Distinguishing cancerous from non-cancerous cells through analysis of electrical noise

We used the ECIS system to collect micro-motion timeseries data, the fluctuations in which are caused by the movements in a confluent layer of live cells. The system can be modeled as an RC circuit [3,4,5,6]. The cells are cultured on a small gold electrode (5 × 10 -4 cm 2 ), which is connected in series to a 1-Megaohm resister, an AC signal generator operating at 1 volt and 4000 Hz, and finally to a large gold counter-electrode (0.15 cm 2 ). This network is connected in parallel to a lock-in amplifier, and the in-phase and out-of-phase voltages are collected once a second, from which we extract time series of resistance and capacitive reactance (Figure 1a). In ECIS experiments, the fluctuations in complex impedance come * corresponding author: davidra@ewald.cas.usf.edu primarily from changes in intercellular gaps and in the narrow spaces between the cells and the small gold electrode [4,5,6]. A current of about one microamp is driven through the sample, and the resulting voltage drop of a few millivolts across the cell layer has no physiological effect: this is a noninvasive, in vitro-technique. An ovarian cancer line (SKOV3) and a normal human ovarian surface epithelial (HOSE) cell line (HOSE15) were provided by Dr. Samuel Mok at Harvard Medical School. These cells were grown in M199 and MCDB 105 (1:1) (Sigma, St. Louis, MO) supplemented with 10% fetal calf serum (Sigma), 2mM L-glutamine, 100 units/ml penicillin, and 100 microgram/ml streptomycin under 5% CO 2 , and a 37 • C, high-humidity atmosphere. For ECIS micro-motion measurements, cells were taken from slightly sub-confluent cultures 48 hours after passage, and a mono-disperse cell suspension was prepared using standard tissue-culture techniques with trypsin/EDTA. These suspensions were equilibrated at incubator conditions before addition to the ECIS electrode wells. Confluent layers were formed 24 hours after inoculation, resulting in a density of 10 5 cell/cm 2 .

Figure 1a shows a representative 4096-second run (just over one hour) measuring the real part of impedance as a function of time; the example shows a HOSE culture, but to the eye, SKOV cultures do not appear very different. While the example shows increasing resistance with time, others show a decrease; at this time scale, there is no evidence for an overall trend. We collected, under similar conditions, 18 time series for HOSE cultures, of which 16 went for 8192 seconds and two for 4096 seconds. Each 8192-second run was split in two halves, so that effectively we had thirty-four 4096-second runs; however, where appropriate in the analysis below, we discard the second halves of the longer runs in order to avoid inadvertently introducing correlations. Similarly, for SKOV cultures we took data in eight 8192-second runs and ten 4096-second runs, yielding effectively twenty-six 4096second runs. We numerically differentiated the resistance and capacitance time series to obtain noise time series for FIG. 1: Scheme of data extraction from noise. (a) Time series of resistance for one of the experimental runs. Taking the discrete time derivative and normalizing to zero mean and unit variance gives the noise, (b). The power spectrum of noise is shown in (c), using overlapping windows of 256 points in order to reduce scatter. Fits to the first hundred and last hundred frequencies estimate low-and high-frequency powerlaws, f -α . White noise would have appeared frequencyindependent (α = 0). The Fourier transform of the power spectrum gives the autocorrelation, (d), which we fit to a shifted power-law decay and extract the measure β0. As explained in the text, subtle differences in the univariate noise distribution (e) (smoothed) discriminate between cancerous and non-cancerous micromotion.

each, which we normalized to zero mean and unit variance (Figure 1b).

We seek information from the normalized noise series. The first question to pose is whether the noise can distinguish cancerous from non-cancerous cultures, but more generally the measures we extract may be used to test models of cell micromoti

…(Full text truncated)…