Amoeboid Chemotaxis

Abstract

Chemotaxis is the directed movement of a cell towards a gradient of chemicals such as chemokines or growth factors. This phenomenon can be studied in organisms ranging from bacteria to mammalian cells, and here we will focus on eukaryotic amoeboid chemotaxis. Chemotactic responses are mediated by two major classes of receptors: GPCR's and RTK's, with multiple pathways signaling downstream of them, certain ones functioning in parallel. In this review we address two important features of amoeboid chemotaxis that will be important for further advances in the field. First, the application of in vivo imaging will be critical for providing insight into the functional requirements for chemotactic responses. We will briefly cover a number of systems in which in vivo imaging is providing new insights. Second, due to the network-type design of signaling pathways of eukaryotic chemotaxis, more refined phenotypic analysis will be necessary, and we will discuss recent analyses of the role of the phosphoinositide 3-kinase pathway in this light. We will close with some speculations regarding future applications of more detailed in vivo analysis and mechanistic understanding of eukaryotic amoeboid chemotaxis.

Introduction

Amoeboid cells display chemotactic responses through rearrangement of the cytoskeleton.^1,2 The actin and microtubule networks are flexible and distinct; cell movement is based on the dynamic growth and turnover of these structures. Chemotaxis of amoeboid cells plays an important role in normal functions such as food sensing (for single cells), embryogenesis, development and wound repair.^3,4 Chemotaxis is also involved in pathological conditions including inflammatory diseases (rheumatoid arthritis, multiple sclerosis, etc.),⁵ Alzheimer's disease6 and tumor metastasis.^7–9

Extracellular chemotactic cues are transduced by receptors on the cell surface into signals leading to cellular responses such as polarization and reorganization of the cytoskeleton.¹⁰ These signals are quite complex but cooperatively they lead to cell migration.¹ Migration on a two-dimensional (flat) surface is currently best characterized and can be viewed as composed of four overlapping processes: cell polarization and protrusion of the membrane at the leading edge, formation of contacts with the substratum at the protrusion, contraction and translocation of the cell body, and detachment of adhesions at the rear with retraction of the trailing edge.^10,11 Different cell types will show variations on this theme. For example, more rapidly moving cells, such as Dictyostelium discoideum or neutrophils, can have weaker adhesion sites, resulting in cell motion depending more on the rate of protrusion.¹¹

Two general classes of chemoreceptors have been studied in detail—G protein coupled receptors (GPCRs) and receptor tyrosine kinases (RTKs). GPCRs tend to be the major mediators of chemotactic responses on rapidly moving cells such as neutrophils and Dictyostelium; RTKs tend to be used as receptors on slower moving cells such as fibroblasts and epithelial cells.^5,12 However, many cell types express both types of receptors; for example, macrophages express CXCR4 and CSF1-R, fibroblasts express EDG receptors and PDGFR^13,14 and tumor cells can express CCR7 and EGFR.^12,15 Some of the types of pathways downstream of these two receptor types are the same, such as phosphoinositide 3-kinases (PI3Ks) and phospholipases (PLCs, PLAs), although they use different isoforms of the signaling molecules (see Fig. 1 for a summary of chemotactic signaling pathways). However, pathways may be more strongly activated in one type (for example, the ras/MAP kinase pathway tends to be more strongly activated by the RTKs).^16–19 There also tends to be a difference in the dynamics of responses, with GPCR's having the capability for rapid responses as well as strong amplification of the signal.²⁰ This raises one of the questions that will be important in the coming years—are there functional differences in the types of chemotactic responses that are initiated by GPCR's compared to RTK's? Are these differences due to the different biological functions (such as recruitment of leukocytes to a site of infection vs migration of an epithelial cell to recover a wounded surface) that signals from specific receptors need to fulfill? Such questions will need in vivo analysis, and an overview of some of the in vivo models currently in use will be covered in the next section.

Figure 1.

A model summarizing some of the pathways described in this review leading to chemotaxis and cell motility. A number of parallel pathways are activated upon receptor binding chemoattractant, including PI3K, PLC, PLA₂, GC and ras (red arrows). The PI3K pathway (activated directly and also via ras and rac) results in increased PIP₃, producing activated small G proteins cdc42 and rac, both of which have been shown to activate actin polymerization through WASP/WAVE family protein stimulation of the Arp2/3 complex. The PLC pathway releases cofilin bound to PIP₂, allowing it to cleave actin and stimulate new barbed ends for actin polymerization. PTEN, a PIP₃ phosphatase which can be localized by PIP₂, is more concentrated towards the rear of the cell, helping to reinforce a gradient of PIP₃ in the cell membrane that is highest near the front of the cell. Ca released by the PLC product IP₃ synergizes with the activation of PLA₂ for stimulation of chemotaxis as discussed in the text (the direct downstream targets of PLA₂ are still unclear but one downstream response is actin polymerization). Activation of ras also stimulates MAP kinase activity. Pathways stimulated by MAP kinase include activation of MLCK, resulting in contractility (in part via stress fibers), as well as calpain activation, resulting in cleavage of integrins for detachment. Activation of GC leads to increased cGMP, also producing increased MLCK activity and cell contraction. The net result is a bias towards activation of actin polymerization on the side of the cell closest to the chemoattractant source and detachment/contraction on the side away from the stimulus, producing cell polarization, protrusion and chemotactic movement.

Complementary to our understanding of the in vivo significance of chemotaxis for different receptors or cell types will be dissecting the contributions of specific pathways for a specific receptor system in a single cell type. In the E. coli or S. typhimurium bacterial chemotaxis systems, a reasonable approximation can be made of one major signaling pathway leading from the receptors to the motors, and that the motor function is separated from the upstream signaling pathways.²¹ These two assumptions have simplified the analysis of the mechanisms of bacterial chemotaxis significantly (although at finer levels of analysis potential complications exist). Neither assumption is true for amoeboid chemotaxis. Viewing an amoeboid cell as a plasma membrane enclosing a cytoskeletal system composed of actin and microtubules, the receptors in the plasma membrane activate pathways that then impinge on the cytoskeleton and adhesion systems.^20,22 However, the cytoskeleton and adhesion systems have the potential to regulate the signaling pathways as well: some guanine nucleotide exchange factors can be bound to microtubules and F actin, and adhesion sites can stimulate the same pathways that are activated by the receptors, such as the Ras and PI3K pathways.^20,22 Thus the motor system is likely to be an integral part of the signaling system. In addition, as noted in the previous paragraph, downstream of the receptor are multiple signaling pathways that are activated essentially in parallel. This then can complicate the process of determining the mechanism of detecting the gradient, since inactivation of any one pathway may only generate a partial defect. The ability to unravel such parallel pathways then will rely upon the development of appropriate genetic screens and quantitative assays. An example of such a case will be provided in the section on PIP₃ signaling.

Back to the Future (In Vivo)

Although in vitro studies (for this review we consider using cells in culture to be in vitro studies) provide the opportunity for more accurate measurement under controlled conditions, chemotactic responses have developed and been selected for under in vivo conditions. Thus a clear understanding of the differences between GPCR and RTK signaling may require returning to the in vivo conditions under which the behavior may have been selected for during evolution. A first step is shifting away from the rigid, two dimensional world of glass or plastic surfaces to studying cell movement and motility in three-dimensional matrices. A flexible, enclosing matrix that is more similar to the in vivo microenvironment produces remarkable changes in cell structures, for example with fibroblast focal adhesions becoming much less prominent.^23–25 Using three-dimensional matrices, differences in what have been termed ‘amoeboid’ and ‘mesenchymal’ cell motility have been discovered based on their differences in integrin, protease and cadherin dependence.^26,27 It has been shown that fibroblasts²⁸ and tumor cells²⁷ tend to crawl through three-dimesional matrices using mesenchymal motility that is integrin and proteolysis dependant with recruitment of cell surface matrix-degrading proteases from the matrix metalloproteinase (MMP) and serine protease families (urokinase-type plasminogen activator uPA/uPAR).²⁶ Leukocytes, on the other hand, lack stable focal contacts and tend to exhibit amoeboid motility that is not based on proteolytic activity but still can be integrin dependent.²⁶ But when proteolytic activity is blocked, tumor cells can transition to a more amoeboid shape and motility without a loss of speed when moving through the matrix. For tumor cells to crawl through a collagen matrix in an amoeboid fashion, the cell generates enough actomyosin force to deform the matrix to allow it to crawl through.²⁹

Moving the study of chemotaxis into living organisms through advances in imaging technologies has become advantageous because it allows integration of studying the cell in its natural environment with manipulation of both the chemokines and the ability of cells to sense them. Genetic manipulation of cells within Drosophila and Danio rerio is leading to a better understanding of how cells move in vivo during morphogenesis.

In the fruit fly, Drosophila, green fluorescent protein (GFP) tagging has been used to follow border cell movement during oogenesis. Border cells are part of the follicular epithelium, which migrate as a cluster of cells from the anterior part of the egg chamber to the oocyte border at cell division stage 9.³⁰ The effects of genetic manipulation of factors that regulate border cell motility, such as the receptors controlling chemotaxis and their respective ligands,³¹ can now be followed in real time using conventional microscopy. The results from these experiments suggest a reevaluation of the models for the migration of these cells.³⁰ They have unexpectedly demonstrated that inhibition of the guidance receptors EGFR and PVR did not suppress cellular protrusion, as was expected from studies of fixed samples and the hypothesis that receptor activation stimulates protrusion.³⁰ In fact, more protrusions were produced upon receptor inhibition, but extending in all directions. Thus the chemoattactant receptors were mainly inhibiting inappropriate protrusions. Additionally, it was found that the leading cell constantly changes and that “shuffling” and “tumbling” of the border cells in the cluster is part of this remarkable dynamic migration process.³²

Live imaging in vivo has been used to demonstrate for the first time the role of Rho signaling in matrix and cell-cell interactions of hematocytes in Drosophila embryos.³³ This has confirmed the role of Rac in those cells as essential for their motility. A surprising result was that Cdc42 is dispensable for hemocyte chemotaxis in embryonic Drosophila wounds, despite previous in vitro evidence showing that blocking Cdc42 signaling in cultured cells such as Bac1.2F5 macrophages and fibroblasts inhibits their ability to respond to chemotactic signals.^33–35

Using the zebra fish, Danio rerio, transgenic lines that express GFP only in neutrophils provide a vertebrate system that has been used to advance our understanding of the inflammatory response to wound healing in vivo.³⁶ Before in vivo imaging was used to study this system, it had been generally thought that the process by which neutrophil-mediated inflammation is resolved involves apoptosis of these cells in the wounded tissue. However, in vivo time-lapse microscopy of this system revealed that perhaps inflammation is resolved instead by retrograde migration of neutrophils back to the vasculature.³⁶

With the advent of multiphoton microscopy, intravital imaging of mammalian systems has greatly improved and has opened up new ways to explore chemotaxis, cell-cell interactions and the metastatic cascade within the in vivo microenvironment. Multiphoton microscopy involves imaging with an IR laser at higher wavelengths of light that allow for deeper tissue penetration, less phototoxicity and less bleaching.³⁷ Further, the second harmonic signal generated by the IR excitation allows for the visualization of collagen fibers within the tissue.³⁸ A variety of cell types have been looked at using this technology ranging from immune cells to tumor cells. T-cell interaction with B-cells^39,40 has been imaged using endogenous EGFP-labeled B-cells and exogenously labeled T-cells introduced by tail vein injection, which allows for the visualization of both cell-cell interactions and localization of cells within the living lymph node. With such methods, in vivo imaging has provided us with evidence that dendritic cells and natural killer cells form direct, stable contacts with each other.⁴¹

Studying tumor cell motility in vivo at the primary tumor is another example of the advantages of in vivo imaging. Overexpression of VEGF-C on tumor cells caused hyperplasia to occur in lymph nodes, increasing lymphatic flow and tumor cell metastasis to the lymph node.⁴² Blocking VEGFR-3, a receptor for VEGF-C, reduced flow and tumor cell drainage into the node. Collectively, these intravital imaging results revealed that while VEGF-C did not confer a growth or survival advantage to B16F10 melanoma or T241 fibrosarcoma cells, it is implicated in the metastatic process to the lymph node, and could therefore be a target for the initial steps in lymphatic metastasis.⁴² With respect to intravasation and distant metastasis, in the primary tumor, the roles of cell motility and the EGF receptor in intravasation have been studied. Imaging of tumor cells in situ revealed that cell motility, orientation and polarization to vessels, are correlated with enhanced intravasation and metastasis.^43,44 Overexpression of the EGF receptor can enhance metastasis through increased motility, orientation towards blood vessels and intravasation.^45,46

The (Signaling) Matrix

A number of intriguing experiments originally suggested that PI3K function and PIP₃ signaling could play an important role in amoeboid chemotaxis. The case for the importance of PIP₃ for cell polarization was made most dramatically in Dictyostelium and neutrophils. In Dictyostelium, the use of a PH domain-GFP fusion enabled direct visualization of PIP₃ during chemotactic stimulation.⁴⁷ PIP₃ would rapidly accumulate on the sides of cells closest to the chemoattractant source. Blocking actin polymerization using latrunculin produced round cells without the complexity of time-varying protrusions but even in these cells dramatically increased PIP₃ levels were seen on the side of the cell next to the chemoattractant source (PIP₃ polarization) that would rapidly track the source as it was moved around the cell.⁴⁷ Similar results were also presented for neutrophils.⁴⁸ The rapid polarization of PIP₃ concentration even in the absence of cell movement or actin polymerization indicated that PIP₃ polarization could be an important event in sensing the gradient and that cytoskeletal rearrangements and cell movement might then follow the PIP₃ signal.⁴⁷ This hypothesis was supported by studies in which suppression of PIP₃ hydrolysis (through disruption of the 3′ phosphatases PTEN or SHIP1), mistargeting of PI3K or inhibition of PI3K by genetic deletion or drug treatment was reported to inhibit chemotactic responses.^49–51 An important role for PIP₃ also fits well with its function in regulating the rho family small G proteins, especially rac, which clearly regulate actin polymerization.

However, further study raised questions about how important PIP₃ signaling really was for detection of the chemoattractant gradient, or the accuracy of the cell's compass. Ferguson et al. did not find such dramatic inhibition of chemotaxis; deletion of PI3Kγ in neutrophils affected cell speed more than navigation in the right direction.⁵² Hoeller and Kay most thoroughly tested the role of PIP₃ in chemotaxis of Dictyostelium cells.⁵³ They removed all five type 1 PI3Ks present in the Dictyostelium genome as well as PTEN. Although there was no PIP₃ production or polarization of PIP₃ towards the chemoattactant source, F-actin still showed a polarized response to a chemoattractant gradient. In addition, the chemotactic response to a strong source showed no difference between wild-type and the quintuple PI3K/PTEN mutant. However, the mean velocity of movement was reduced to around 40% of wild-type in the absence of a gradient and 75% of wild-type during chemotactic movement.⁵³ Andrew and Insall also found that inhibition of PI3K affected spontaneous motility; in particular the frequency of pseudopod formation was affected, but not the accuracy of preferentially maintaining the pseudopod that has extended in the correct direction.⁵⁴

This paradox has been resolved, perhaps completely, by the Devreotes and van Haastert groups in two complementary approaches.^16,19,55 They find that there are (at least) two parallel pathways that are important in gradient sensing: one dependent on PI3K, the other dependent on PLA₂.

Chen et al. used a genetic strategy to explore this conundrum, using aggregation as a rapid surrogate measure of chemotaxis.¹⁶ They screened randomly mutated cells for defective aggregation only in the presence of the PI3K inhibitor LY294002. This identified a phospholipase A₂ homologue (termed plaA), whose importance was confirmed by using homologous recombination to selectively remove the plaA gene.¹⁶ Similarly, PI3K mutants treated with the PLA inhibitor bromoenol lactone (BEL) showed strong defects in aggregation. Chemotaxis assays then confirmed that the aggregation defects reflected defects in chemotaxis. PIP₃ production was normal in plaA mutants, as was attractant-stimulated actin polymerization.¹⁶ However, using cells labeled with arachidonic acid to follow lipid metabolism, they found that plaA mutants showed a strong decrease in attractant-stimulated phospholipid metabolism. The specific arachidonic acid metabolite that is altered in plaA cells has not yet been identified. Treatment of plaA mutants with LY resulted in a 50% decrease in total actin polymerization (F-actin) induced by chemoattractant, suggesting that the initial actin polymerization response is partly dependent on both the PLA₂ and PI3K pathways and that ablation of both pathways simultaneously would be required to completely inhibit this response.¹⁶

van Haastert et al. used theoretical systems analysis to guide a directed combination of mutant and pharmacological manipulations.¹⁹ From a theoretical analysis of the requirements for chemotactic signal transduction, they resolved that appropriate second messengers should accumulate transiently with a time constant of around five seconds after stimulation. The signaling pathways whose outputs fit this constraint were PI3K, PLC, PLA₂ and cytosolic calcium. Single and multiple combinations of mutants and drugs specific for individual pathways, comprising 70 different experimental conditions and a total of 500 measurements were used to evaluate the roles of these pathways in chemotaxis. Blocking all four pathways resulted in complete inhibition of chemotaxis, while allowing the PI3K or PLA₂ pathway alone to function was sufficient for nearly normal chemotaxis in steep chemoattractant gradients. However, in shallow gradients of chemoattractant, inhibition of either PI3K or PLA₂ was sufficient to strongly inhibit chemotaxis. Intriguingly, inhibition of the PLC pathway alone significantly reduced polarized localization of PIP₃; however chemotaxis was not affected, since the PLA₂ pathway was still functioning normally.¹⁹ This suggests that increased levels of PIP₂ in the PLC mutants target increased levels of PTEN to the plasma membrane, resulting in levels of PTEN which inhibit the normal chemoattractant-stimulated localized PIP₃ production, due to PI3K activation.⁵⁵ Conversely, an IP₃ receptor mutant was insensitive to inhibition of PLA₂, but lost chemotactic ability upon inhibition of PI3K.^19,56 This suggests that the proper functioning of the PLA2 pathway is dependent on intracellular calcium.

Further exploration of the PLA₂ pathway will be important—identification of the specific downstream second messenger, how intracellular calcium may regulate this pathway, and the immediate downstream target that couples the PLA₂ pathway to cell motility are the immediate questions that arise. Key to the systems analysis approach was the development of an appropriate assay: high throughput yet flexible in terms of being able to accommodate varying important parameters such as chemoattractant gradient steepness.

The take home lessons from the PIP₃ story are multiple. Some experiments such as the PTEN or SHIP1 knockouts may work as dominant negatives—perhaps by causing pathways activated by PIP₃ (which are sufficient but not necessary under normal circumstances for chemotaxis) to override the other pathways through excessive PIP₃ signaling. Perhaps the most important lesson is that fine control over the assay that is used is critical in determining the ability to dissect a system. Chemotaxis assays are extremely varied: Boyden chamber type assays count the number of cells that cross a filter (which could also be a measure of cell motility rather than chemotaxis), lacking the ability to provide high resolution images of the cells as they are moving.^57,58 In the versions that use a Nucleopore filter, the pores that the cells traverse can be significantly smaller in diameter than the cell, forcing the cell to squeeze itself through the pore; convolving the ability to generate force with the ability to actually detect the gradient.^57,58 The defects in cell translocation and crossing Nucleopore filters exhibited by the PIP₃ mutants may indicate that PIP₃ plays a role in regulating adhesion or force generation. Another type of chemotaxis assay uses a micropipette to generate a gradient near a cell and allows direct visualization of the cell during its responses.^59,60 Depending on the distance from the cell and the concentration in the micropipette, the actual gradient experienced by the cell may vary. In this case, the diameter of the micropipette is a critical regulator of release rate and is difficult to standardize although commercially available pipets can be more reproducible. Thus the micropipette assay is usually used as a crude qualitative stimulus and is hard to use for accurate dose-response analysis. Chambers which provide both a controlled gradient and allow cell visualization, such as the Zigmond chamber, provide a linear gradient which has limited steepness.^57,58 The assay used by van Haastert et al. requires good manual dexterity, close attention to the assay and is not typically used for high resolution imaging.⁵⁵ An appreciation and awareness of the strengths and weaknesses that each assay provides is important for providing the more detailed dissection of responses that the next phase of research in chemotaxis will require. In particular, assays that provide a reliable quantitative measure of kinetics as well as a range of gradient steepnesses are likely to be critical. Printing methods can allow application of defined gradients of stimuli onto surfaces or gels.^61,62 Alternatively, microfluidic methodologies are promising approaches for providing more controlled stimulation conditions. Recently, the Bodenschatz group has used the microfluidic chamber approach to stimulate single cells with high resolution of spatiotemporal control of the chemical stimulus, combining the approach of photochemical uncaging of the stimulus (cAMP) as well as imaging the translocation of a fluorescent fusion protein (CRAC-GFP) in Dictyostelium discoideum cells.⁶³ The Jeon group has demonstrated an application of the microfluidic chamber system that allows for precise gradient control in biological assays without having to change established cell culture methods by incorporating sets of fluidic channels and vacuum networks for the application of microfluidic gradients onto any type of wet cell culture surface.⁶⁴ Such assays will be valuable in testing and extending theoretical analyses in which the binding characteristics of a receptor with defined affinity and saturation binding properties could respond differently to different types of gradients (i.e., linear vs exponential).^65,66

A Brave New (Bioengineered) World

The initial application of an improved understanding of amoeboid cell motility is likely to focus on enhancing our understanding of disease mechanisms, predicting the course of metastasis and the development of specific drugs to inhibit particular pathologies associated with diseases such as cancer, arthritis or asthma. However, by combining the more detailed understanding of in vivo function and dissection of functional contributions discussed in the previous two sections with the advances in genetic manipulation and genomic data now available, novel redesign of the chemotaxis systems will become possible. This is already occurring in the bacterial chemotaxis realm. In the bacterial systems, parts lists of components that one can plug and play with are becoming available.^67,68 By combining a particular binding module with another enzyme module, a new signaling element can be engineered for a novel pathway connection. Currently, the bacterial systems are designed for work outside of human hosts, such as removing substances toxic to the environment.⁶⁹

One possibility with amoeboid chemotaxis will be to extend this approach to cells moving inside the body. For example, currently some anticancer approaches make use of selection for particular types of immune defense cells in vitro and then reinjecting them into the patients in hopes of mounting a more effective attack on the cancer.⁷⁰ This approach has worked for a small number of patients. But a possible limitation in the effectiveness of this approach is that the immune cells do not know the right place to go. If they are engineered to chemotax towards the tumor cells, detecting combinations of compounds selectively released by the tumor cells, a more effective destruction of disseminated tumor cells may be possible. Similar approaches may work for specific long term infections—for example, designing immune cells to track down tuberculosis may prove advantageous. In vivo models of disease will provide valuable opportunities to test such opportunities. The confluence of in vivo imaging with experimental manipulations based on in vitro mechanistic analysis will allow the direct evaluation of predicted effects on amoeboid cell motility and therapeutic effectiveness simultaneously.

Abbreviations

CSF1

colony stimulating factor

CSF1-R

colony stimulating factor-1 receptor

EGF

epidermal growth factor

EGFR

epidermal growth factor receptor

F-actin

filamentous actin

GC

guanylate cyclase

GFP

green fluorescent protein

GPCR

G-protein coupled receptor

IP3

Inositol triphosphate

MAPK

mitogen-activated protein kinase

MLCK

myosin light chain kinase

NK cells

natural killer cells

PDGFR

platelet derived growth factor receptor

PH

Pleckstrin homology

PI3K

phosphatidylinositol 3-kinase

PIP2

phosphatidylinositol (4,5)-bisphosphate

PIP3

Phosphatidylinositol (3,4,5)-trisphosphate

PLA

phospholipase A

PLC

phospholipase C

PTEN

phosphatase and tensin homolog

RTK

receptor tyrosine kinase

SHIP

SH2-containing inositol phosphatase

VEGF

vascular endothelial factor

VEGFR

vascular endothelial factor receptor

WASP

Wiskott-Aldrich syndrome protein

WAVE

WASP-family verprolin-homologous protein