In Vivo Flow Cytometry of Extremely Rare Circulating Cells

Abstract

Circulating tumor cells (CTCs) are of great interest in cancer research, but methods for their enumeration remain far from optimal. We developed a new small animal research tool called "Diffuse in vivo Flow Cytometry" (DiFC) for detecting extremely rare fluorescently-labeled circulating cells directly in the bloodstream. The technique exploits near-infrared diffuse photons to detect and count cells flowing in large superficial arteries and veins without drawing blood samples. DiFC uses custom-designed, dual fiber optic probes that are placed in contact with the skin surface approximately above a major vascular bundle. In combination with a novel signal processing algorithm, DiFC allows counting of individual cells moving in arterial or venous directions, as well as measurement of their speed and depth. We show that DiFC allows sampling of the entire circulating blood volume of a mouse in under 10 minutes, while maintaining a false alarm rate of 0.014 per minute. In practice, this means that DiFC allows reliable detection of circulating cells below 1 cell per mL. Hence, the unique capabilities of DiFC are highly suited to biological applications involving very rare cell types such as the study of hematogenous cancer metastasis.

Figure 1

(a) Schematic of the DiFC instrument. See text for details. (b) DiFC fiber-optic probes contain a source fiber, 8 collection fibers, integrated filters and collection optics. The probes were mounted in a holder (c) and placed on the ventral skin surface of the mouse tail. (inset) stylized cross-sectional view of the tail and DiFC FOV. The ventral caudal artery (VCA) and ventral caudal vein (VCV) are approximately 1 mm deep.

Figure 2

Representative data taken with DiFC from (a) sham PBS-injected control mouse, and (b) CTFR-labeled MM.1 S injected mouse. The peaks indicate cells moving through the instrument field of view. (c) We extracted the amplitude, width and linear speed of detected peaks, given the known fiber-separation. (d) we developed an algorithm to determine the direction of cell travel, and remove unpaired peaks (see text). (e) Examples of forward, reverse, and un-paired peaks are shown.

Figure 3

Example data collected from a mouse injected with 10⁵ CTFR labeled cells and 10⁶ unlabeled cells. (a) Processed DiFC data for 1 channel, showing identified arterial, venous and unmatched peaks in the first 30 minutes of the scan. (b) The count rate per minute in the arterial and venous directions, in the first 30 minutes. (c,d) The corresponding data for 30–60 minutes and (e,f) 60–90 minutes of the scan are shown. The count rate declines over time as cells clear from circulation.

Figure 4

Example processed 1-channel DiFC data collected from a PBS-injected control mouse for (a) 0–30 minutes and (b) 30–60 minutes of the scan. (c,d) example data collected from a second example control mouse which was slightly noisier and produced a number of small false alarm peaks (green circles). However, the peak-matching algorithm eliminated these.

Figure 5

(a) Example data collected from an MM-injected mouse, up to 90 minutes following injection. The detected count rate for fibers 1 and 2, as well as the arterial and venous matched count rates are shown. The distribution of measured (b) peak amplitudes, expressed in PMT current (nA) and SNR (dB), (c) detected peak widths for arterial and venous matched cells, and (d) linear cell speeds for arterial and venous matched cells from all mice in this study are shown. (e) Use of a warming pad over the tail significantly increased the linear flow speed of cells moving in the artery. (f) The measured cell speeds and pulse-widths, in combination with a Monte Carlo simulation of the DiFC collection volume allowed us to estimate the depth of origin of the signals, which in this case was 1.1 mm (dotted line).

Figure 6

(a) False-alarm rate measured from sham-injected control mice for different detection thresholds. As shown, use of the matching algorithm drastically reduced false-alarm signals. At the operating threshold of 250 nA, the FAR was 0.014 per minute. (b) The number of cells in extracted blood was compared to the (c) DiFC arterial count rate during the last 10 minutes of scanning. (d) The result showed a good linear agreement between the count rate and MM cell burden.