Supplementary Appendix Table of contents: Genes selected for real-time quantitative reverse transcriptase PCR………2 Table 1s: Microarrays results for genes selected for RQ-PCR…………….……...5 Table 2s: Gene list of TaqMan expression assays ……………………………...…6 Figure 1s: An example for a RQ-PCR matrix plate……………………………...….7 Figure 2s: Distribution of quantitative real time RT-PCR values before and after normalization for the various gene…………………………………………………….8 Table 3s: Distribution of means of quantitative real time RT-PCR values for the various genes………………………………………………………….………………….9 Table 4s: Summary of the ranks in Multivariate Cox Regression analysis of quantitative real time RT-PCR results (1- low, 2- medium, 3- high).....……………...………..………………………………………………………….9 Figure 3s: Microarrays results verification. Immunostaning with anti-CDH2.….10 Reference......……………...………..…………………………………...……………11 Genes selected for real-time quantitative reverse transcriptase PCR (RQ- PCR) We hypothesized that increased expression of certain genes in primary NSCLC could identify patients at high risk for the development of brain metastasis. The selection of genes for the RQ-PCR studies was based on results from our exploratory microarray gene expression profiling (GEP) studies (table 1s) [1, 2] and published data linking the expression of these genes either to metastasis (there was no published data on brain metastasis) or to survival in lung cancers or other non- hematological cancers. We focused on genes whose expression was higher either in primary NSCLC with brain metastasis and/or in brain metastasis compared with primary NSCLC. Our microarray study included 26 primary NSCLC (6 metastasized to the brain), 8 normal lungs, 7 brain metastases (not from the same primary NSCLC included in the arrays), and one pooled RNA from normal brain. All the microarray results for the chosen genes were verified by RQ-PCR (and in the case of N Cadherin also by immunohistochemistry) on the samples used for the microarray analysis. The twelve genes (the limit of 12 genes was determined by the availability of RNA) belong to three general gene ontology functional categories: cell proliferation, neuronal, and extra-cellular matrix. This is based on the hypothesis that these three functional groups will be enriched with genes associated with brain metastasis. The "cell proliferation group" consisted of KIFC1 (kinesin family member C1), KIF2C (kinesin family member 2C), KIF14 (kinesin family member 14). All are kinesin family proteins known for their importance in chromosome segregation and mitotic progression[3]. KIF14 has been shown to be a prognostic marker in breast and lung cancer[4-6] and KIF2C has recently been reported to be associated with progression of melanoma, a tumor with high frequencies of brain metastasis[7]. While no publication has associated KIFC1 with cancer so far, we have included it because of it's significant statistical association with brain metastasis in our GEP studies and because of the published data linking the kinesin family with metastasis and survival. 2 CCNB2 (cyclin B2) and the newly discovered SIL (SCL TAL1 interrupting locus) gene, are mitotic regulators whose expression has been associated with metastasis and tumor progression[2, 8-11]. TNPO1 (transportin I) and LMNB1 (Lamin B1) are nuclear membrane proteins, whose expression is increased in proliferating cells and has been associated with metastasis and tumor progression [10, 12, 13]. An increased expression of the cell proliferation genes may theoretically be associated with any type of metastasis. We hypothesized that the expression in primary NSCLC of "neuronal" molecules – genes that normally have high expression in brain tissue – may be associated with a predilection to spread to the brain. Indeed, cancers with a neural crest origin such has melanoma and small cell lung cancer metastasize to the brain frequently[14]. CDH2 (N-cadherin) is a calcium-dependent adhesion molecule critical for many aspects of neuronal development through interactions with neural growth factors[15-17]. Its over expression in gliomas, melanomas and breast cancer is associated with increased invasion and metastasis[18-21]. Correlation between N-cadherin expression and angiogenesis in NSCLC tumors has been shown immunohistochemically. In addition, survival in patients with N-cadherin-positive large cell lung carcinomas was shown to be lower than that in N-cadherin negative cases[22]. In our GEP dataset CDH2 was over- expressed in primary lung cancers that spread to the brain as well as in brain metastases themselves (table 1s). These preliminary findings were also confirmed by immuno-histochemistry (figure 3s). SGNE1 (Secretogranin V) that codes a neuroendocrine chaperonin, was chosen because of its increased levels in brain metastases and in primary lung cancers that spread to the brain in our initial GEP cohort (table 1s), and because a higher expression of neuroendocrine genes has been associated with adverse prognosis of lung adenocarcinomas[23]. Finally, the neuronal transcription factor FALZ (Fetal Alzheimer Antigen) was chosen because of its high expression in brain 3 metastases (table 1s) and a previous publication demonstrating correlation between its expression in primary adenocarcinomas and metastatic spread[10] . The third group of genes included two which code for proteins secreted into the extracellular matrix and whose expression was associated with brain metastasis in our initial cohort (table 1s). ADAM8 (A Disintegrin and Metalloprotease 8) is a member of a family of at least 21 transmembrane proteases. Ishikawa et al have shown that the expression of ADAM8 in the serum of lung adenocarcinoma patients correlated with advanced tumor stage and prognosis[24]. ADAM8 expression levels have also been reported to correlate with glioma invasion[25] and to mediate neuron – glia interactions[26]. SPP1 (Osteopontin) is a secreted cytokine that has been associated with increased invasion and metastasis of multiple cancers including lung and breast cancer, melanomas and brain tumors[7, 8, 27-31]. Significantly, Osteopontin expression correlated with brain metastasis of murine breast cancer[32] 4 Table 1s: Microarrays results for genes selected for RQ-PCR Expression in Expression in Brain Primary NSCLC with Metastasis (n=7) and (n=6) and without Symbol Gene name Affy ID Primary NSCLC (n=26) (n=20) Brain Metastasis RMA Log2 Fold RMA Log2 Fold change* change* 348_at, kinesin family KIFC1 349_g_at, 2.10 1.03 member C1 38933_at kinesin family KIF2C 36837_at 1.86 1.20 member 2C kinesin family KIFC14 34563_at 1.78 1.07 member 14 CCNB2 cyclin B2 32263_at 2.14 1.06 TAL1 (SCL) SIL interrupting 32767_at 1.17 1.44 locus 40463_at, TNPO1 transportin 1 40464_g_a 3.34 0.95 t LMNB1 lamin B1 37985_at 2.66 1.41 cadherin 2, type 1, 2053_at, CDH2 1.37 1.89 N-cadherin 2054_g_at (neuronal) fetal Alzheimer FALZ 41091_at 2.64 1.07 antigen secretory granule, SGNE1 neuroendocrine 34265_at 1.33 2.56 protein 1 (7B2 protein) a disintegrin and ADAM8 metalloproteinas 40712_at 1.39 1.33 e domain 8 secreted 2092_s_at, SPP1 phosphoprotein 34342_s_a 1.21 1.67 1 (osteopontin) t * Fold change values were calculated after RMA (Robust Multi-array Average) normalization [33] and are expressed in log2 values. 5 Table 2s: Gene list of Taqman expression assays Symbol Gene name NCBI RefSeq TaqMan Gene number Expression Assay ID KIFC1 kinesin family member C1 NM_002263 Hs00382565_m1 KIF2C kinesin family member 2C NM_006845 Hs00199232_m1 KIFC14 kinesin family member 14 NM_014875 Hs00208408_m1 CCNB2 cyclin B2 NM_004701 Hs00270424_m1 SIL TAL1 (SCL) interrupting NM_003035 Hs00161700_m1 locus TNPO1 transportin 1 NM_002270 Hs00266970_m1 LMNB1 lamin B1 NM_005573 Hs00194369_m1 CDH2 cadherin 2, type 1, NM_001792 Hs00169953_m1 N-cadherin (neuronal) FALZ fetal Alzheimer antigen NM_004459 Hs00189461_m1 SGNE1 secretory granule, NM_003020 Hs00161638_m1 neuroendocrine protein 1 (7B2 protein) ADAM8 a disintegrin and NM_001109 Hs00174246_m1 metalloproteinase domain 8 SPP1 secreted phosphoprotein 1 NM_001040058, Hs00167093_m1 (osteopontin, bone NM_000582 sialoprotein I, early T- lymphocyte activation 1) ACTB actin, beta NM_001101 Hs99999903_m1 HPRT1 hypoxanthine NM_000194 Hs99999909_m1 phosphoribosyltransferase 1 (Lesch-Nyhan syndrome) For each gene, the RefSeq accession for the mRNA sequence used as the basis for the design of the primer and probe sequences. All of the primer and probe sets listed below were designed and manufactured by Applied Biosystems as part of the TaqMan Gene Expression Assays. These assays can be found at https://products.appliedbiosystems.com and they are currently not approved for clinical usage. 6 Calibration sample ACTB Negative control HPRT CDH2 SIL Tissue sample Figure 1s: An example for a CDH2 and SIL matrix on 96-well plates which contained duplicates for 10 lung tumor samples, a sample of H1299, and a negative control (wells without cDNA). 7 Figure 2s : Distribution of quantitative real time RT-PCR values before and after normalization for the various genes: A - CDH2 ,B - ADAM8 ,C - SIL ,D – TNPO1 ,E – LMNB1 ,F – CCNB2 ,G – KIFC1 ,H – KIF2C ,I – KIF14 ,J – FALZ ,K – SGNE1 ,L- SPP1. The upper panels represents the absolute values, the lower panels represents the values after normalization 8 Table 3s: Distribution of means of quantitative real time RT-PCR values for the various genes GENE N ABSOLUTE VALUES VALUES AFTER NORMALIZATION MEAN SD MEAN SD KIFC1 140 0.437 0.509 -0.520* 0.352 KIF2C 142 0.160 0.208 -0.981* 0.390 KIF14 140 0.470 0.603 -0.598* 0.533 CCNB2 142 0.393 0.358 -0.545* 0.351 SIL 142 0.390 0.357 -0.54** 0.327 TNPO1 142 3.060 3.412 0.359* 0.298 LMNB1 142 0.676 0.452 -0.239* 0.233 CDH2 142 0.854 2.295 -0.516* 0.540 FALZ 142 2.422 2.366 1.44** 0.590 SGNE1 142 7.999 13.720 0.627* 0.460 ADAM8 141 82.194 71.86 8.318** 3.618 SPP1 142 414.177 528.82 2.311* 0.553 * - Calculated using log (base 10) of absolute value ** - Calculated using square root of absolute value Table 4s: Summary of the ranks in Multivariate Cox Regression analysis of quantitative real time RT-PCR results (1- low, 2- medium, 3- high) CENSORED CASES RANK SCORE NO.