We designed a microfluidic tool that enables high-precision analysis of the directional choices made by moving cancer cells (Fig. 1). Inside these devices, cells are loaded through a small inlet and driven through high-aspect-ratio channels sideways to four side channels. These side channels are connected to the migration channels and, through these, to an external reservoir of fresh media. After loading, the cells travel through the high aspect ratio channels and settle at the entrance of the migration channels. While the cross-section of the migration channels is smaller than that of the suspended cells, the settled cells are trapped at the entrance. The cell trapping also restricts the fluid flow during cell loading from the inlet to the migration channels without cells, ultimately helping distribute the cells uniformly through the device at the entrance of the migration channels. After adhering to the bottom of the device, the epithelial cells start moving toward the external chamber. The moving cells are confined in channels and pass through successive where they could enter either dead-end or through channels (Fig. 1A). The design of the device includes a narrow cell loading chamber such that cells introduced into the device are within 200 µm from the entrance to the migration channels, increasing the overall yield (Fig. 1B).
We compared the yield of cells passing through channels with one, three, five, and seven bifurcations. We reasoned that increasing the number of bifurcations could increase the precision of evaluating each cell's performance by increasing the number of repeated observations in similar conditions along the channel. However, while the number of cells reaching the end of channels after a large number of bifurcations decreases, the contribution of later bifurcations to the information generated by the devices also decreases. Moreover, the added migration distance also extends the duration of experiments. Thus, for the purpose of design optimization, we compared the throughput of four designs with 1, 3, 5, and 7 bifurcations. We observed a significant drop in the cellular yield between devices from 5 and 7 bifurcations per channel (Fig. 1B). Thus, we chose for this study to use devices with five bifurcations, which cells traverse in approximately 6 hours.
We quantified the bias of the cells moving through successive bifurcations as orientation factor b and calculated the fraction of cells through each bifurcation, assuming independent decisions and uniform probability of cells taking the through path (Fig. 2A). We represented the changes in the fraction of through cells for orientation factors b with values between 1 and 0.5. A value of the orientation factor b = 1 would indicate cells that are biased towards the through channel all the time, and b = 0.5 cells that make random decisions.
For each set of experimentally determined values, we determined the value of the orientation factor b that best fits the predicted and experimental values. We then employed the orientation factor to compare the performance of cells in different conditions. We found that PC9 cells in growth media display a strong bias towards the through channels. Out of 796 cells moving through devices, 459 reached the end, representing 58% of the cells (Fig. 2B). Our calculated orientation factor for these cells was b = 0.9. If cells made random choices at each bifurcation (b = 0.5), we would have expected 3.1% of cells to go through the entire channel. The twenty-fold difference in the fraction of cells exiting the channel confirmed that a significant bias towards the through channels exists for PC9 cells in media.
We compared the cells moving through the channels with bifurcations in the presence and absence of glucose and glutamine. We found that out of 580 cells to enter the channels in the presence of media with dialyzed FBS, DMEM, glucose (5.5 mM), and glutamine (2.0 mM), 358 will go through all five bifurcations (b = 0.92). If glutamine is absent from the media, the orientation decreases significantly, and only 140 of 482 cells go through (b = 0.70, Fig. 2B). In the absence of glucose, the fraction of cells to go through is comparable to that in the controlled media (b = 0.90). The effect of removing both glucose and glutamine from media is comparable to the removal of glutamine alone (b = 0.70).
We measured the ATP levels and cell speed for PC9 and HMEC cells in the presence of glucose and glutamine in media. We found no significant differences in the ATP levels in the absence of glutamine and glucose and no correlation between the ATP levels and the changes in orientation (Fig. 3A). We also measured the levels of EGF in the media of cells cultivated in the presence and absence of glutamine. We found no significant differences in EGF levels (data now shown).
We also analyzed the effect of glucose and glutamine on migrating non-tumor human mammary epithelial cell line (HMEC). Using a DMEM media as a reference, we calculated that the orientation factor (b = 0.88) decreased both in the absence of glucose (b = 0.67) and the absence of glutamine (b = 0.74). When both glucose and glutamine are absent, an even smaller number of cells migrate, and none of them reaches the end of the channel (b = 0.59). We found that the ATP levels in HMEC cells did not change significantly when glucose and glutamine were removed and increased the most when both glucose and glutamine were removed (Fig. 3B). The interchangeable role of glutamine and glucose in non-tumor epithelial cells (HMEC) and the unique role of glutamine in tumor cells (PC9) point to some contributions from the Krebs cycle. While PC9 cells cannot use glucose in the Krebs cycle, non-tumor cells could use both nutrient sources.
We tested the effect of alpha-ketoglutarate (AKG) on the orientation factor of PC9 cells. We reasoned that AKG could enter the Krebs cycle as an alternative to glutamine in cells that could not use glucose. We found that the orientation factor is corrected by AKG (b = 0.79, Fig. 4). The cells exposed to a combination of AKG and glutamine displayed a modest improvement in orientation (b = 0.88), suggestive for an interchangeable role of glutamine and AKG in cell directionality.