Local opto-PI3K stimulation steers migrating HL-60 cells

Previous work has found that local activation of phosphoinositide 3-kinase (PI3K) signaling induces the generation of actin-based protrusions such as pseudopodia and growth cones [10,24,30]. Since polarized cells differ in the biochemical compositions of their fronts and backs, we first sought to understand whether the response to local PI3K activation spatially varied in migrating HL-60 cells. Toward this end, we used an optogenetic approach to control PI3K signaling (opto-PI3K) in time and space. This enabled us to probe migrating cells’ biochemical and directional responses to user-defined changes in PIP₃ at different subcellular locations (S1A Fig).

Our opto-PI3K approach uses an improved light-induced dimer (iLID) system [31] to recruit endogenous PI3K to the plasma membrane of HL-60 cells in response to blue light (Fig 1A). We first verified the effectiveness of our optogenetic system using spatially patterned blue light stimulation (470 nm) and PHAkt-Halo, a live-cell biosensor for PIP₃ production [32–34] in stationary cells (S1B–S1D Fig and S1 Video). Upon local activation of PIP₃ production (as evidenced by local accumulation of PHAkt-Halo), we observed a concurrent local increase in the localization of Pak-PBD-mCherry. This live-cell biosensor predominantly recognizes active Rac in HL-60 cells [35] and typically localizes to the fronts of migrating cells. Inspired by previous work in smart microscopy control systems [36–40], we implemented a computer vision–based feedback system that tracks the dynamic features of migrating cells (primarily location and movement, biosensor polarity, and visibility in total internal reflection fluorescence (TIRF) microscopy) and delivers automatically updated, spatially patterned optogenetic stimuli (Fig 1B). Using this system, we can reproducibly track and perturb PIP₃ signals in time and space in moving cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Computer-guided, spatially controlled optogenetic stimulation of PI3K signaling controls cell directionality.

(A) Schematic depicting spatial control of PI3K recruitment using the iLID optogenetic system. A membrane-bound, light-responsive “anchor” (AsLov2-SsrA-BFP-CAAX) exposes the SsrA peptide upon blue light (470 nm) illumination. This exposed peptide can then recruit SspB. By attaching signaling “cargo” to SspB, localization of the cargo from the cytoplasm to the plasma membrane is placed under blue light control. We used the iSH2 domain of P85 beta (which binds endogenous PI3K) as our signaling cargo. (B) Schematic depicting the computer vision–guided, closed-loop feedback control of optogenetic stimulation. This automation enabled us to dynamically reposition optogenetic inputs for moving cells, enabling improved throughput and reproducibility for a given stimulation assay compared to manual inputs. Cells expressed fluorescent translocation-based biosensors for PIP₃ (PHAkt-Halo labeled with JF646) and Rac activity (mCherry-Pak-PBD). Since these biosensors were imaged with 640 nm and 561 nm light, respectively, they could be imaged without activating the blue light–sensitive optogenetic system. (C) Cells expressing the opto-PI3K constructs (shown in 1A) and biosensors for PIP₃ and Rac activity were allowed to migrate freely for 6 minutes. The cells generally maintained their existing directionality during this time. The Rac biosensor is shown in the Single Cell Assay subpanel. In both the Single Cell Assay and the Migration Paths subpanels, the spatial data have been rotated and translated such that the average directionality in the first minute of imaging is toward the top of the figure, and the location of the cell at the 1-minute mark is at the intersection of the gray, dashed lines. Scale bars: 50 μm (D) When cells are stimulated at one side with blue light starting at the 1-minute mark, they reorient their fronts in the direction of the stimulus. The Rac biosensor is shown in the Single Cell Assay subpanel. Scale bars: 50 μm (E) In the absence of blue light stimulation, migrating cells maintain their current axis of polarized signaling. These kymographs represent an average of the radial distribution of reflected blue light (Average Light Input, arbitrary units) or normalized fluorescence intensities (Average Normalized PIP₃ Response, Average Normalized Rac Response) over time across all measured no-stimulation control cells (as shown in C, upper-right). Peripheries are aligned so that the initial direction of the cell is 0°. (F) In the local stimulation condition, cells reorient both their directionality (as shown in 1D) as well as their biosensor distributions. Like 1E, these kymographs show the averaged radial distribution of measured quantities around the peripheries of migrating cells. There is a local increase in PIP₃ at the −90° location, corresponding to the site of optogenetic stimulation. The underlying data for this figure can be found in S1 Data. iLID, improved light-induced dimer; iSH2, inter-SH2; PIP₃, phosphatidylinositol 3,4,5-triphosphate; PI3K, phosphoinositide 3-kinase.

https://doi.org/10.1371/journal.pbio.3002307.g001

Local opto-PI3K stimulation at a given subcellular location (e.g., side, front, back) caused reproducible behaviors across migrating cells. Compared to unstimulated, freely migrating cells (Fig 1C), local optogenetic stimulation of PI3K signaling at one lateral edge of a migrating cell tended to cause that cell to turn toward and persistently migrate in the direction of the optogenetic stimulus (Fig 1D). Local activation at the fronts of cells (S2A Fig) caused highly directed movement in the direction of stimulation, and this motion was more directed than either unexposed control (Fig 1C) or uniformly stimulated cells (S2B Fig). Local activation at the backs of cells (S2C Fig) caused them to either perform a “u-turn” [41] or depolarize, in which case the cell footprint disappeared from the TIRF plane. The directional stimuli generally caused continuous turning responses as opposed to a depolarization–repolarization response (S2 Video), and the rates of these turning responses varied depending on the location of the stimulus (S2D Fig). We could even induce cells to turn continuously by updating the position of the stimulus to continuously target one edge of the cell (S3 Video). To characterize the biochemical responses of these cells, we collected fluorescence data from the periphery of moving cells and mapped it onto a standard frame of reference (angle from cell centroid) to enable averaging within and comparisons between conditions (Figs 1E, 1F, and S3). With these controllable perturbations and measurable outcomes, we could now ask which spatial and temporal features of PIP₃ signaling contributed to these cell behaviors.

Sensitivity to local PIP₃ stimulation is spatially biased to the lateral edges in polarized cells

We next sought to determine whether responsiveness to optogenetic stimulation and downstream activation varied spatially in polarized cells, which are known to differ in the biochemical compositions of their fronts and backs. As other researchers have observed with various signaling inputs, we found the backs of cells to be relatively insensitive to stimulation via opto-PI3K [21,23,26–29]. Cells took longer to reorient (S2D Fig) in the direction of the stimulus and displayed u-turn behaviors in response to stimulation at the back. Restricting signaling competency to the fronts of cells could be one mechanism to ensure persistent movement, though, at an extreme, this strategy could cause polarity locking [3]. To quantify the spatial extent of this bias in cells moving in an approximately two-dimensional environment, we took advantage of our ability to simultaneously stimulate multiple locations in moving cells using our computer vision–driven system. This enabled us to expand upon results from previous work on back insensitivity for cells constrained in one-dimensional channels [28,29].

When we activated both lateral edges of a migrating cell in a 180°-opposed fashion, cells stably turned toward only one of the two stimuli (Fig 2A and S4 Video). When the opposed stimuli were oriented perpendicularly relative to the initial direction of motion, cells broke symmetry evenly, with half of the cells migrating to the left and the other half migrating to the right. This effect was not predicted by initial differences in PIP₃ accumulation as both the “winning” and “losing” side had similar initial local PIP₃ dynamics (Figs 2B and S4). By varying the angle of these opposed competing signals relative to the initial axis of movement, we found that cells were more likely to turn toward stimuli closer to the existing front (Figs 2C and S5). At the extreme, all cells stimulated simultaneously at their front and back continued moving in their initial direction. These results support the notion that the back of the cell responds less strongly to our optogenetic input.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Optogenetic control of PI3K signaling reveals spatial differences in responsiveness to input signals in migrating cells.

(A) When migrating cells are stimulated simultaneously at both lateral sides (at ± 90°) with 180°-opposed blue light opto-PI3K stimuli, they orient themselves in the direction of only one of the two stimuli. Scale bar: 50 μm. (B) Local PIP₃ biosensor dynamics are initially similar whether a cell ends up orienting toward a particular stimulation site (“Winning” Side) or away from it (“Losing” Side). Mean local responses are shown with 95% CI shaded. (C) When the angle of 180°-opposed stimulation sites is varied relative to the initial direction of the cell, a preference for frontward signals is revealed. (from left to right, n = 41, n = 14, n = 10, n = 10, n = 6). n.s.: not significant meaning p > 0.05 under the null assumption that cells are as likely to respond to one stimulus as the other. P < 0.05 inidicated by * symbol. Proportions shown with 95% CI error bars. (D) The stable winner-take-all nature of the 180°-opposed stimuli is a consequence of spatial effects rather than an intrinsic ability of the cells to ignore a second local stimulus. If two spots are located at ± 45° (right) rather than ± 90° (left), cells frequently change their spot prioritization during the course of the 5-minute assay. Proportions turning toward frontward stimuli with 95% CI are shown. The p-value shown is the result of a two-sample Z-test for proportions. Scale bars: 50 μm. (E) The frontward bias shown in 2C does not cause globally stimulated cells (right) to persist in their current direction of movement compared to front-simulated cells (left). Scale bars: 50 μm. The underlying data for this figure can be found in S1 Data. PIP₃, phosphatidylinositol 3,4,5-triphosphate; PI3K, phosphoinositide 3-kinase.

https://doi.org/10.1371/journal.pbio.3002307.g002

Cells stably moved toward only one of the two optogenetic stimuli, indicating that this assay created a winner-take-all scenario. Surprisingly, however, the frontward advantage was gradual. In a straightforward winner-take-all system, as in the case of coupled local positive feedback and global negative feedback, any slight initial advantage should be amplified, leading to switch-like advantages. Thus, we might have expected cells to reliably turn toward more front-oriented signals, even for small angles. We sought to resolve this apparent contradiction by further exploring the winner-take-all nature of two-spot competition assays.

Fronts of polarized cells show behaviors consistent with local negative regulation

The observed stability of migration decisions in response to 180°-opposed stimuli could represent an intrinsic ability of cells to commit to one site of PIP₃ stimulation by suppressing signaling elsewhere. Alternatively, it could represent an initial commitment that becomes stable due to the insensitivity of the back of the cell to stimulation. To differentiate between these possibilities, we altered the stimulation assay to ensure that both stimulation sites would be continuously near the fronts of cells. We accomplished this by comparing cell behaviors for assays in which activation spots were placed at the cell edges either ±90° or ±45° from the initial axis of movement (Fig 2D). In the ±45° assay, if cells were to direct their fronts toward one stimulus, the other stimulus would then be located at the lateral edge of the cell, which should be sensitive to stimulation compared to the back of the cell. During the 5 minutes of frontward-oriented opposed-stimulus exposure, 48% of cells in the ±45° assay showed some degree of reversal between the two orientations, compared to 8% in the ±90° assay (see S6 Fig for a description of our classification strategy). This result suggests that the insensitivity of the backs of cells drives the previously observed stability in the ±90° spot competition assay. The long-term stability is not due to an intrinsic ability to suppress signaling at a distance. On the contrary, the orientation switching observed in the ±45° assay suggests the existence of a factor that locally inhibits front signaling, thereby destabilizing fronts on a slow timescale and making them less able to suppress signaling at distant sites over time.

One additional piece of evidence in support of a local inhibition at cell fronts is the difference in directional responses of the cells to local front stimulation compared to global stimulation assays (Fig 2E). In the global stimulation assay, cells tend to deviate from their initial direction significantly more than cells driven to migrate forward via local optogenetic stimulation. Local inhibition at the center of the fronts of cells could account for these larger responses at regions just outside this area. This is also apparent when examining fold changes in biosensor enrichment around the peripheries of these cells. In both front-stimulated and globally stimulated cells, Rac signal does not significantly increase at the exact fronts of cells during the first minute of stimulation, but the edges of the fronts do increase (S3C and S3D Fig). Overall, our data are consistent with a combination of insensitivity at the backs of cells [21,23,26–29] and local inhibition–based insensitivity at the extreme fronts of cells.

Our data suggest the existence of local inhibition at the fronts of cells, as predicted by Meinhardt [3]. This front-based inhibition is distinct from previously observed inhibition (or lack of responsiveness) at the backs of cells and is instead expected to enable directional plasticity. To directly demonstrate local negative regulation at the fronts of cells, we next sought to remove confounding effects from global negative feedback.

Direct demonstration of local negative feedback for Rac activation

In HL-60 cells with an intact cytoskeleton, protrusion generation causes rapid, actin-dependent increases in membrane tension throughout the cell [10]; thus, we expect cells to limit protrusion growth via mechanically mediated global negative feedback [5,11,42,43]. Since both global and local inhibition act on and are acted on by components of the positive feedback loop (Fig 3A), they are difficult to disambiguate in the intact system. To isolate the effects of local inhibition, we blocked actin-based protrusions with latrunculin, preventing activation of mechanically gated global inhibition. Inhibiting this particular node is also expected to interrupt front-based positive feedback [8], leaving only the putative influence of local negative regulation (Fig 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Using opto-PI3K to reveal locally acting inhibition of Rac activation.

(A) Cell polarity depends on a combination of positive and negative feedback loops with different spatial ranges of action. Locally acting positive feedback (green shaded portion, mediated through phospholipids, GTPases, and actin polymerization) locally amplifies the signals that drive cell protrusion. Actin-based protrusions increase the tension in the plasma membrane to globally inhibit secondary sites of protrusion (purple-shaded region). These two feedback loops (short-range positive, long-range negative) are thought to form the core of a polarity circuit in many cellular contexts but cannot explain the flexibility of cell polarity unless other feedback circuits are also included. An additional locally inhibitory feedback loop has been proposed (blue-shaded region) to avoid the tendency of this core circuit to lock on a given direction of polarity and ignore new stimuli. Direct evidence of this local inhibitor in neutrophil polarity is lacking. (B) Because both local positive feedback and global negative feedback are thought to operate through actin-dependent processes, we hypothesized that treating cells with latrunculin B to block actin polymerization would enable us to observe evidence of local inhibition without the confounding effects of these feedback loops. (C) To demonstrate the absence of global inhibition in latrunculin-treated cells, we exposed them to either one or two spots of opto-PI3K activation and compared the resulting dynamics on both sides. In the absence of a global inhibitor, the responses for a single opto-PI3K stimulus should be the same whether or not there is a secondary site of opto-PI3K activation (in contrast to competition between these sites in the presence of the global inhibitor, as in Fig 2A–2C). When the right sides of cells are compared in each type of stimulation, the presence or absence of the local right stimulus makes a large difference in the observed signaling dynamics. However, the dynamics on the left sides of cells do not differ based on the presence or absence of a distant site of stimulation. Thus, there is no inhibitory action-at-distance, which would be expected from a global inhibitor. The p-values above each graph indicate the probability of observing an absolute difference greater than or equal to that observed by a permutation test. Mean local responses are shown with 95% CI shaded. (D) To demonstrate the presence of local inhibition, we first exposed single cells to a local opto-PI3K stimulus on their left sides, then allowed cells to recover briefly, and then exposed them to a global opto-PI3K stimulus. We reasoned that if the decline in signaling on the left side were due to a local inhibitor, the impact of that inhibition should be observable via a weaker response on the prestimulated side. We observed this evidence of local inhibition at the level of Rac activation (green: Rac is biased toward the right side in phase IV) but not at the level of the opto-PI3K input (orange: no bias of PIP₃ to the right side in phase IV). Mean local responses are shown with 95% CI shaded. The underlying data for this figure can be found in S1 Data. PIP₃, phosphatidylinositol 3,4,5-triphosphate; PI3K, phosphoinositide 3-kinase.

https://doi.org/10.1371/journal.pbio.3002307.g003

To avoid previously observed oscillatory dynamics of Rac activity in latrunculin-treated differentiated HL-60 cells [20–22], we elected to use undifferentiated HL-60 cells, which retain the ability to generate protrusions [10] in response to opto-PI3K stimulation but are basally quiescent. When we locally stimulated one side of these latrunculin-treated opto-PI3K cell with blue light, we observed a sustained increase in PIP₃ (as evidenced by PHAkt enrichment) and a transient increase in Rac activity (as evidenced by Pak-PBD enrichment) (Fig 3C, blue curves in left plots). These dynamic responses were largely confined to the stimulated side of the cell; the side of the cell away from the stimulus experienced much weaker responses of a similar character (Fig 3C, blue curves in right plots). To demonstrate that global negative feedback (long-range inhibition) is not present in this experimental context, we tested whether including second spot of activation showed any evidence of “action at a distance.” We found that the local dynamics at the first side were not measurably impacted by the presence of a second stimulus on the opposite side of the cell (Fig 3C, orange curves in left plots). Finally, the global inhibitor is thought to be involved in protrusion-site competition, enabling a winner-take-all behavior. We compared the local responses of migrating cells stimulated with two sites of optogenetic activation with that of latrunculin-treated cells (S7 Fig). In migrating cells, the Rac response shows a mutually exclusive distribution, suggesting a winner-take-all system. In the latrunculin-treated cells, Rac responses are highly correlated, suggesting little-to-no “action-at-a-distance.”

These results, combined with the transient nature of the Rac response in latrunculin-treated cells, suggest a nonglobal, actin-independent mode of inhibition. To determine the spatial range of this inhibition, we subjected cells to a local optogenetic stimulus until inhibition was observed and then applied a secondary global stimulus after a brief rest period. During this experiment, we measured the asymmetry in signaling by comparing the dynamics on the prestimulated half of the cell and the other side. As expected, local optogenetic recruitment of PI3K resulted in steady and reversible polarization of our PIP₃ reporter (Fig 3D, orange curve). The polarization of the Rac reporter, however, showed different dynamics (Fig 3D, green curve). Upon initial activation (time = 0), the Rac signal becomes transiently hyperpolarized before stabilizing. During the subsequent recovery and global activation phases of the assay, the Rac signal overshoots and becomes slightly polarized toward the previously unstimulated side of the cell before trending back toward a nonpolarized state. An explanation for this Rac-specific polarity reversal is the presence of one or more local Rac inhibitor(s) accumulating at the prestimulus site, accounting for both the early decline from hyperpolarization and the overshooting behavior observed later.

By applying temporally spaced global pulses in latrunculin-treated cells, we identified a 90-second recovery half-time of the Rac response (S8 Fig). Since cells in this latrunculin-treated context do not exhibit global inhibition, this recovery time instead appears to be related to the timescale of recovery from local inhibition (Fig 3A and 3B). Within its full signaling context (in the absence of latrunculin), this local inhibition of Rac activity likely represents a Meinhardt-style local inhibitor, preventing front-locking associated with local-positive-feedback, global-negative-feedback systems.

Local inhibition operates through negative feedback, linking front signaling to input rates

Beyond its potential role in avoiding the locking of cell polarity decisions, we examined how local Rac inhibition could influence a cell’s interpretation of guidance cues over time. Depending on the signaling topology of the inhibition, different strategies of cell guidance can be achieved. [44–48]. For example, bacteria chemotax by using integral negative feedback to temporally sense spatial signal gradients while they move through them [49,50]. The internally measured rate of change of the external signal is used to modulate tumbling frequency and accomplish a biased random walk [51]. We wondered whether the local front-based inhibition that we identified might play an analogous role in neutrophil interpretation of guidance cues through local temporal sensing. To investigate this possibility, we took advantage of the titratability and controllability of our system to understand how PIP₃ input dynamics regulate the Rac response. If Rac operates through an integral negative feedback circuit, it should respond to the rate of change in PIP₃ (analogously to [52,53]).

When we exposed latrunculin-treated cells to a rapid increase in PIP₃, there was a transient increase in Rac activation (Fig 4A, left). We independently confirmed this behavior using an ELISA-like Rac activation assay (Rac 1,2,3 GLISA assay, Cytoskeleton) (S9A and S9B Fig). We took advantage of the precision of our microscopy-based assay to expose cells to a slowly ramped increase in PIP₃ rather than a sudden increase, eventually reaching the same steady state. In this case, we observed little to no Rac activation (Fig 4A, right). Thus, the Rac response in these conditions appears to depend on the rate of change of PIP₃ rather than absolute levels of PIP_3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Local inhibition on Rac signaling enables rate detection that operates through negative feedback.

(A) When latrunculin-treated opto-PI3K cells are exposed to a sudden, step-like increase in blue light (left panel), the PIP₃ biosensor rapidly approaches steady state, and the Rac biosensor has a rapid increase followed by a decline. In contrast, when the light levels are increased slowly (right panel), the PIP₃ biosensor slowly accumulates to a similar steady state as the step input, but the Rac biosensor response is dramatically attenuated compared to the Rac response seen following a step input of PIP₃. Rac dynamics are thus correlated with the rate of change of PIP₃ rather than the absolute levels of PIP₃ in this context. Mean local responses shown with 95% CI shaded. (B) Treating cells with EHT1864, an inhibitor that prevents Rac from signaling to downstream effectors, disrupts the decline in Rac activity that eventually follows a step-like increase in PIP₃. This suggests that Rac activity is essential for Rac inhibition—a negative feedback loop. Adaptation via negative feedback enables global rate-sensing in bacterial chemotaxis, and this may represent a similar strategy for interpreting guidance cues in amoeboid chemotaxis. Mean responses shown with 95% CI shaded. The underlying data for this figure can be found in S1 Data. PIP₃, phosphatidylinositol 3,4,5-triphosphate; PI3K, phosphoinositide 3-kinase.

https://doi.org/10.1371/journal.pbio.3002307.g004

Rate-sensitive responses are a general feature of negative feedback–based adaptive systems, and the Rac activity response to differing PIP₃ input dynamics is consistent with adaptation through negative feedback, as opposed to other circuit topologies like incoherent feedforward [44–48]. To further evaluate whether our system is operating through negative feedback, we tested the effect of Rac inhibition on the dynamics of the response. If the system operates through a Rac-dependent negative feedback loop, then inhibition of Rac’s ability to signal to its effectors should disrupt Rac negative feedback and impair adaptation. We inhibited the ability of Rac to signal to its downstream effectors by treating cells with EHT1864 [54] and observed a dose-dependent shift from transient to sustained Rac activation following an opto-PI3K step input (Fig 4B). (See S9C and S9D Fig for more details of the dose response.) These data suggest that Rac activity is required for Rac inhibition, consistent with Rac-induced Rac negative feedback.

Though Pak-PBD primarily recognizes active Rac in HL-60 cells [35], the domain also has affinity for Cdc42 in vitro and in cells [55,56]. This could complicate our interpretation of Pak-PBD–based measurements. To test whether our biosensor-based results were Rac specific, we verified the transient Rac activation in response to opto-PI3K stimulation with a Rac-specific GLISA assay (S9A and S9B Fig), as mentioned previously. Additionally, we repeated a subset of our experiments in Cdc42-null PLB cells expressing the opto-PI3K system and the Pak-PBD biosensor (PLB cells are a subclone of HL-60 cells, a kind gift from Sean Collins). We found that Cdc42 is not required for the opto-PI3K–induced transient Pak-PBD response (S10A Fig), nor is it required for opto-PI3K–directed guidance (S10B and S10C Fig). Though Cdc42 is an important regulator of cell movement under physiological conditions [22,57,58], optogenetic control of PI3K signaling may isolate a Rac-dominant axis of signaling. The principles learned from this particular signaling axis may reveal general principles of signaling organization across other axes as well.

By operating through adaptive negative feedback, a Meinhardt-style local inhibitor in amoeboid cells could act analogously to the negative feedback system used by bacteria to accomplish chemotaxis through rate measurement. This rate dependence is a proposed behavior in the pilot pseudopod model for gradient interpretation, as initially proposed by Gerisch [25]. The pilot pseudopod model proposes that migrating cells can navigate chemoattractant gradients across many orders of magnitude by maintaining (randomly initiated) pseudopodia that experience a temporal increase in signal as they extend up a gradient at the expense of those that experience a temporal decrease in signal when extending down a gradient. A Meinhardt-style local inhibitor, operating in an adaptive negative feedback mechanism, could be used to detect rates-of-change in local signals and, thus, mechanistically enable pilot pseudpod behavior.

Fronts of polarized cells respond to input signal rates, consistent with a pilot pseudopod model

Our work demonstrates that local negative feedback sensitizes biochemical front signals to temporal changes in signaling inputs, consistent with a pilot pseudopod model of gradient interpretation. We returned to our computer vision–based stimulation assays in the context of moving cells to extend our biochemical signaling observations to cell guidance decisions. We sought to disambiguate the role of temporal changes in input signals from their absolute levels by performing an experiment analogous to our shadow experiment (Fig 3D). Since the backs of cells are relatively insensitive to stimulation, we focused on stimulation of the lateral edges of migrating cells.

When we persistently activate cells with local opto-PI3K stimulation at one lateral edge, we can direct them to continuously turn clockwise (for right edge stimulation) or counterclockwise (for left edge stimulation. See S3 Video). By switching cells from local to global activation, both edges of the cell attain the same final level of signaling but through different histories. The prestimulated edge will have a smaller increase than the stimulus-naive edge (Fig 5A). If cells respond to signaling levels without history dependence, they should respond to the global stimulus as they did previously (Fig 2E, right). In contrast, if cells respond to temporal changes in input signals (such as through the local Rac negative feedback loop we have demonstrated), we expect cells to reverse directionality due to the larger increase on their stimulus-naive sides. Our experiments confirm the local temporal interpretation of PIP₃ inputs (Fig 5B and S6 Video) in a manner analogous to our shadow experiments in latrunculin-treated cells (Fig 3D and S5 Video).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Cell fronts respond to local temporal changes in input signals, consistent with a modified pilot pseudopod model of cell guidance.

(A) To probe how cells decode the spatial and temporal features of input signals, we designed a test of the pilot pseudopod model. Cells were first tracked in I, then continuously stimulated (via opto-PI3K) at their left edge in II, and finally exposed to global illumination in III. This stimulation regime is expected to cause both sides of the cell to arrive at the same level of PIP₃ with different local rates of change, enabling us to determine the relative contributions of spatial and temporal features of input signals to cell orientation. (B) A tracked and stimulated cell responding to the stimulus scheme described in Fig 5A. See S6 Video for an animated version of this subpanel. Scale bar: 50 μm. (C) Tracks of multiple cells (n = 30) responding to the stimulus scheme described in A show a net bias toward reversal following global stimulation. See S7 Video for an animated demonstration of this behavior. Scale bar: 50 μm. (D) Angular velocities were estimated for cells at the beginning of each of the phases depicted in Fig 5A. P values shown describe a t test for whether the mean of each distribution is zero. Cells transition from counterclockwise motion to clockwise motion following the switch from local stimulation (phase II) to global stimulation (phase III). (E) The PIP₃ signals on the left and right sides of migrating cells were calculated and then fit with a time-delayed saturable function. Sample means are shown as dots, and the 95% CI is shown as the shaded area. The fit functions are shown as dashed lines. As expected, both sides approach the same steady state at the end of phase III, but their histories differ. The changes in cell directionality correlate with the local derivative of input signals rather than the absolute levels of the input signal, consistent with a pilot pseudopod model of gradient interpretation. The underlying data for this figure can be found in S1 Data. PIP₃, phosphatidylinositol 3,4,5-triphosphate; PI3K, phosphoinositide 3-kinase.

https://doi.org/10.1371/journal.pbio.3002307.g005

We estimated the average angular velocity (in rotations per minute) of cells during each of the three phases of the assay across replicates of this assay (Fig 5C and S7 Video). We found statistically significant changes in these velocities (Fig 5D). The p-values shown indicate the likelihood of observing these distributions under the null hypothesis of an average angular velocity of zero. We verified that the PIP₃ biosensor matched the expected temporal profile in Fig 5A on the population level (Fig 5E). Though the left and right sides of migrating cells reach similar steady states, the rates of change differ between the two sides. Turning behaviors of the cells coincide with the temporal changes in the input signal rather than the absolute levels of the signal. As further validation, we measured initial rates of PIP₃ accumulation (as assayed via PHAkt recruitment) on each side for individual cells during the assay. In most cases, the initial rates matched the expected behavior (S11 Fig).

By performing an experiment where we subject cells to local temporal changes in PIP₃ in the absence of absolute differences in PIP₃ across the cell, we demonstrate the sensitivity of cells to local signaling changes for cell guidance, consistent with a pilot pseudopod model of gradient interpretation.

Cell culture

HL-60 cells were grown in RPMI 1640 media supplemented with l-glutamine and 25 mM HEPES (Mediatech) and containing 10% (vol/vol) heat-inactivated fetal bovine serum (Gibco). Cultures were maintained at a density of 0.2 to 1.0 million cells/ml at 37°C/5% CO₂. HL-60 cells were differentiated by adding 1.5% (vol/vol) DMSO (Sigma-Aldrich) to actively growing cells and incubated for 5 days. HEK293T cells (used to generate lentivirus for transduction of HL-60 cells) were grown in DMEM (Mediatech) containing 10% (vol/vol) heat-inactivated fetal bovine serum and maintained at 37°C/5% CO₂.

Plasmids

Plasmids were assembled using a Golden Gate–based modular cloning toolkit [73,74].

The two key constructs for opto-PI3K system are based on the iLID system [31] and include a mTagBFP2-tagged, membrane-localized, light-sensitive component: AsLov2-SsrA-mTagBFP2-CAAX and an EGFP-tagged translocatable PI3K-binding component: ISH2-EGFP-SspBMicro. Upon 470-nm light exposure, the translocatable component localizes to the plasma membrane, and the ISH2 domain recruits endogenous PI3K to the plasma membrane [8,37].

Biosensors for PIP₃ [32–34] and active Rac [35] were also created using the modular cloning kit and were designed to be continuously imaged during experiments without activating the blue light–sensitive optogenetic system. For experiments shown in this paper, PIP₃ was detected using a Halo-tagged [75] PH domain from Akt labeled with Janelia Fluor 646 [76], and active Rac was detected using an mCherry-tagged Pak-PBD domain from Pak1.

Transduction of HL-60 cells

HEK293T cells were seeded into 6-well plates and grown until about 80% confluent. For each well, 1.5 μg pHR vector (containing the appropriate transgene), 0.167 μg vesicular stomatitis virus-G vector, and 1.2 μg cytomegalovirus 8.91 vector were mixed and prepared for transfection using TransIT-293 transfection reagent (Mirus Bio) per the manufacturer’s instructions. After transfection, cells were grown for additional 3 days, after which virus-containing supernatants were harvested and concentrated 20-fold using a Lenti-X Concentrator (Takara Bio) per the manufacturer’s instructions. Concentrated viruses were frozen and stored at −80°C until needed. For all transductions, the thawed virus was mixed with 3 × 10⁵ cells in growth media supplemented with polybrene (8 μg/ml) and incubated overnight. Cells expressing desired transgenes were isolated by FACS.

Preparation of cells for microscopy

Cells expressing Halo fusion proteins were stained with 10 nM JF646 (Janelia) for 10 to 15 minutes at 37°C in complete media and then rinsed once with complete media (RPMI with 10% FBS) before placing cells in reduced-serum media for migration assays.

We used an under-agarose preparation [77] with reduced serum (RPMI with 2% serum) to keep cells adjacent to the coverglass for TIRF imaging and optogenetic stimulation. This preparation involved layering 2% to 2.5% low melting point agarose onto cells after they had been allowed to attach to the glass for 10 to 15 minutes.

Microscopy hardware

Hardware used for the experiments included an Eclipse Ti inverted microscope equipped with a motorized laser TIRF illumination unit, a Borealis beam-conditioning unit (Andor Technology), a 60× Plan Apochromat TIRF 1.49 NA objective (Nikon), an iXon Ultra electron-multiplying charge-coupled device camera, and a laser merge module (LMM5; Spectral Applied Research) equipped with 405-, 488-, 561-, and 640-nm laser lines. All hardware was controlled using Micro-Manager [36,38] (University of California, San Francisco), and all experiments were performed at 37°C and 5%CO₂.

Activity of opto-PI3K was controlled via a 470-nm (blue) LED (Lightspeed Technologies) that transmitted light through a custom DMD (Andor Technology) at varying intensities by connecting the LEDs to the analog outputs of a digital-to-analogue converter and setting the LED voltages using serial commands via custom Python code. Our microscope is equipped with two stacked dichroic turrets such that samples can be simultaneously illuminated with LEDs using a 488-nm longpass dichroic filter (Chroma Technology) in the upper turret while also placing the appropriate dichroic (Chroma Technology) in the lower turret for TIRF microscopy.

Automated microscopy and light activation pipeline

Images were collected, and light was spatially patterned using custom tracking and light-patterning code written in python and making use of Pycromanager [40]. The code we used for our experiments and analysis is available on Zenodo (10.5281/zenodo.8217768).

In brief, a program scanned locations in a prepared well while running a segmentation algorithm at each location. If a segmented object was found to be an appropriate size and intensity, the program would then image that object for 30 seconds to test whether it was moving at a speed appropriate for a migrating cell. If so, it moved that object into the center of the field of view and initiated a tracking and stimulus protocol to track the cell over time and deliver programmed stimuli described in this paper.

Image and data analysis

Measurement of Rac-GTP by GLISA

For biochemical verification of Rac-GTP dynamics in response to acute optogenetic activation of PIP₃, we used a colorimetric Rac 1,2,3-GLISA (Cytoskeleton).

After testing several cell densities to determine the linear range of the assay, we settled on using 2 × 10⁵ undifferentiated HL-60 cells per sample. So, 2 × 10⁵ cells in 10 μL DPBS were placed in centrifuge tubes and exposed to a saturating dose of blue LED light for varying amounts of time. Cells were lysed in 120 μL ice cold GL36 buffer, and 100 μL lysate was transferred into ice-cold tubes and immediately snap-frozen in liquid nitrogen. The remaining 20 μL were used to verify that each of the samples had similar protein concentrations (of approximately 0.16 mg/mL).

Once all samples were collected, the 100 μL aliquots were thawed in a room temperature water bath and processed according to the manufacturer’s instructions.