Permissivity of the NCI-60 cancer cell lines to oncolytic Vaccinia Virus GLV-1h68

Cancer cell lines

Fifty-nine available cell lines from the NCI-60 panel were purchased from NCI-Frederick Cancer Center DCTD Tumor/Cell Repository. GI-101A cells were kindly provided by Dr. A. Aller, Rumbaugh-Goodwin Institute for Cancer Research, Inc., Plantation, Florida and Huh7.5.1 by Dr. Richard Wang, Department of Transfusion Medicine, NIH, Bethesda, MD. Three autologous melanoma cell lines (888-MEL, 1858-MEL and 1936-MEL) from temporally distinct cutaneous metastases were derived as previously described [18, 19]. MIAPaCa2, HT29, A549, OVCAR3, Panc-1, Siha, MDA-MB-231, NCI-H1299 and PC-3 were purchased from American Type Culture Collection (Manassas) in the past and have been previously characterized for their in vivo responsiveness to viral oncolytic therapy with GLV-1h68 by our group [12].

All cells were cultured in Roswell Park Memorial Institute Medium (RPMI) supplemented with 10% FBS, 10 mM HEPES, 1% antibiotic/antimycotic solution. All cell cultures were carried out at 37°C under 5% CO₂. During prolonged cell culture and immediately before DNA, RNA isolations and viral infection, all cells were tested for mycoplasma with the Venor^®GeM Mycoplasma Detection kit (Sigma, St Louis, MO). In no occasion contamination was detected. Original HLA class I and II loci sequence-based typing was performed on the NCI-60 as previously described [20].

Viral constructs and infections

The construction of mutant VACV GLV-1h68 was described previously [7]. Vaccinia Virus WR-GFP and VSV-GFP were kindly provided respectively by Drs. N. Restifo, NCI, Bethesda, MD and S. Balachandran, Fox Chase Cancer Center, Philadelphia, PA [21]. Before carrying out the experimental procedure, cells were thawed and cultured for three days in RPMI supplemented with antibiotics. One day following the beginning of culture cells were harvested for DNA and RNA isolation while a separate aliquot was exposed to viral infection. For DNA and RNA isolation, 2 × 10⁶ cells were used and 5 × 10⁵ cell were used for viral infection.

Cells were infected with VACV GLV-1h68, at multiplicity of infection (MOIs) of 0.3 and 0.6 or with Vaccinia WR and VSV virus using MOI at 10 and 0.002 respectively. After infection the cells were incubated in cell culture medium at 37°C for 20 hrs. One hour after infection the culture media was replaced with fresh media and cells were incubated at 37°C for 20 hrs.

Quantitative real-time PCR validation of VACV 1h68 gene expression

Differential expression of viral genes in infected cells was detected by using TaqMan^® Gene Expression Assays (Applied Biosystems, Foster City, CA). RNA was reverse transcribed as above with random oligo primers in 20 μl final volume. Primer Express 2 (PE2) (Applied Biosystems, Foster City, CA) was used to generate primers and a TaqMan probe specific for the virus sequence; real-time PCR was performed on thermal cycler 7900HT (Applied Biosystems). Differences in expression were determined by the ratio of Ct values of viral genes over those of endogenous control (18s rRNA).

Plaque forming assay

A375, DU-145 and A549 cell lines were infected with GLV-1h68 at an MOI of 0.6. At 8, 20 and 36 hours after infection, tumor cells were harvested and lysates were prepared by freeze-thawing and sonication in 1 ml of RPMI media with 10% FBS. The viral titers were determined by plaque forming assay using TK-cells. TK-cells (kindly provided by Dr. Jack Bennink, NIAID Bethesda, MD) were cultured in DMEM media with 10% FBS in 6 well plates. When cell density reached 80%-90% confluence, they were washed with 0.1% BSA in PBS and infected with 0.5 ml of diluted cell lysate in PBS with 0.1% BSA for 2 hours at 37°C incubator. Two ml of cell culture media were then added to each well and incubated at 37°C and 5% CO2 for 48 hours. Media were removed from the cells culture and stained with crystal violet (Sigma, St. Luis, MO) and washed with PBS. The titer of the virus was determined by the plaque numbers of the highest dilution. Each cell lysate were titered in duplicate.

FACS-analysis of GFP Protein expression

Twenty hours following viral infection, cancer cells were trypsinized (when applicable), fixed by paraformaldehyde, washed with AUTOMacs running buffer and GFP expression was analyzed with a FACSCalibur flow cytometer. Data were analyzed by the FlowJo software. To ensure between-batch reproducibility, all FACS experiments included BD's Rainbow Calibration Particles (6 peaks) as standards and measured signal intensities were converted in molecules of equivalent fluorescein, phycoerythrin, or allophycocyanin (MEFL, MEPE, MEAPC). The flow cytometer settings were kept constant during the complete experimental procedure and the cell line A549 was included in all experimental batches as a positive control for viral infection to asses inters experimental consistency.

Nucleic acid isolation and preparation

Automated DNA isolation was performed from non-infected human cancer cell lines using Fujifilm's Quickgen DNA Whole Blood kit and Nucleic Acid Isolation System-810. Total RNA (tRNA) from tissue cultures was isolated with the Qiagen miRNeasy Mini kit and its quality was tested with the Agilent Bioanalyzer 2000 (Agilent Technologies). For expression studies, tRNA was amplified into antisense RNA (aRNA) as previously described [22, 23]. Reference for human arrays was obtained by pooling PBMCs from 4 normal donors.

Plaque forming assay on cells cultured with specific culture media

HT29, Panc 1, NCI-H1299 and A549, cell lines were simultaneously cultured in four different cell culture media which were prepared as followed:

Transcriptional analysis

Array quality was documented as previously described [24]. For 36k human array hybridization, a two color system was used; both reference and test aRNA were directly labeled using ULS aRNA Fluorescent Labeling kit (Kreatech Diagnostics, Amsterdam, The Netherlands) with Cy3 for reference and Cy5 for test samples and co-hybridized to the slides. After 20 h incubation at 42°C the arrays were washed, dried and scanning using the Agilent scanner. VACV-gene expression was assessed by a custom-made VACV array platform (VACGLa520445F, Affymetrix, CA) including 308 probes representing 219 genes that covered the combined genome of several VACV strains, the Renilla luciferase-Aequorea green fluorescent fusion gene specific for GLV-1h68, and 337 human or mouse "housekeeping" genes (393 probes) [12]. Five μg tRNA were amplified using the GeneChip^® One-Cycle Target Labeling and Control kit (Affymetrix, Santa Clara, CA). After 16 h incubation in the hybridization oven at 45°C, the arrays were washed and stained in the Fluidics station using the GeneChip^® Hybridization, Wash, and Stain Kit (Affymetrix).

Data processing and statistical analysis

Transcriptional data were uploaded to the mAdb databank (http://nciarray.nci.nih.gov) and further analyzed using BRBArrayTools developed by the Biometric Research Branch, NCI (http://linus.nci.nih.gov/BRB-ArrayTools.html) [25], Partek Genomics Suite (St Louis, MO) or TreeView software [26]. Gene ratios were average corrected across experimental samples and displayed according to uncentered correlation algorithm. Class comparison was performed using parametric unpaired Student's t test or 3-way ANOVA. Adjustments for multiple test comparisons were based on univariate and multivariate permutation test. Gene function interpretation was based on Ingenuity Pathway Analysis (IPA, Ingenuity Systems). Data retrieved from the Affymetrix platform was normalized using median over entire array as reference because of single color labelling technology.

Validation of tumor cell lines

However, their patterns cannot be extrapolated to other homonymous cell lines of the same tissue origin at large because it was impossible to clearly identify and classify those cell lines.

GFP gene expression correlates with viral transcription

To test whether green fluorescent protein (GFP) expression by GLV-1h68 correlated with the global transcriptional activity of GLV-1h68 within infected cells, we hybridized RNA from HT-29 and GI-101A cells infected with GLV-1h68 at different time points to a customized whole genome VACV chip and we monitored the expression of viral genes including Ruc-GFP-fusion. GFP expression was highly correlated with the expression of the remaining 219 genes included in the chip. In particular, GFP expression occurred in the intermediate-late phase of viral genome expression following the expression of early-intermediate genes (like Interferon resistant protein or Interf) and together with other intermediate-late (IMV surface protein or IMV) genes (Figure 1A). Moreover, to confirm these finding, we infected 46 cell lines with GLV-1h68 for 20 hours and tested the expression of GFP and the two viral nonstructural proteins Interf (or K3L) and IMV (or A27L) by RT-PCR. A tight correlation was observed between the expression of GFP and Interf (R² = 0.83) or GFP and IMV (R² = 0.96), (Figure 1B, C). Since the transcriptional expression of GFP is observed in intermediate-late phase and correlated well with the expression of other early or late vaccinia genes at the time point studied, we concluded that GFP is a reliable index of active intra-cellular virally-driven transcription and, indirectly, represent a marker of viral replication.

GFP gene expression correlates with viral copy number estimated by plaque forming assay

To investigate whether GFP positive cells harbor replicating viral particles, we titered the number of plaque forming unit in lysate of infected cells and correlated with the expression level of GFP.

A375, DU-145 and A549 cell lines (showing respectively low, intermediate or high GFP by RT-PCR) were infected with GLV-1h68 for 8, 20 and 36 hours. At the indicated time points, infected cells were collected and viral titers were extrapolated in a plaque assay using TK-cell lines. A375 showed the lowest permissivity to VACV compared to DU-145 and A549 confirming a correlation with RT-PCR based estimates of GFP expression (R² = 0.76) (Figure 1D).

GFP gene transcription correlates with its protein expression

To determine whether the GFP gene transcription correlates with its protein expression, seventy-four cancer cell lines were infected with GLV-1h68 and GFP expression was evaluated by FACS analysis. The infection of the 74 cell lines was performed in batches of five; A549, evaluated to be highly permissive to GLV-1h68 replication RT-PCR, was repeatedly used as a positive control and as cross-reference among different batches. The cell line demonstrated a highly comparable behaviour in each batch (data not shown). Since different cell lines were characterized by different auto-fluorescence, GFP intensity of uninfected cells was subtracted from GFP intensity of infected cells. In addition, cells were observed to vary not only for fluorescence of intensity (MFI) within the GFP positive cells but also for the percentage of fluorescent cells. Therefore for each cell line we calculated the infectivity index as the product the two parameters (MFI of GFP and the percentage of fluorescent cells). Although this index was used for subsequent analyses, either MFI or percent of infection tightly correlated with the parameter and the salient conclusions would not change based on the adoption significantly. We observed a continuous spectrum of GFP expression among cell lines (Figure 1E, Table 2). Infections were performed using multiplicity of infections (MOIs) 0.3 and MOI 0.6. An almost perfect correlation (R² = 0.96) between the infectivity index obtained in the two conditions was observed with the higher MOI consistently yielding a 2-fold higher intensity (Figure 1F). This suggests that concentrations of virus used to test cellular permissivity were within a flexible dynamic range. Since results obtained with the two different MOIs were almost identical, the subsequent discussion will be focused on MOI 0.6 data even though same conclusions could be drawn using the other MOI. Moreover, a good correlation (R² = 0.66) was observed between GFP protein and RNA levels (Figure 2C) which was more significant when two outlier leukemia cell lines (HL-60 and MOLT4) completely resistant to infection were removed from the analysis (R² = 0.75).

Cell line	Ruc-GFP Index	1/Δ Ct	Cell line	Ruc-GFP Index	1/Δ Ct
*MOLT4	242.25	0.05	SNB75	22849.59	0.114
*HL-60	890.01	0.038	OVCAR-3 p42	26919.33	0.102
*CCRF-CEM	890.69	0.095	MDA-MB231 p41	30471.13	#
*K562	1185.59	0.066	OVCAR-8	31066.5	0.142
*SW-620	1223.86	#	ACHN	31639.56	#
*SR	1389.99	0.079	RXF-393	35969.66	0.095
*NCI-H522	1723.7	0.081	EKVX	36117.84	0.111
*OVCAR5	1961.62	#	A498	36486.43	#
*HCT15	2121.68	0.104	Huh 7.5.1	39321.66	0.116
*LOX-IMVI	2591.66	#	NCI-H23	40056.95	#
T47D	2831.5	0.093	PC3 p35	40124.72	#
NCI-H1299	3499.99	#	SF539	46092.96	#
RPMI-8226	3891.89	0.085	NCI-ADR_RES	46569.81	0.138
HOP92	4430.72	#	TK-10	49256.78	0.136
UO-31	5960.36	#	SN12C	50505.78	0.175
A375	6049.59	0.094	MCF-7	55230.93	0.119
HCC2998	6537.28	0.093	SF 295	56037.99	#
786-0	7141.29	0.094	888-MEL	56437.28	0.131
MDA-MB 231	7500.93	#	DU-145	56733.35	0.15
Panc1	7876.6	#	KM12	57112.82	0.144
SK-Mel 5	7917.7	0.089	PC3 p7	57919.09	0.118
SF268	8703.17	#	NCI-H322M	69687.9	0.229
NCI-H460	9247.41	0.097	Siha	72182.24	#
UACC 62	10437.13	0.092	SK-OV-3	73220.28	0.129
U-251	11245.39	#	SK-MEL2	75407.23	0.133
NCI-H226	12182.98	0.108	HT29 p155	84833.55	#
Hs578T	13912.24	#	M14	94724.62	0.171
GI-101A	14791.4	#	*Colo 205	101576.11	0.142
UACC 257	15405.44	0.105	*OVCAR-4	111530.36	0.141
Caki I	15869.9	0.102	*1858-MEL	129262.97	0.164
SK-Mel 28	16737.85	0.097	*397-MEL	134615.99	0.139
MIA PaCa2	17376.47	#	*IGR-OVI I	141059.48	0.147
HOP62	19768.46	0.126	*HT29 p6	159709.89	#
MDA-MB435	20698.92	#	*HCT 116	406258.9	0.141
1936-MEL	21079.94	#	*A549 p120 avg	409912.11	0.25
SNB19	21349.31	#	*OVCAR-3 p7	425559.88	#
BT549	22182.84	0.113	*A549 p4	425962.87	#

To test the correlation between protein and mRNA levels, expression of Ruc-GFP was screened by RT-PCR in 46 cell lines post-infected. RT-PCR values are expressed as 1/ΔCt (Ct GFP-Ct 18S). Cells further used in microarray class comparison and class prediction analysis are indicated with *.

GFP protein expression data were ranked among the 74 cell lines according to infectivity index (Figure 2A). Six out of 6 cancer cell lines of haematopoetic derivation were completely resistant to GLV-1h68 infections. The permissivity to GLV-1h68 based on GFP intensity was not affected by cells size. In fact, cells with similar size such as SK-MEL-2, and SK-MEL-5 (both very large) or A375 and M14 (both very small), showed different permissivity to viral infection. Moreover, permissivity to GLV-1h68 replication was not tissue-specific nor patient-specific as three autologous melanoma cell lines derived from the same progenitor cell clone but from temporarily distant metastases (Mel-888, MEL 1858 and MEL 1936) [18] displayed discordant behavior. Side by side comparison of 1/ΔCt expression levels according to RT-PCR (Figure 2B) demonstrated a pattern of expression similar to the infectivity index calculated according to GFP expression levels.

Detection of cells infected with GFP-carrying vaccinia provides a fast and sensitive method to measure virus replication [27] that in our hands correlated well with transcriptional expression of other VACV genes as well as the commonly used but more laborious plaque forming assay that yields information about viral particle production following cellular infection.

Indexes are similar in other vaccinia strain and Vesicular Stomatitis Virus infected cells

In order to test whether differential replication of GLV-1h68 in various cell line is virus-specific, 6 cell lines were infected in parallel with GLV-1h68 or a different Vaccinia strain (VACV-WR) and 4 cell lines with GLV-1h68 or Vesicular Stomatitis Virus (VSV) [21], both expressing the GFP gene. Twenty hours after infection, infectivity index was calculated (Additional File 1A) revealing a good correlation in cell-specific infectivity driven by the three viruses (R² = 0.84 and R² = 0.87, comparing GLV-1h68 to VACV-WR or VSV respectively) suggesting that a common cancer cell phenotype may influence the replication of these viruses.

Usage of specific cell culture media does not affect permissivity to GLV-1h68 infections

Since the main study was performed in standardized conditions using identical but arbitrarily selected culture media for all cell lines, we tested whether the usage of different cell cultured media could have impact on cells permissivity to GLV-1 h 68 infections. A549, HT29, Panc1 and NCI-H1299 cell lines (showing high, intermediate and low GFP by FACS analysis) were simultaneously cultured in cell specific culture media (as originally recommended by the providing sources) and infected with GLV-1h68. Twenty hours after infection cells were collected and viral titers were extrapolated in a plaque assay using TK-cell lines.

NCI-H1299 and Panc 1 showed lowest permissivity to VACV compared to HT-29 and A549. These observations confirmed the results obtained by FACS analysis and ruled out variation in permissivity to VACV infection due to cell culture conditions (Additional File 1B).

Transcriptional profiles do not differ among cancer cell lines with different permissiveness to GLV-1h68

In order to identify correlates of permissivity to viral infection mediated by GLV-1h68, we performed transcriptional profiles of 74 cancer cell lines prior infection. The 74 mycoplasma-free cell lines used in this study were cultured in identical conditions and subjected for gene expression profiling analysis using 36 K oligo human array. The complete dataset, was filtered (80% gene presence across all experiments) to enrich for informative transcripts obtaining a total of 32,863 transcripts.

Nevertheless, three genes: Growth Differentiation Factor 15 (GDF-15), CD9 and Integrin B5 (ITGB5), known to play an important role in cell movement and adhesion, ranked at the top of those up-regulated in permissive cell lines (Table 3).

In addition, genes involved in IRF (Interferon Regulatory Factor) activation by cytosolic pattern recognition receptors and in RNA polymerase II activation complex, were found to be down regulated in permissive cell lines. Those genes included Interferon Induced with Helicase C domain 1 (IFIHI), RNA Polymerase II (POL II), and Reticuloendotheliosis viral oncogen homolog (v-REL) which are also involved in Interleukin 12 (IL-12) signaling and macrophage activation (Table 3).

The same 371 genes served as a platform to generate a heat map based on archival NCI-60 cell lines (59 out of the 74 cell lines considered in our study) transcriptional data derived from Genomics and bioinformatics groups, LMP, CCR, National Cancer Institute, Bethesda, MD (http://discover.nci.nih.gov/cellminer/loadDownload.do. This independent dataset demonstrated a comparable pattern of gene expression (Figure 3A, B). Thus, it is likely that the transcriptional profiling obtained in our cell lines is representative and confirmatory of the transcriptional profile of NCI-60 panel cell lines obtained in other experimental conditions. Based on the 317 differentially expressed genes with functional annotations, Ingenuity Pathways Analysis (IPA) constructed a primary network centered on NF-kB signaling (Figure 3C) and a second ranking network centered on TGF-β and Interferon-α/β signaling (Figure 3D). In either case, the majority of transcripts were down-regulated in permissive cell lines as it could have been predicted in teleological terms. In fact, expression of v-REL, IFIHI and POL II and other genes involved in IRF 3/7 signaling and activation of innate immunity were down-regulated in permissive cell lines while few transcripts that were up regulated such as GDF-15 had known inhibitory effects on innate immune function. Similarly, the canonical pathway related to assembly of the RNA polymerase II complex was predominantly down-regulated in permissive cell lines (Figure 3E). Several other statistical approaches to identify predictors of permissivity yielded similar results.