated oncogenesis, we used cDNA arrays to identify potential Myc target genes comprehensively. We also determined the extent to which N-Myc and c-Myc modulate common pathways by studying another childhood tumor, medulloblastoma. Expression of c-myc is correlated with outcome in patients with this common brain tumor.19–21 Our results indicate that a number of diverse pathways are modulated by Myc and that c-Myc and N-Myc share large numbers of mutual targets. Our results also suggest that distinct targets of N-Myc and c-Myc may exist, or, more probably, that the ability of Myc to modulate the expression of its target genes is influenced strongly by the cell type and environment. Although, overall, there was a significant correlation between myc and target transcript expression among cohorts of tumors, great variability in levels of target expression was seen among individual tumors. This variability in target expression patterns may offer an explanation for some of the clinical heterogeneity that is seen in patients with neuroblastoma and may prove to be of prognostic significance in the future.

MATERIALS AND METHODS

Cell Lines and Media

The cell lines, D283 MED, Daoy, and D425 MED all were derived from childhood posterior fossa medulloblastomas. The PFSK cell line was derived from a hemispheric peripheral neuroectodermal tumor (PNET). The characteristics and growth conditions of these cell lines have been described previously.19 NGP and NLF cell lines were derived from neuroblastomas with amplification copy numbers of 150 and 30 of the N-myc oncogene, respectively. SK-N-RA, SK-N-SH, and LAN-6 cell lines were derived from neuroblastomas with a single copy of N-myc. All neuroblastoma cell lines were grown in RPMI 1640 media supplemented with 10% fetal calf serum, 2 mM glutamine, 50 U/mL penicillin, and 50 μg/mL streptomycin. The N-myc-inducible SHEP-21N cell line (kindly provided by Manfred Schwab, German Cancer Research Center, Heidelberg, Germany) was maintained in RPMI 1640 media supplemented with 10% fetal calf serum, 100 U/mL penicillin, and 100 μg/mL streptomycin. N-myc expression was extinguished by the addition of 1 μg/mL tetracycline (Fluka, Ronkonkoma, NY) to the cell culture media.

Western Blot Analysis

Western blot analyses were performed as described previously.22 Monoclonal antibodies to N-Myc and c-Myc were used at a dilution of 1:1000 (Oncogene Research, San Diego, CA) followed by incubation with goat antimouse secondary antibody coupled to horseradish peroxidase (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) at a dilution of 1:5000.

RNA Isolation, Probe Preparation, and Hybridization

RNA purification, cDNA synthesis, and probe hybridization were performed according to protocols described previously.23 cDNA was synthesized with indodicarbocyanine (Cy5)-deoxycytidine triphosphate (dCTP) (red) labeling of Myc-expressing cell lines (NGP, NLF, LAN-6, D425 MED, and D283 MED) and indocarbocyanine (Cy3)-dCTP (green) labeling of nonexpressing/low-expressing cell lines (SK-N-RA, SK-N-SH, PFSK, and Daoy). Differentially labeled targets from Myc-expressing and nonexpressing cell lines were cohybridized to glass slides containing 4608 separate cDNAs spotted in duplicate (Gen III; Molecular Dynamics, Sunnyvale, CA). Expression ratios were quantified with ArrayVision software. Signals were normalized by adjusting for the relative fluorescence from each channel (Myc expressor, Cy5; Myc-nonexpressor, Cy3).

Target Gene Selection

Transcript expression was compared between cell lines with high levels of Myc protein and cell lines with low or absent Myc protein expression: NGP versus SK-N-RA, NLF versus SK-N-SH, LAN-6 versus SK-N-SH, D425 MED versus Daoy, D425 MED versus PFSK, and D283 MED versus Daoy. The average expression ratios for genes spotted in duplicate were determined for all comparisons. Genes selected as potential targets were those with a ≥ 1.65-fold difference in expression in 2 of 3 comparisons between Myc-expressing cell lines and nonexpressing cell lines (e.g., NGP vs. SK-N-RA and NLF vs. SK-N-SH). A 1.65-fold difference in expression was used as a cut-off level, because differences less than this may be attributed to experimental variation itself.24 The requirement for two of three cell line comparisons was chosen to allow for the biologic heterogeneity among the individual cell lines.

Real-Time Polymerase Chain Reaction

Two real-time polymerase chain reaction (PCR) methods were used for the validation of target expression. Relative transcript levels in tumor cell lines (NGP, NLF, SK-N-RA, and SK-N-SH) and SHEP-21N cells were determined by a comparative C_T method.25 The average cycle of threshold (ΔC_T) values from triplicate experiments of target expression were normalized to the ΔC_T value of β2-microglobulin (β2-MG) by subtracting ΔC_{T, β2-MG} from ΔC_{T, target}. To calculate relative transcript levels of target genes, the ΔΔC_T value (which is equal to ΔC_T,target−ΔC_T,β2-MG) of SK-N-SH was subtracted from the ΔΔC_T value of SK-N-RA, NLF or NGP. In SK-N-SH, the N-myc transcript level was below the detectable limit, and the ΔΔC_T value of SK-N-RA was used. The relative transcript level was determined by evaluating the following expression: 2^{− ΔΔCT}. To determine the relative transcript levels of target genes in the SHEP-21N cell line, the ΔΔC_T value of time 0 point was subtracted from the ΔΔC_T values of the 24 hour and 2-week time points.

Transcript numbers were measured in tumor samples using standard methods to allow comparison with previous studies.23 All reactions were run in triplicate, and target transcripts were normalized to β2-MG and quantified based on standard curves. Primer sequences for N-myc and the targets are shown in Table 1.

Table 1. Primer Sequences for N-myc and Target Genes

Target	Forward primer	Reverse primer
FBL: fibrillarin; PTMA: prothymosin α; HSPCB: 90-kilodalton heat-shock protein; p68: RNA helicase p68; ID2: inhibitor of DNA binding 2; STMN1: stathmin 1; POLD2: DNA polymerase δ subunit 2; PCNA: proliferating cell nuclear antigen; HDAC2: histone deacetylase 2; CNTFR: ciliary neurotrophic factor receptor; B2-MG: β2-microglobulin.
N-myc	5′-CAAGGCTGTCACCACATTCAC-3′	5′-TCTTTAGCAACTGCTGCTGTC-3′
FBL	5′-GCATCCTCGATCACAGGAATG-3′	5′-AAGCTAGCAGCAGCAATCCTG-3′
PTMA	5′-GAAGAAGAGGAAGAAGGTGGG-3′	5′-CCTCCCCTGTTGCAAATTCTC-3′
HSPCB	5′-ATGCCTGAGGAAGTGCACCAT-3′	5′-CCAGACTTGGCAATGGTTCCC-3′
p68	5′-ATGTCGGGTTATTCGAGTGACCG-3′	5′-CCATGACATTTGCAGGGAAATTGG-3′
ID2	5′-GGTCCGTTAGGAAAAACAGCC-3′	5′-GGGAATTCAGAAGCCTGCAAG-3′
STMN1	5′-AAGGATCTTTCCCTGGAGGA-3′	5′-TGTGCCTCTCGGTTCTCTTT-3′
POLD2	5′-ATGTTTTCTGAGCAGGCTGCC-3′	5′-TGGGGGAGCAGGTTGTGCTC-3′
PCNA	5′-AGGGCTCCATCCTCAAGAAGG-3′	5′-TGGTGCTTCAAATACTAGCGC-3′
HDAC2	5′-CGTGTAATGACGGTATCATTCC-3′	5′-ACCAGATAATGAGTCTGCACC-3′
CNTER	5′-CCACCTACATTCCCAACACC-3′	5′-GGGCTACCACATTTTCTGGA-3′
β2-MG	5′-ACCCCCACTGAAAAAGATGA-3′	5′-CTCAGATACATCAAACATGG-3′

Tumor Samples

Total RNA harvested from 27 neuroblastoma tumor samples was obtained from 2 separate tissue banks maintained by the Children's Cancer Group. Specifically, the samples consisted of RNA from seven International Neuroblastoma Staging System Stage III tumors; seven Stage IV, N-myc-amplified tumors; and seven Stage III and six Stage IV nonamplified tumors.

Time Course Experiments in SHEP-21N Cells for Target Gene Validation

SHEP-21N cells were plated at 3 × 10⁵ cells per T75 flask. N-myc expression then was extinguished with the addition of tetracycline, as described above. After removal of tetracycline to induce N-myc expression (time 0), cells were harvested at 24-hour and 2-week time points. Total RNA then was purified, and real-time PCR using a comparative C_T method was performed, as described above.

Statistical Analysis

The N-myc expression level in tumor samples was log-transformed to improve normality. The difference in median N-myc transcript number between Stage III tumors and Stage IV tumors (samples were balanced with respect to gene amplification status) was determined. Significance values were calculated using the Wilcoxon Mann–Whitney test. The difference in median N-myc transcript number between all amplified and nonamplified tumors also was determined using this approach. The Pearson correlation coefficient was computed to test the strength of the association between N-myc and individual target transcript expression in tumor samples. The squared correlation coefficient (Δ²) was used as an estimate of the fraction of the variability in the target gene expression attributed to N-myc expression.

RESULTS

Identification of Myc Targets using cDNA Arrays

Myc protein expression was determined by Western blot analysis in tumor cell lines. We compared the gene expression profiles of several Myc-expressing versus nonexpressing cell lines, derived from human tumors, using cDNA arrays (Fig. 1). The analysis identified 121 potential N-Myc activation targets and 9 repression targets in N-Myc-expressing cell lines (Tables 2, 3). Using the same selection criteria, 721 c-Myc activation targets and 47 c-Myc repression targets were identified in a panel of medulloblastoma/PNET cell lines (data not shown). Sixty of the 121 potential N-Myc target genes (50%) were up-regulated in c-Myc-expressing cell lines (Tables 2, 3).

Figure 1. cDNA array analysis of (A) neuroblastoma and (B) medulloblastoma cell lines. A hierarchical cluster analysis of the 4608 array elements discriminated between cell lines with high levels (e.g., NGP, NLF, LAN-6, 425, 283) and low levels (e.g., SK-N-RA, SK-N-SH, Daoy, PFSK) of Myc protein expression. Individual genes are shown in rows, and cell lines are shown in columns. Genes that were overexpressed in the Myc-expressing cell lines relative to the low-expressing cell lines are depicted in red, genes that were underexpressed in the Myc-expressing cell lines relative to the low-expressing cell lines are depicted in green, and genes that were expressed equivalently in both cell populations are depicted in black. Duplicate hybridizations were performed for all comparisons, and the average values of expression for individual genes are shown.

Table 2. Genes Activated by Myc

Target gene	UniGene no.70	GenBank accession no.71	Expression ratioa						C-Myc target?
			Neuroblastoma cell lines			Medulloblastoma cell lines
			NGP/SKNRA	NLF/SKNSH	LAN-6/SKNSH	425/324	425/PFSK	283/324
a Expression ratios represent the relative change in the expression of individual genes in comparisons of high versus low Myc-expressing cell lines The average value for a duplicate hybridization is shown for each gene. b Replicate clones of individual genes.
Protein synthesis
RPL4	286	NM_000968	2.00	1.61	1.81	4.30	2.32	4.15	Yes
GARS	75280	NM_002047	1.92	2.52	1.14	3.02	1.86	1.45	Yes
RPS5	76194	NM_001009	2.10	1.96	0.86	3.56	1.86	3.13	Yes
RPS5b	76194	NM_001009	1.73	2.02	1.96	3.81	1.87	3.07	Yes
RPL27	111611	NM_000988	1.68	2.14	2.64	15.36	3.97	5.07	Yes
RPL8	178551	NM_000973	1.34	1.92	2.16	7.23	2.62	3.62	Yes
RPL11	179943	NM_000975	1.37	1.94	1.81	6.81	2.70	3.22	Yes
RPL13	180842	BC000851	3.03	0.80	2.04	7.88	3.85	3.63	Yes
RPLP2	351937	NM_001004	2.26	1.14	1.84	6.39	1.59	4.13	Yes
Protein folding/turnover
CCT3	1708	NM_005998	3.33	2.14	1.72	4.99	1.53	3.06	Yes
HSPCB	74335	M16660	2.34	2.42	1.11	10.27	2.65	8.00	Yes
UBB	183842	NM_018955	1.40	1.89	1.77	3.93	1.80	2.52	Yes
RBX1	279919	BC001466	1.27	2.89	2.30	2.77	1.61	1.09	No
HSPCA	289088	AF028832	0.74	2.78	1.83	4.38	1.07	0.86	No
Metabolism
HMT-1	44592	AB019038	1.59	1.74	1.66	3.34	1.45	2.38	Yes
DHCR24	75616	NM_014762	2.07	0.75	2.55	0.70	0.93	0.86	No
PTGS1	88474	M59979	2.81	2.11	1.59	2.57	1.74	1.33	Yes
DGKI	242947	AF061936	1.73	1.66	1.35	2.03	1.61	1.88	Yes
GAPD	169476	AF261085	3.74	2.40	2.07	1.12	0.81	0.46	No
GAPDb	169476	AF261085	1.69	0.59	1.67	0.49	0.71	0.67	No
GAPDb	169476	AF261085	4.19	2.59	3.67	1.13	0.85	0.59	No
GAPDb	169476	AF261085	4.28	2.68	2.70	1.12	0.83	0.51	No
GAPDb	169476	AF261085	4.06	2.71	2.83	1.09	0.82	0.50	No
Nucleotide/DNA synthesis, repair, recombination
ATP5G3	429	NM_001689	2.02	2.29	1.52	6.46	1.87	1.77	Yes
UNG2	3041	M87499	1.69	2.45	1.20	2.80	1.47	1.16	No
FEN1	4756	X76771	1.44	2.36	1.65	1.83	1.59	1.24	No
RFC4	35120	NM_002916	1.66	2.21	1.72	1.36	1.10	1.09	No
ATP synthetase	73581	AA178987	1.80	1.76	1.06	3.29	1.33	1.88	Yes
APX	73722	D13370	2.18	2.04	1.31	2.63	1.99	2.35	Yes
POLD2	74598	HSU21090	1.95	1.85	2.02	4.75	4.46	1.79	Yes
POLD2b	74598	HSU21090	1.78	2.07	1.46	4.76	3.99	1.58	Yes
PCNA	78996	NM_002592	2.34	2.58	2.07	3.02	1.58	1.27	No
PCNAb	78996	NM_002592	1.69	2.48	1.55	2.18	1.53	1.21	No
PCNAb	78996	NM_002592	2.34	2.24	1.12	2.28	1.37	1.13	No
ATP5F1	81634	BC005366	1.66	1.77	0.92	3.48	1.38	1.37	No
TYMS	82962	NM_001071	1.05	2.06	1.66	2.41	1.21	1.07	No
ATP5G2	89399	NM_005176	1.88	1.61	1.82	2.85	1.78	3.18	Yes
PRKDC	155637	HSU47077	2.03	1.76	1.02	2.42	1.04	1.60	No
Transcription
HDAC2	3352	NM_001527	2.08	2.13	1.18	2.81	1.54	1.92	Yes
POLR2E	24301	NM_002695	1.82	1.77	1.91	4.24	3.14	1.77	Yes
P65NFKB	75569	M62399	1.74	1.76	1.04	2.76	1.35	1.90	Yes
ID2	180919	D13891	2.65	3.59	1.25	1.08	0.85	0.72	No
ATF4	181243	NM_001675	2.10	2.34	2.61	5.75	5.54	4.72	Yes
ELF3	166096	AF017307	1.82	1.66	1.47	2.96	1.60	2.14	Yes
SOX8 SRY	243678	AF164104	2.99	1.79	1.07	9.04	4.00	8.16	Yes
RNA processing
SNRPC	1063	NM_003093	1.92	2.41	1.75	2.71	1.34	0.92	No
BAT1	55296	NM_004640	1.93	1.61	1.83	2.31	1.36	1.52	No
P68	76053	NM_004396	1.65	2.10	1.82	4.17	2.05	0.97	Yes
P68b	76053	NM_004396	1.82	1.66	2.73	2.74	1.90	1.76	Yes
SNRPB	83753	J04564	2.07	1.96	1.34	2.75	1.42	1.35	No
FBL	99853	NM_001436	2.66	2.07	1.10	6.35	2.86	3.78	Yes
HNRPA2B1	232400	D28877	1.82	2.62	1.73	8.61	2.25	0.96	Yes
Signal transduction
RAB5C	479	NM_004583	1.68	2.24	1.97	4.15	1.53	1.42	No
FGFR1	748	M34185	0.96	1.72	1.94	2.04	0.81	1.00	No
CD9	1244	M38690	1.33	3.43	2.46	1.19	1.33	0.68	No
PRKCSH	1432	NM_002743	1.80	3.10	0.92	3.61	1.69	1.30	Yes
WBP1	7709	BC010012	2.20	3.03	0.77	4.52	1.63	0.71	No
STAC	56045	NM_003149	4.21	2.00	1.23	1.03	1.51	1.08	No
CD14	75627	HUMCD14MCA	1.72	1.93	2.25	4.00	1.92	3.10	Yes
DBI	78888	NM_020548	2.18	2.34	0.70	19.96	12.85	1.09	Yes
PPP1CC	79081	NM_002710	1.57	2.34	1.72	5.98	2.37	5.69	Yes
HDGF	89525	NM_004494	1.94	1.74	1.43	1.84	1.29	1.46	No
PPP2R1A	173902	JO2902	2.03	1.36	1.68	1.47	1.19	1.72	No
NME2	275163	M36981	1.67	1.88	1.35	10.48	3.02	2.57	Yes
PTTG1	252587	AF095287	2.37	2.73	1.93	2.29	1.18	0.72	No
Cytoskeletal regulation
ARPC3	6895	NM_005719	1.58	1.89	1.88	2.11	1.28	0.59	No
TUBA1	75318	BC009238	2.14	3.25	2.18	3.84	1.03	0.78	No
PFN1	75721	HUMPROF	2.01	2.11	0.99	0.96	0.28	0.38	No
MYL6	77385	HUMMYLCC	1.81	1.79	2.03	3.63	1.19	0.63	No
STMN1	81915	J04991	1.50	2.49	2.48	6.16	3.27	1.98	Yes
STMN1b	81915	J04991	1.19	1.73	1.72	1.99	1.24	1.70	Yes
STMN1b	81915	J04991	1.48	2.70	2.89	7.16	4.83	1.71	Yes
STMN1b	81915	J04991	0.83	2.09	2.11	1.69	2.18	2.38	Yes
ARHGEF1	252280	NM_004706	1.66	2.58	1.48	7.46	2.92	1.48	Yes
TUBA3	272897	AF141347	2.92	2.62	2.99	3.95	0.98	0.76	No
TUBB-5	274398	AK001295	1.90	2.49	1.39	1.93	0.91	0.76	No
Proliferation, differentiation, apoptosis
MAF	30250	AF055376	1.84	1.93	2.42	3.35	2.27	2.55	Yes
PIG3	50649	NM_004881	1.81	1.86	0.81	0.96	0.96	0.71	No
CNTFR	194774	NM_001842	2.30	1.64	1.25	9.60	4.29	4.12	Yes
TNFAIP3	211600	NM_006290	2.19	1.41	1.98	1.67	1.17	0.85	No
PTMA	250655	M14630	2.13	2.84	1.16	6.30	3.70	2.85	Yes
Cell cycle control
SAM68	119537	NM_006559	1.94	1.80	1.34	2.45	1.81	1.71	Yes
CDKN1A	179665	L25610	1.18	1.82	2.71	0.93	0.75	0.62	No
CDKN1Ab	179665	L25610	1.13	1.73	1.79	0.69	0.75	0.60	No
CDC2	334562	Y00272	1.91	2.28	1.74	1.43	1.20	1.04	No
Miscellaneous
CHGB	2281	NM_001819	1.66	0.30	8.76	1.03	1.10	1.11	No
PIN	5120	HSU32944	1.46	2.00	2.02	2.92	1.72	0.80	Yes
PNN	44499	HSU77718	1.68	1.92	1.71	2.39	1.49	1.58	No
COX7A2	70312	NM_001865	1.90	1.90	0.97	3.39	1.63	0.91	No
COX7A2b	70312	NM_001865	1.83	1.77	0.91	3.12	1.44	0.99	No
ARF1	74571	M84326	1.58	1.73	2.00	4.64	1.83	2.33	Yes
BSG	74631	NM_001728	4.30	1.75	2.39	4.99	3.51	3.04	Yes
TXN	76136	AY004872	2.07	2.08	1.73	3.88	2.45	1.07	Yes
COX6C	351875	NM_004374	1.87	1.97	1.46	4.62	1.30	2.12	Yes
TFRC	77356	X01060	1.83	3.14	1.20	3.30	3.03	1.90	Yes
S100A4	81256	BC016300	1.43	2.11	1.78	1.00	0.92	0.58	No
CD53	82212	NM_000560	1.16	2.22	2.46	2.30	1.15	0.94	No
PFKP	99910	D25328	1.76	2.18	1.29	1.29	1.07	0.93	No
CLTA	104143	BC009201	1.28	1.77	1.74	3.14	1.41	1.22	No
SMG1	110613	AB061371	2.51	1.21	1.66	5.03	2.11	6.53	Yes
FTL	111334	M11147	2.14	1.39	1.70	1.08	1.56	1.06	No
H2AFZ	119192	M37583	2.05	2.56	1.33	6.57	2.01	1.43	Yes
H2AFZb	119192	M37583	1.75	2.88	1.76	5.41	1.72	1.27	Yes
AP2S1	119591	NM_004069	3.21	2.62	2.16	1.19	0.86	0.62	No
OPHN1	128824	HSJ001189	1.14	1.70	1.71	1.99	2.67	2.72	Yes
SFRS7	184167	NM_006276	1.72	1.88	0.84	2.35	1.18	1.25	No
BCRP1	268763	AF068235	1.48	1.74	1.72	2.26	1.42	1.07	No
LOC56993	285005	AB041906	1.22	1.86	1.69	2.52	1.54	1.68	Yes
Unknown
E2IG5	5243	NM_014367	2.68	2.89	1.34	1.58	1.96	2.02	Yes
FLJ11196	6166	AK002058	1.67	1.66	1.28	0.91	1.24	0.90	No
REA	7771	NM_007273	2.36	1.84	1.30	3.58	2.01	2.75	Yes
SMBP	8203	AF116347	3.94	2.68	1.91	1.15	0.98	0.70	No
PQBP1	30570	NM_005710	2.0	2.16	1.16	3.33	2.48	1.98	Yes
FLJ31235	49433	AK055797	2.03	1.29	2.24	3.34	2.37	3.43	Yes
DOK2	71215	AF034970	4.34	3.03	0.96	1.06	0.90	0.80	No
MGC5350	71331	NM_030920	1.69	2.80	1.22	2.00	1.06	1.28	No
LSM4 U6	76719	HSA238096	1.84	2.09	1.41	2.38	1.38	1.57	No
MEG3	112844	AB032607	2.31	0.98	4.50	0.74	0.78	0.62	No
NESP55	113368	AF105253	2.62	0.79	2.04	5.05	2.70	3.10	Yes
MLLT2	114765	NM_005935	2.81	2.46	1.73	1.35	1.08	0.76	No
ESTs	132168	AI401581	1.65	2.65	0.96	20.41	12.11	1.08	Yes
RAP1B	156764	BC000176	3.27	2.04	1.15	2.40	1.09	0.63	No
TROAP	171955	NM_005480	1.68	1.87	1.42	2.07	1.38	1.22	No
HSMNP1	179666	NM_018478	1.95	0.72	1.71	0.75	0.93	0.93	No
LOC51142	180859	NM_016139	2.29	2.64	1.11	6.36	2.50	1.18	Yes
LOC51142b	180859	NM_016139	1.77	2.79	1.79	5.94	2.42	1.12	Yes
PCCX2	199009	AB031230	3.69	2.03	1.09	1.51	1.00	1.14	No
ESTs	240833	AI080703	1.86	1.75	2.22	1.78	1.15	1.62	No
KIAA0788	246112	AB018331	2.26	2.77	1.62	2.18	1.44	1.03	No
ESTs	269157	R79837	1.11	1.79	1.85	2.47	2.98	1.92	Yes
MGC1346	273234	BC007321	1.15	1.89	1.68	1.46	1.46	1.13	No
MGC12992	278242	NM_032342	2.71	3.03	2.55	4.35	0.96	0.88	No
FLJ11827	288534	AK021889	2.82	1.12	1.81	3.01	1.58	2.93	Yes
DKFZP586A0522	288771	AK023693	1.11	1.79	1.98	1.15	0.91	1.12	No
DC50	324521	AF271779	1.67	1.67	0.89	2.69	1.54	0.62	No
FLJ30010	9585	AK054572	1.27	2.14	1.88	9.52	3.58	2.56	Yes
E2IG3	279923	NM_014366	2.12	1.74	0.72	4.56	2.30	2.39	Yes

Table 3. Genes Repressed by Myc

Target gene	UniGene no.70	GenBank accesion no.71	Expression ratioa						C-Myc target?
			Neuroblastoma cell lines			Medulloblastoma cell lines
			NGP/SKNRA	NLF/SKNSH	LAN-6/SKNSH	425/324	425/PFSK	283/324
a Expression ratios represent relative change in the expression of individual genes in comparisons of high versus low Myc-expressing cell lines. The average value for a duplicate hybridization is shown for each gene. b Replicate clones of individual genes.
IGF-2 (XM_006402)	251664	AA570502	0.51	0.36	0.68	0.52	0.17	0.74	Yes
TIMP-3	245188	H00393	0.19	0.59	0.91	1.70	3.92	1.38	No
TIMP-3b	245188	AA044130	0.19	0.54	0.66	1.37	1.95	1.30	No
Apolipoprotein CI	268571	H78683	31.42	0.18	0.58	1.01	0.48	0.84	No
TRAM-like protein	153954	N98550	0.59	0.54	1.73	0.71	0.79	1.00	No
α-PDGF receptor precursor	74615	AA041261	0.81	0.49	0.61	0.63	0.76	0.87	No
α-PDGF receptor precursorb	74615	AA043451	0.73	0.29	0.49	0.68	0.72	0.79	No
Homo sapiens 5T4 gene for 5T4 oncofetal antigen	82128	AA046300	0.83	0.47	0.37	0.67	0.87	0.60	No
Small inducible cytokine A	303649	AA047099	0.73	0.61	0.39	0.47	0.59	0.19	Yes
Disabled Drosophila homolog 1	8108	T73039	0.93	0.26	0.43	0.13	0.81	0.15	Yes
H. transglutaminase mRNA	8265	AA488308	1.03	0.14	0.50	0.59	0.83	0.73	No

Validation of Target Genes

To validate the microarray data initially, we used quantitative PCR to measure transcript levels for N-myc and selected target genes in the cell lines used for expression profiling. In selecting targets for validation, we chose a combination of previously reported Myc targets (e.g., 90-kilodalton heat-shock protein [HSPCB], inhibitor of DNA binding 2 [ID2], prothymosin α [PTMA], proliferating cell nuclear antigen [PCNA], and fibrillarin [FBL]) as well as future targets of interest (e.g., stathmin 1 [STMN1], ciliary neurotrophic factor receptor [CNTFR], histone deacetylase 2 [HDAC2], and RNA helicase p68 [p68]). In all cases, higher levels of target transcript expression levels were seen in the N-myc-amplified cell lines compared with the single-copy cell lines, validating the microarray expression data (Table 4).

Table 4. Relative Increase in Target Gene Expression Levelsa

Cell line	N-myc	PCNA	HSPCB	POLD2	p68	ID2	FBL	PTMA	HDAC2	STMN1	CNTFR
BD: below detectable limit; PCNA: proliferating cell nuclear antigen; HSPCB: 90-kilodalton heat-shock protein; POLD2: DNA polymerase δ subunit 2; p68: RNA helicase p68; ID2: inhibitor of DNA binding 2; FBL: fibrillarin; PTMA: prothymosin α; HDAC2: histone deacetylase 2; STMN1: stathmin 1; CNTFR: ciliary neurotrophic factor receptor. a The relative increase was calculated based on the transcript copy level in the SK-N-SH cell line (except for N-myc, for which the copy level in SK-N-RA was used).
Neuroblastoma tumor cell line
SK-N-SH (single copy)	BD	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00
SK-N-RA (single copy)	1.00	1.73	1.40	2.33	1.46	0.59	1.49	1.44	0.60	2.71	5.98
NLF (amplified)	142.02	26.17	15.03	20.39	8.11	6.41	33.36	29.24	19.87	30.48	42.52
NGP (amplified)	5293.48	10.93	13.93	31.12	6.77	2.83	67.65	15.67	14.08	21.41	18.00
SHEP-21N cell line
Tetracycline on	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	1.00	—
Tetracycline off, 24 hs	54.57	1.21	1.67	1.12	1.33	1.33	1.46	1.47	3.51	1.51	—
Tetracycline off, 2 ws	123.64	3.58	3.03	1.02	1.39	2.30	2.46	3.43	2.83	1.95	—

We also verified our microarray data in tumor samples. Similar ranges of N-myc transcripts were seen for Stage III tumors and Stage IV tumors, with Stage III tumors exhibiting more variability (Fig. 2A). There was no evidence of a difference in median N-myc transcript expression between Stage III tumor samples and Stage IV tumor samples that were balanced with respect to amplification status (P = 0.98). A highly significant difference in the median N-myc transcript number was observed between amplified tumors and nonamplified tumors (P = 0.0003), as expected, with much higher N-myc mRNA expression in amplified tumors (Fig. 2B). A significant overall correlation (log-transformed values) between N-myc and selected target transcript numbers was seen in the 27 tumor samples (Table 5). Although the transcript expression of the targets was correlated significantly with the expression of N-myc, it also appeared that for most of the targets, < 50% of the variability in expression was attributed to N-myc expression, suggesting that other factors that are unique to individual tumors may regulate target transcript expression. A subset analysis of the initial 20 tumor samples (10 N-myc-amplified samples and 10 nonamplified samples) from a single tumor bank also was carried out and showed both the correlation between N-myc and selected target gene transcript number, normalized to β2-MG, and also the variability in expression patterns among individual samples (Fig. 3).

Figure 2. Correlation between N-myc RNA expression, tumor stage, and amplification status in 27 neuroblastoma samples from 2 tumor banks. (A) Box plot showing the ranges of the log of N-myc transcript copy number (normalized to β2-microglobulin) for Stage III and Stage IV tumor samples. (B) Box plot showing the correlation between the log of N-myc transcript copy number and N-myc amplification status.

Table 5. Correlation between N-myc and Targets in Tumor Samples

Target	Correlation	R^2	P value
R^2: correlation coefficient; FBL: fibrillarin; PCNA: proliferating cell nuclear antigen; CNTFR: ciliary neurotrophic factor receptor; TIMP3: tissue inhibitor of metalloproteinase-3; STMN1: stathmin 1.
FBL	0.788536	0.621789	0.000001
PCNA	0.436081	0.190167	0.022972
CNTFR	0.729435	0.532075	0.000016
TIMP3	0.398687	0.158951	0.039407
STMN1	0.419498	0.175979	0.029388

Figure 3. Variability in target gene expression in tumor samples. Individual plots show the range and mean transcript copy number (log-transformed values) for (A) N-myc and selected target genes ([B] fibrillarin [FBL]; [C] proliferating cell nuclear antigen [PCNA]; and [D] ciliary neurotrophic factor receptor [CNTFR]) in 10 amplified neuroblastoma samples and 10 nonamplified neuroblastoma samples from a single tumor bank. Short horizontal bars indicate the mean level of transcript expression.

Target transcript expression levels also were determined at serial time points after the removal of tetracycline using the myc-inducible SHEP-21N cell line system12 (Fig. 4A). An increase of up to 3.6-fold in the transcript expression levels of the selected targets was observed, closely paralleling the expression of N-myc for the majority of tested targets (Table 4, Fig. 4B).

Figure 4. Target transcript expression after tetracycline-dependent expression of N-Myc in SHEP-21N cells. (A) Immunoblots of SHEP-21N cell lines. Top: N-Myc expression level detected by immunoblotting at 0 hours, 24 hours, and 2 weeks after tetracycline removal in SHEP-21N cells. Bottom: An equivalent amount of total cellular protein (10 μg) was analyzed by immunoblotting with anti-actin antibody as a control for protein loading. (B) Time course showing the relative increase in target transcript expression with the induction of N-myc transcript and protein expression (see also Table 4). Relative transcript levels of selected target genes (normalized to β2-microglobulin) were determined by real-time polymerase chain reaction analysis using a comparative C_T method. Relative transcript expression levels were determined after the withdrawal of tetracycline, at the indicated time points (0, 24 hours and 2 weeks). Black bars: + tetracycline; gray bars: − tetracycline for 24 hours; white bars: − tetracycline for 2 weeks; PCNA: proliferating cell nuclear antigen; HSPCB: 90-kilodalton heat-shock protein; POLD2: DNA polymerase δ subunit 2; p68: RNA helicase p68; ID2: inhibitor of DNA binding 2; FBL: fibrillarin; PTMA: prothymosin α; HDAC2: histone deacetylase 2; STMN 1: stathmin 1.

DISCUSSION

Neuroblastoma is a clinically heterogeneous childhood malignancy with vastly different treatment outcomes that depend on clinical, genetic, and biologic features of the disease.26–32 These differences have prompted a search for variables that more accurately predict outcome and direct optimal treatment assignment for patients with neuroblastoma.

Amplification of the N-myc protooncogene, which is observed in approximately 25% of neuroblastomas, has been the key genetic feature that has been used to stratify patients into risk groups for treatment.10, 11, 33–35 It is noteworthy, however, that some tumors show increased amounts of N-myc mRNA or protein expression in the absence of gene amplification, with controversial prognostic implications.16, 36 This finding suggests that variables other than N-myc expression also may be important in predicting prognosis. Because Myc proteins are transcription factors, we hypothesized that the activation of target genes may serve as a better reflection of the biologic activity of Myc and that individual differences in downstream effectors may explain some of the clinical heterogeneity characteristically observed in neuroblastoma.

Previously published reports of both N-Myc targets and c-Myc targets come from a variety of experimental models, the majority of which have been in vitro cell line systems.12, 37–41 Our putative targets, like those identified by Schuldiner and Benvenisty, were determined exclusively in human myc-induced tumor cell lines.42 We used a number of approaches, including tumor samples, to define potential targets with the rationale that if the same differentially regulated genes were identified within a number of different cellular backgrounds, then it is likely that their representative pathways play a fundamental role in tumorigenesis.

Many of our findings were consistent with previous reports. We found that Myc activates more targets than it represses38, 40 and that ≈ 50% of N-Myc targets were shared by c-Myc.37 This was not surprising given the functional redundancy of c-myc and N-myc.9 The number of c-Myc targets greatly exceeded the number of N-Myc targets in our analysis. The reasons are unclear but may be secondary to differences between the medulloblastoma and neuroblastoma cell lines or to the differences between the transactivation potential of c-Myc and N-Myc. It has been shown previously that the transforming potential of c-Myc exceeds that of N-Myc.43

The Myc target genes identified in the current study are involved in diverse pathways. They primarily include genes involved in cellular growth and metabolism, cell proliferation, and cell cycle regulation. Although there have been several reports of c-Myc targets, few comprehensive analyses of N-Myc targets have been completed to date.37, 42, 44 Some of the first N-Myc targets to be identified were PTMA, ornithine decarboxylase, and ID2.12, 45 Our chips contained PTMA and ID2, and both genes were up-regulated in Myc-expressing cell lines. Two previous reports by Lasorella et al. also showed that ID2 is an effector of N-Myc, whereas two more recent reports have failed to demonstrate an association between ID2 transcript levels and either N-myc amplification status or mRNA expression.45–48 We cannot explain the differences in these findings; however, in our PCR assays in neuroblastoma cell lines, we normalized N-myc to β2 MG, whereas Wang et al. used 18S RNase as an internal control when examining the correlation between N-myc and ID2 transcript expression.47 Furthermore, it has been shown that ID2 expression levels vary considerably, depending on the cell line growth conditions.48FID2 was defined as a target in our array analyses in neuroblastoma cell lines in which N-Myc protein was expressed differentially. This was verified in a cell line system when N-Myc protein was induced maximally. It is noteworthy that Wang et al. showed up-regulation of ID2 protein with increasing N-myc expression in an inducible cell line system.47

Consistent with previous reports, a large percentage of the differentially regulated genes in our analysis modulate cellular growth and metabolism, with roles in ribosome biogenesis, translational regulation, protein synthesis and processing and glycolysis. The recent finding that c-Myc activates RNA polymerase III transcription directly may explain this finding.49 We showed induction of eight ribosomal protein genes, the RNA helicase, p68, genes encoding two ribonucleoproteins, and FBL. Genes involved in protein folding and degradation, HSPCB and ubiquitin B (UBB), were induced as well as the gene encoding the glycolytic enzyme, glyceraldehyde-3-phosphate dehydrogenase (GAPD). The vast majority of these genes also were defined as c-Myc targets. These findings are in accord with the known role of Myc in regulating the increase in cell mass and size that is likely to be necessary for cell cycle progression and division.50

In addition to these common findings, we have identified several new potential target genes of interest. Future confirmation will be needed, however, to determine if, in fact, these are direct targets of N-Myc. Ciliary neurotrophic factor (CNTFR) is one example. CNTF affects the survival and differentiation of several classes of neurons through binding to its receptor, CNTFR.51–53 It has been shown that the α subunit of CNTFR, which confers specificity to the receptor complex, is expressed in several neuroblastoma cell lines and activates known signaling pathways in these cell lines.54 The detailed molecular mechanisms through which the CNTFR complex influences cell survival and differentiation currently are unknown.

Another newly identified N-Myc target is stathmin. STMN1 is an abundant cytoplasmic phosphoprotein that plays an important role in controlling cellular proliferation by regulating the dynamics of the microtubules during assembly of the mitotic spindle.55 STMN1 is expressed widely in a variety of human tumors,56 and it has been shown that inhibition of STMN expression in leukemia cell lines inhibits cell growth.57 It is noteworthy that, STMN1 was used recently for molecular-based therapeutic approaches, which makes this a target of interest.58, 59

HDAC2 also is an interesting target: several studies have suggested the possible role of histone deacetylases in human malignancies.60 Increasing evidence of the involvement of histone deacetylases in transcriptional repression has been reported. Because many histone deacetylases function as transcriptional corepressors and/or play roles in chromatin remodeling, it also will be interesting to investigate whether HDAC2 is involved in N-myc autoregulatory mechanisms.

We also identified repression targets of potential interest, including tissue inhibitor of metalloproteinase 3 (TIMP-3). TIMPs play complex and sometimes paradoxical roles in regulating the extracellular matrix, tumor growth, invasiveness, and metastasis.61, 62 Among the members of the TIMP family, TIMP-3 has unique proapoptotic functions.63, 64 Recently, studies showed that TIMP-3 induced apoptosis through death receptor–mediated mechanisms.65 In addition, it has been shown that adenoviral transfer of TIMP-3 into cells lines decreases tumor invasiveness,63 and a loss of TIMP-3 function has been implicated in tumorigenesis.66 The therapeutic potential of modulation of TIMP expression also is being explored currently.67

Despite challenges in identifying Myc target genes,4, 68, 69 we have shown that cDNA arrays are a useful tool for identifying potential Myc targets, and it is noteworthy that we have identified several new potential targets of interest. We envision that some of these targets may be of prognostic markers offering a reflection of the biologic effect of Myc, and our objective is to verify their expression levels in larger cohorts of patient samples in the future. It also is hoped that newly identified targets and effectors farther downstream may be modulated to provide new therapeutic approaches.

Acknowledgements

The authors thank the Microarray Core Facility at the Huntsman Cancer Institute and the Children's Oncology Group Neuroblastoma Reference Laboratory for providing tumor samples, Dr. Manfred Schwab for providing the SHEP-21N cell line, Dr. Henry Friedman and Dr. Daniel Fults for kindly providing the medulloblastoma cell lines, Dr. Kenneth Boucher for statistical analysis, and Drs. John Maris and Susan Cohn for helpful comments in their review of the article.