A Haptotaxis Assay for Leukocytes Based on Surface-Bound Chemokine Gradients Free

Chemokines are small, secreted signal proteins that play a key role in the human immune response, as well as inflammatory and autoimmune diseases and development (1–4). A well-known and thoroughly studied representative is CXCL8 (IL-8). Since its first description in 1987 it has been known as a neutrophil-activating protein that leads to two in vivo effects on neutrophils, chemotaxis and release of granule enzymes (5, 6). CXCL8 mediates its chemotactic effect on neutrophils by binding its two cognate G protein–coupled receptors, CXCR1 and CXCR2 (7–9). Endothelial cells present glycosaminoglycans (GAG) on their cell surface to which CXCL8 is able to bind via its C-terminal GAG binding site (10, 11). Due to the GAG interaction, a stable CXCL8 gradient is formed on the surface of the endothelial cells and activated neutrophils migrate along this gradient toward the site of inflammation (12, 13). Its critical role in acute inflammation (14) makes it an interesting and important target for therapy due to the involvement of CXCL8 in many chronic inflammatory diseases like rheumatism (15, 16), chronic bronchitis (17), multiple sclerosis (18), and Crohn’s disease (19). An increased understanding of the leukocyte response to immobilized chemokines will contribute to the development of new diagnostic tools and therapeutic approaches.

Chemotaxis, that is, cell migration in response to gradients of soluble chemoattractants, is a commonly studied cell response. Several migration assays like the Boyden chamber (20), the Zigmond chamber (21), the Dunn chamber (22), and the under-agarose assay (23) have been established for this purpose. However, most of these assays are liable to limitations such as a lack of the control due to temporal instability of the gradient or the observed cell behavior. Another approach for generating gradients is the use of microfluidic devices that enable the composition of controllable, spatially and temporally defined concentration gradients (24–27). Therefore, it is applied for chemotaxis experiments and allows the study at single-cell levels and in real time (28–30). These assays do not fully reflect the in vivo situation where the migration of neutrophils takes place on the endothelium triggered by the activation of surface-bound chemokines (31). An assay for migration on a surface poses a basis for research of the response of neutrophils to immobilized chemokines and would complement the existing chemotaxis assays. Microcontact printing has been widely used to immobilize protein gradients and patterns on surfaces (32–34) and to determine the influence of immobilized proteins on cell adhesion, growth, and migration (35, 36). It was shown that the biological activity of the chemokine can be preserved upon immobilization on a surface (37, 38). Microfluidic networks can also be used for patterning protein gradients onto surfaces (39–41). The use of hydrogels for microcontact printing instead of polydimethylsiloxane has been previously reported (42, 43). Mayer et al. (44) have shown that gradients of BSA formed in hydrogels by diffusion could be transferred onto surfaces by stamping.

In this study, we demonstrate that continuous chemokine gradients formed by diffusion of the protein in agarose gel stamps can be transferred onto a petri dish surface by stamping. Thus, defined and reproducible protein gradients are obtained and the adhesion, activation, formation of lamellipodia, and migration of cells in response to the immobilized protein can be examined as shown for the example of human neutrophils responding to CXCL8. Therefore, this setup is suitable for a migration assay based on immobilized protein to study haptotaxis of leukocytes in response to surface-bound chemokine gradients. Agarose stamping of diffusion gradients is very simple and inexpensive compared with commercial devices and other established methods like microfluidics, which require complex channel systems and setups. Combined with an image processing and statistical evaluation algorithm, our method yields information about cell behavior at any given time point, and permits the study of single-cell responses and statistical analysis of migration.

The plasmid pET-22b containing the sequence for CXCL8 was transformed into the expression strain Escherichia coli BL21 (DE3) RIL (45). A modified protocol of Wiese et al. (45) was used for the expression and purification of the protein. The cells were grown in lysogeny broth medium with 60 μg/ml ampicillin at 37°C to an OD₆₀₀ of 0.6, and the expression was induced with 0.1 mM isopropyl β-D-1-thiogalactopyranoside. After 3 h of expression at 30°C the cell suspension was centrifuged for 45 min at 4500 rpm and 4°C. The cell pellet was resuspended in buffer A (40 mM Na₂HPO₄, 90 mM NaCl, pH 7.4), supplemented with 1 mM EDTA, 0.2 mg/ml lysozyme, 0.1 mg/ml DNase I, and incubated on ice for 1.5 h. After adding 0.5% Triton X-100, the cell suspension was sonified three times for 30 s (Sonoplus; Bandelin Electronics, Berlin, Germany). After incubation with additional DNase I for 30 min at room temperature the suspension was centrifuged at 4500 rpm and 4°C for 45 min. The supernatant was purified via HPLC on an ÄKTA purifier 10 system (GE Healthcare, Freiburg, Germany) using a 5-ml HiTrap SP FF column (GE Healthcare) with a gradient of 0–30% buffer B (1.5 M NaCl and 40 mM sodium phosphate, pH 7.4) in buffer A. The protein fractions were desalted by a centrifugal filter with a 3000 MWCO membrane (Viaspin 20; Satorius Stedim Biotech). Protein concentration was determined by measuring the absorption at 280 nm.

A 3% aqueous solution of agarose (NEEO; Carl Roth GmbH) was heated in a microwave, cast into the cavities of a 384-well plate (Greiner Bio-One), and centrifuged for 5 min at 800 × g. After cooling at 4°C to harden the agarose, the stamps were gently pulled from the wells with a forceps and cut to a length of 5 mm. The resulting stamping area is ∼3.6 × 3.6 mm. The stamps could be stored in the covered well plates at 4°C for up to several weeks.

A total of 5 μl of a 15 mg/ml protein solution in low-salt PBS was transferred onto a glass slide and the stamp was carefully placed on the drop so that the long side of the stamp was in contact with the surface. After incubation, the stamp was removed, washed with deionized water, and dried. The stamp was carefully placed into a petri dish (35 × 10 mm; Cellstar; Greiner Bio-One) with the stamping area, that is, the short site of the stamp, facing downward (Fig. 1). Light pressure was applied to improve the contact, the stamp was incubated for 60 s, and the stamped area was marked.

For the modified protocol, the stamp was treated as described earlier and after soaking with the protein solution completely into the stamp for 60 min, water was added to the bottom of the stamp. After incubation, the stamp was washed with deionized water and dried. The stamping procedure was performed as described earlier.

The CXCL8 surfaces were incubated with 10 μg/ml FITC-labeled anti-human CXCL8 (Biolegend) in PBS for 1 h. After washing with PBS, the fluorescent Ab was detected by fluorescence microscopy (Zeiss Observer). For the fluorescence imaging a ×2.5 magnification was used that resulted in a detected area of 3581 × 2683 μm². The images were analyzed using the software ImageJ (46) and the Plugin loci_tools. The relative fluorescence intensity across the gradient was calculated and plotted to evaluate the chemokine cross sections.

A total of 5 μl CXCL8 solutions in PBS with defined concentrations was allowed to dry on a stamped agarose spot on a petri dish surface. The dried spots were incubated with 10 μg/ml FITC-labeled anti-human CXCL8 (Biolegend) in PBS for 1 h. After washing, fluorescence was detected by fluorescence microscopy (Zeiss Observer). The fluorescence intensities were obtained by the software ImageJ (46). After calculating the amount of CXCL8 per area (mm²), a calibration curve was plotted with which the amount of immobilized CXCL8 on the gradient was estimated.

Buffy coats from human blood donors were obtained from the blood donor center of the DRK Frankfurt. Equal volumes of buffy coat, PBS, and 6% dextran 500 solution were incubated for 30 min at 37°C. The supernatant was centrifuged for 10 min at 15°C and 240 × g. The pellet was washed with PBS and gently resuspended in 10 ml PBS, transferred onto 30 ml Lymphoprep (Axis-Shield PoC), and centrifuged for 20 min at 4°C and 600 × g with the brake turned off. The supernatant was carefully removed and the red pellet was washed with PBS and resuspended in 5 ml PBS. After adding 25 ml deionized water and waiting for 30 s, 17.2 ml of a 3.6% sodium chloride solution was added and the cell suspension was centrifuged for 10 min at 240 × g. The white pellet was washed with 30 ml PBS and chemotaxis buffer (RPMI 1640 medium containing 0.2% BSA).

Isolated neutrophils were incubated on the petri dish surface for 30 min at 37°C. Afterward the cell suspension was removed, and the surface was washed with PBS thoroughly and covered with 3 ml chemotaxis buffer. The marked area of the gradient was positioned under a phase-contrast objective. For one experiment an objective with ×20 magnification was used, whereas all other experiments were done with a ×10 magnification. The observable area was ∼895 × 671 μm². The sample was aligned so that the steepest increase of the gradient, determined by the cross sections of the gradient, coincided with the vertical of the screen. The cell behavior was observed by taking a picture every 60 s for a time span of 3 h.

Cells on the recorded images were tracked using the software ImageJ (46) with the plugin MTrackJ, and the tracked data were used to generate cell trajectories for statistical analysis of directionality and speed of cells migrating in response to the immobilized chemokine. All further calculations were performed with MATLAB (MATLAB and Statistics Toolbox Release 2012b; The MathWorks, Natick, MA). Because it is difficult to align the microscope perfectly with the gradient, the direction of the gradient may vary by several degrees between individual measurements. To obtain comparable results, in a first step we rotated the obtained cell vectors of one measurement such that the mean angle of all cell trajectories with respect to the y-axis is zero. These modified data were subsequently used for all statistical calculations. Note that this procedure does not affect the migration values; for example, FMI and directness calculated from the data as the exact direction of the gradient are determined this way. The cosθ, the forward migration index in y direction (FMI), and the sideward migration index in x direction (SMI) for one single cell were calculated as follows (47):

The directness for one single cell has been calculated accordingly:

The values given throughout the text are the mean values over all cells within one experiment (12–173 cells) with the SEM for these values.

For a quantitative analysis of chemotaxis in response to a defined gradient, we propose a chemotactic precision index (CPI), which is discussed in more detail in the 11Results and 16Discussion:

The speed of the cells was calculated for each time step, and the SD was determined assuming a normal distribution. The Rayleigh test for the evaluation of the directed migration was performed as suggested by Moore et al. (48). This procedure accounts for different lengths of the cell vectors and weights them accordingly, resulting in more accurate values.

For cell migration experiments, continuous protein gradients immobilized noncovalently on a surface had to be created. To generate these gradients, we used the method described by Mayer et al. (44), who produced gradients of BSA in an agarose gel by diffusion. These gradients were then transferred by light pressure to a glass surface functionalized with aldehyde groups for covalent attachment. This procedure was optimized for our purpose to yield gradients of CXCL8 (Fig. 1). To obtain a suitable agarose stamp for cell experiments in terms of size and smoothness, we used 384-well plates as a mold for the stamps into which the hot agarose solution was cast and hardened (Fig 1, step 1). After pulling the cooled agarose stamps off the well, we obtained cuboids with a quadratic base area of 3.6 × 3.6 mm (Fig. 1, step 2). By incubating the stamp in chemokine solution, a gradient was formed by diffusion in the gel (Fig. 1, step 3). By gently pushing the stamping area of the agarose on a petri dish surface (Fig. 1, step 4), the protein could easily be immobilized due to the transfer of a thin layer of agarose without the need for covalent immobilization (Fig. 1, step 5).

The polystyrene petri dish surface (Cellstar; Greiner Bio-One) is functionalized with hydroxyl and carboxyl groups for hydrophilicity. We assumed that CXCL8 adhered to the negatively charged surface because of electrostatic interaction of its positive charge at physiological pH. However, dried CXCL8 from a drop could not be detected on the petri dish surface after immunostaining, demonstrating that it was not bound on the polystyrene surface. Therefore, binding was associated with the agarose, and only the stamping process with the chemokine in the agarose gel resulted in an immobilized protein on the surface. Hence we assume that the protein is embedded in the agarose transferred to the surface.

For evaluation of cell migration in response to CXCL8, the amount of immobilized CXCL8 on the surfaces was estimated. We prepared a fluorescence intensity calibration curve with defined amounts of CXCL8 per area detected by a fluorescently labeled anti-CXCL8 Ab (data not shown). Using this curve, we were able to estimate the absolute amount of CXCL8 immobilized on the surface and its decrease along the gradient from fluorescence intensities per area after immunostaining.

Due to the different size and properties of CXCL8 in comparison with BSA and the differences in the agarose used by Mayer et al. (44), the optimal diffusion time for CXCL8 within the gel had to be determined. Because of its smaller size we expected shorter diffusion times for CXCL8 in comparison with the times of 4–7 h for BSA reported by Mayer et al. (44). We tested 60, 75, and 90 min for diffusion, and an FITC-labeled anti-CXCL8 Ab was used to detect the chemokine under the fluorescence microscope. The obtained fluorescence images of the CXCL8 gradients were used to compile the cross sections of the gradients for comparison of the gradient shapes resulting from the different diffusion times (Fig. 2B).

The stamping of CXCL8 resulted in homogeneous protein coatings (Fig. 2A). For all tested diffusion times, the cross section of the fluorescence intensity profiles resembled a sigmoidal curve with a plateau on both sides and an approximately linear slope in the center. Interestingly, the selected diffusion times seemed to have no major influence on the inclination of the gradient that ranged between 22 and 23 ng/mm (Table I). For 75- and 90-min diffusion time, a linear area of 500 μm between 2000 and 2500 μm measured from the edge of the stamp area was identified, and we chose the diffusion time of 75 min for testing cell behavior.

To obtain a linear gradient over a larger distance, we used a modified procedure. We soaked the stamp in CXCL8 solution for 60 min (Fig. 1, step 3). After adding water to the bottom of the stamp, we extended the incubation time to 2 or 3 h (Fig. 2D, 2E). This resulted in linear gradients over almost the whole width of the stamp between 0 and 3600 μm, but with a more shallow inclination of 0.008 or 0.005 ng/mm at 2 or 3 h respectively (Table I). For cell experiments, we decided to apply the gradient of 2-h incubation time.

To test the homogeneity of the stamped protein surfaces and the reproducibility of the stamping procedure, we generated five stamped surfaces by the use of one single stamp (Fig. 2C). The SDs of the cross sections of the five gradients were relatively low, which indicated good reproducibility. One stamp could be used to produce several immobilized gradients on one or several surfaces. Furthermore, to test the reproducibility of the stamp fabrication and the formation of the gradient, we compared three surfaces stamped with different stamps prepared on different days (Fig. 2F). The SDs were slightly higher compared with the results of repeated stamping but demonstrated the good reproducibility of the whole stamp and gradient fabrication procedure.

The protein is immobilized noncovalently by the deposited agarose so that some CXCL8 and/or agarose may dissociate and diffuse from the surface if the protein gradient is covered by cell culture medium for several hours during haptotaxis experiments. Therefore, we tested the stability of the immobilized gradient in medium by evaluating the fluorescence intensity profiles before and after 3 h of incubation. The cross section of the fluorescence intensity after incubation with medium showed a similar curve shape as the untreated CXCL8 gradient, but the inclination of the fluorescence intensity gradient decreased to almost half the previous value, from 23 to 13 ng/mm (Fig. 3). To ensure that the loss of CXCL8 on the surface did not affect cell behavior and cell migration, we selected the linear region between 1500 and 2500 μm for observation of the cells because in this region the gradient was retained during the 3-h experiment.

For haptotaxis experiments, we applied the well-studied cell system of primary human neutrophil granulocytes that have two G-coupled receptors, CXCR1 and CXCR2, for interaction with the chemokine CXCL8. Neutrophils show a strong response after activation with CXCL8 and are able to migrate along an appropriate gradient with high velocity. These cells are very sensitive, they become activated quickly during the isolation process, and their characteristics are dependent on the physical state of the donor (49, 50). Nevertheless, neutrophils are widely used for migration studies and provide significant results if a sufficient number of independent experiments are carried out.

We first tested neutrophil behavior on the petri dish surface and on the pure agarose. Over the course of 60 min, neutrophil granulocytes exhibited the same behavior on the plain petri dish as on the agarose-coated surface (data not shown). Cells could adhere, spread over the surface without showing polarization, and show activation by formation of lamellipodia. A few cells migrated randomly but remained within a maximum radius of 15 μm. The only difference between the two surfaces was that some cells lost interaction with the agarose surface during the experiment.

Experiments with neutrophils were then performed on immobilized CXCL8 gradients to ensure that the immobilized amount of protein was sufficient to activate the cells, to test whether the conformation of the immobilized protein was appropriate to activate its receptors, and to determine whether the protein distribution would be recognized by the cells as a gradient. On the shallow CXCL8 gradient built by the modified procedure (Fig. 2E), the cells adhered and showed activation by the formation of lamellipodia, but no different cell behavior in comparison with the petri dish or agarose surface was observable (data not shown). Observation of the cells adhering to the sigmoidal gradient built by the method of Mayer et al. (44) revealed that the number of adherent cells depends on the amount of CXCL8 on the surface (Fig. 4). On the plateaus of the gradient, no directed cell migration could be observed, but the activation was visible by the formation of lamellipodia. This may be explained by the shallow concentration gradient in these regions or too high or too low CXCL8 concentrations for induction of migration. It is well-known that chemotaxis is dose dependent and follows a bell-shaped curve whereby the cells migrate less at low or high concentrations of CXCL8 (51, 52). In the area of the linear slope, migrating neutrophils could be observed. Therefore, the immobilized sigmoidal CXCL8 gradient formed by a diffusion time of 75 min proved to be suitable to induce migration of neutrophil granulocytes and was chosen for the following haptotaxis experiments.

Three independent cell experiments were conducted to evaluate haptotaxis of the cells. Cell responses were observed over a time span of 3 h. The linear region of the gradient between 2000 and 2500 μm was observed. The initial parameters, consisting of an observed area of 448 × 336 μm² and time intervals of 5 min between every frame (Fig. 5C), proved to be not ideal. The migration could not be followed over the whole time of 3 h because of migration out of the visual field of several migrating cells. Therefore, in the following independent experiments, a surface area of 895 × 671 μm² was observed with time intervals of 60 s over 3 h. This procedure resulted in a better resolution of the migrating cells (Fig. 5D, 5E).

After rotating the obtained cell vectors, the cell trajectories from all experiments indicated a directed migration along the CXCL8 gradient that proceeded along the y-axis. It is noticeable that the paths of the migrating cells toward higher CXCL8 amounts were very straight and hardly any haptokinesis was observable (Fig. 5).

The Rayleigh test (48) for the evaluation of the migration vectors in uniform circular distribution indicated levels of significance for all experiments, with p < 0.001 corresponding to a specific migration along one direction.

The migrating fractions of adherent cells lay between 30% for the first two and 50% for the third experiment. Nonmigrating cells adhered to the surface, but they were uniformly distributed and not polarized (Fig. 5B). They were tethered to the petri dish surface without any observable migration. In comparison with the adherent cells on the petri dish surface, the nonmigrating cells were more spherical and less spread out, illustrated by the size of the cells (Fig. 5A, 5B). The migrating cells were polarized and showed a leading edge consisting of lamellipodia followed by the cell body and a terminal uropod (Fig. 5B).

For statistical analysis of chemotaxis, diverse chemotactic indices are presented in the literature based on the assumption that the observed phenomenon is chemotaxis rather than chemokinesis. The evaluation of haptotaxis occurs equivalent to chemotaxis as generally two-dimensional systems are examined.

Agrawal et al. (49) and Tharp et al. (53) described a chemotaxis index that results from the cosine of the angle θ between the direction of the gradient and the mean vector of migration (Fig. 6, Eq. 1). This value ranges from −1 to +1, whereby +1 indicates perfect migration along the gradient (54, 55). This index exhibits two disadvantages. First, the direction of the gradient has to be known exactly to avoid major errors. Second, the evaluation of chemotaxis occurs on the basis of a small angle. If there is less migration along the y direction, corresponding to the gradient direction, a small value for θ will result as well and the index will incorrectly indicate strong chemotaxis. In addition, in the case of random migration, that is, chemokinesis, the average of this angle will approach 0 and consequently cos θ will approach 1. Taken together, this value is neither suitable to decide whether there is chemotaxis nor can it be used for quantitative analysis.

Related problems occur with the FMI (Fig. 6, Eq. 2) (28, 47, 56, 57). This index is simply a measure of the migration along the gradient and disregards the migration in the perpendicular direction. Another approach uses the directness as a measure of chemotaxis that considers only directed migration (Fig. 6, Eq. 5). Thus, it bears these problems like the FMI with the disregard of the x direction of migration as well (58). Directness weights the migration in x and y direction equally, and only information about the ratio of the path length in x and y direction is considered, whereas the overall direction of migration itself is not examined. In the case of completely linear movement along the gradient, both FMI and directness are 1. However, high values of FMI in the range of 0.7 do not sufficiently prove migration along the gradient. For example, an FMI of 0.7 and a value of 1 for directness are obtained for a high value of 0.7 for the SMI in x direction. Hence the direction of the migration would not be along the gradient, but along a path in a 45° angle to the gradient. Accordingly, the migration in x direction has to be included as well. Only high values of FMI together with low SMI values perpendicular to the gradient demonstrate highly directed migration in response to a given chemoattractant gradient. The FMI and the directness contain important information, but not the precision by which the cells follow the direction of a given gradient.

The chemotaxis coefficient based on the Keller and Segel model for chemotaxis is much more detailed because it includes the concentration of the chemoattractant, the gradient steepness, and other cell-dependent parameters by which chemotaxis can be described (59, 60). The coefficient is useful but very specific. Due to its dependence on the exact experimental setup it cannot be used to compare chemotaxis in different setups.

To compare the chemotactic precision, we propose a CPI as the product of the cosine of the angle between the vectors of FMI and directness with the FMI and one minus the absolute value of the SMI (Fig. 6, Eq. 6). The cosine of the angle ϕ specifies whether the obtained data are a result of chemotaxis as small angles represent directed migration along the gradient. This angle is multiplied by the FMI, which gives information about how strong migration along the gradient is. By additional multiplication of one minus the absolute value of SMI, information about migration in x direction is included in the CPI. The angle between the FMI and the directness could be replaced by the angle θ used in the literature if the direction of the gradient is exactly known (54, 55). In the case of chemokinesis, only random movements are observed and the angle may adopt any value, whereas the FMI will exhibit a low value and the CPI will be low as well.

Possible limiting cases are provided to illustrate the weakness of the directness and the FMI as chemotaxis indices in comparison with the CPI (Fig. 7B). Case 1 shows a directness of 1 with an FMI of 0.2 that results in an SMI of 0.98 and a CPI of 0.0008. This case shows that directness alone is not sufficient to draw conclusions about chemotaxis, because the main direction of migration distinctively continues along the axis of abscissae, perpendicular to the gradient. It is further shown that an FMI of 0.2 (as published by Lin et al. [28]) does not indicate chemotaxis, but neither excludes it. Nevertheless it is not possible to quantify chemotaxis by utilization of the directness or the FMI. In case 2, a directness of 1 with an FMI of 0.7 is shown, which results in an SMI of 0.7. In this case, a preferred direction of migration exists, but this direction is rotated by 45° rather than the gradient direction itself. The directness and FMI both have high values; however, there is no chemotaxis along the gradient, as indicated by a CPI of 0.14. Case 3 indicates that low values of directness result in low values for the FMI and subsequently also for SMI. Therefore, the CPI is very low, classifying cell movement as no chemotaxis. In case 4, the directness has a value of 0.35, as in an example reported by Sackmann et al. (58). An FMI of 0.3 results in an SMI of 0.18, which leads to a low CPI of 0.21. These data reflect the limits of CPI. In this case, directed migration is present, but it is too weak to produce satisfactory results in a chemotaxis assay and the CPI as well. Case 5 confirms the limits of the CPI as weak chemotaxis leads to low values for the CPI. With a directness of 0.2 and an FMI of 0.2 (SMI is 0), the CPI has just a value of 0.2.

Calculating these values for our experiments provides values for the directness in the range of 0.66 to 0.78, clearly indicating a preferred direction of migration (Fig. 7A). Likewise, the high values for FMI in the range of 0.64 to 0.75 clearly emphasize directed migration along the gradient. For the SMI, we obtained very low values between 0.03 and 0.06. Using the values for directness FMI and SMI, the CPI was calculated by Eq. 6 (Fig. 6). For the first experiment with only 12 cells a value of 0.68 was obtained, whereas the values for the second experiment with 31 cells was 0.71 and for the third with 173 cells was 0.59. Due to similar values of directness and FMI for the experiments, it is not surprising that the values obtained for the CPI do not differ much from each other.

The speed of neutrophils provides information about the magnitude of chemotaxis/haptotaxis. Therefore, the speed distributions of the cell migration experiments were compiled (Fig. 8). The mean speed of the cells lay between 3 and 4 μm/min, whereas the range of the speed was between 0 and a maximum of 10 μm/min. During the whole experiment no significant change in the average speed over time was observed (Fig. 8D), so the cells were apparently not affected by the decrease of the CXCL8 amount described earlier. Therefore, sufficient CXCL8 remained on the surface for activating neutrophils.

We applied the method from Mayer et al. (44) to build protein gradients in agarose gels to form surface-bound chemokine gradients. With adjusted parameters, gradients with defined, reproducible shape were generated. We obtained a steep inclination but a short linear region. By extending the linear region, the inclination became shallower. Thus, the shape of the gradient can be varied via the incubation protocol. However, the variability of the printable shape method is limited. Microfluidic devices are more versatile because they allow building almost any shape of soluble gradient in which the linear region and the inclination can be accurately adjusted (25, 26, 60). However, designing and manufacturing the respective gradient generator in micrometer scale may be time consuming and requires specialized equipment and experience, whereas the diffusion gradient stamping method is very simple and inexpensive in processing. In particular, there is no need for special equipment. As opposed to classical methods like the Boyden (20), Zigmond (21), and Dunn chambers (22), the stamping method leads to defined immobilized gradients that are reproducible in shape. The temporal and spatial stability is not entirely ensured as the CXCL8 amount on the surface decreases with exposure to medium over 3 h. Although the shape of the gradient remains similar, the absolute amounts of CXCL8 decrease. The diffusion could possibly lead to a soluble gradient in close proximity to the surface and in this way exert influence on the activation and migration of the cells. In vivo the situation is similar as the cells are exposed to an immobilized gradient, as well as to a soluble gradient (10, 61). Cells exposed to these surfaces recognize these gradients and do not alter their velocity as absolute CXCL8 amounts on the surface decrease while the gradient is maintained.

The high inclination of the gradient obtained by using the method of Mayer et al. (44) seemed to be very important for cell behavior because the shallow gradient with an inclination of 0.008 ng/mm was not able to elicit a response in neutrophils different from the uncoated surface. It is known that the neutrophil migration is dependent on the inclination of the gradient and the CXCL8 concentration: with microfluidic channels, a concentration range between 0 and 5–500 ng/ml over a linear gradient region of 500 μm led to CXCL8-induced neutrophil migration (53, 60). Thereby a range of 0 to 50 ng/ml equates to an inclination of 100 ng/mm on a gradient of 500-μm length. The inclination of 23 ng/mm obtained by our method is roughly one quarter of the inclination obtained by microfluidics that led to chemotaxis of neutrophils (60), but it was sufficient to induce migration as opposed to the shallow gradients with an inclination of 0.008 ng/mm (55). The immobilization strategy mimics the in vivo situation of leukocyte migration in response to surface-bound chemokines as the chemokine is immobilized via GAGs on endothelial cells.

Notably, only a part of 30 to 50% of the cells migrated along the gradient, whereas the rest were still viable but without polarization (Fig. 5A). The cells showed activation of CXCL8 by formation of lamellipodia but in its spherical morphology. The CXCL8 gradient could not induce cell migration in these cells or the cells were not able to recognize the gradient. In the first reported chemotaxis experiment with CXCL8 in a multiwell chamber assay, Yoshimura et al. (5) already observed that just 54% of the neutrophils migrated at the optimal CXCL8 concentration of 10 nM. Keenan et al. (62) made similar observations on cell culture dish surfaces where a significant part of 38% of all cells showed no migration but changing motility and formation of lamellipodia. This could be an indication that the surface influences the cells in a way that prevents some cells for migrating by forming a stronger binding between the cell and the surface. In transwell assays to study the chemotaxis of neutrophils in response to CXCL8, we obtained similar results. The fraction of migrating cells lay between 30 and 70%. The reason for the nonmigrating cells is unclear, but there are some possible causes. We assumed that the sensitivity and quick activation of the neutrophils, as well as the physical state of the donor, determine how many cells are susceptible to CXCL8 (49, 50). Furthermore, adhesion proteins play an important role and mediate leukocyte migration (63, 64). The influence of adhesive ligands on cell migration depends sensitively on several integrin–ligand interactions like ligand levels, integrin levels, and their binding affinities (65). By the integration of adhesive ligands on the gradient surface of our assay, a larger fraction of cells might be able to migrate and this issue will be a matter of further research.

For statistical analysis of haptotaxis, migration trajectories were compiled and a chemotaxis index for the evaluation had to be selected. Evaluation of commonly used indices like the angle θ, the directness, and the FMI showed that these indices could not give quantitative information about the precision of chemotaxis, because they do not take all parameters of migration into account (see Fig. 7B). Therefore, we introduce an alternative, the CPI for the evaluation of chemotaxis. The CPI includes all necessary information of migration: the angle φ between the directness and the direction of the gradient, as well as the FMI and the SMI. For our experiments, we obtained CPI values ≥0.6, indicating highly precise migration in the direction of the gradient and showing that the developed haptotactic assay poses a powerful tool to investigate haptotaxis.

Our CPI bears several advantages over the indices reported in the literature. First, the value is simple to calculate based on the established chemotaxis indices: directness, FMI, and SMI. Second, the resulting value between 0 and 1 allows an intuitive evaluation of the precision by which cells follow the direction of a given gradient; high values indicate strong and precise chemotaxis, whereas low values show weak chemotaxis and may hint chemokinesis. Third, the index is independent of the underlying chemotaxis assay if the obtained data permit the calculation of the index. Nevertheless, the CPI has its limits: the case of weak migratory responses, when the readout of a migration assay may not be meaningful and an alternative assay should be used.

In addition, the speed of the cell migration during the first, second, and third hour of the experiment was considered, as well as the speed over the whole experiment time span. Frevert et al. (66) reported similar speed distributions for neutrophils in a range up to 10 μm/min. The mean speed of the cells lay between 3 and 4 μm/min, and these results were in agreement with the literature, which describes average values of 2.5 μm/min (66) and speeds between 2 and 10 μm/min depending on the slope of the gradient and the CXCL8 amount (28, 53, 60, 62).

In summary, we have demonstrated a very simple method for studying haptotaxis of leukocytes in response to surface-bound chemokine gradients. The transfer of diffusion gradients in agarose gels to surfaces works well for the chemokine CXCL8. By using time-lapse microscopy, information on the single-cell level of a large number of cells can be gathered easily and statistical data about this cell population can be derived. Only small amounts of cells, protein, and other reagents are needed, making this assay easy and inexpensive to perform.

The assay is very versatile and gives the opportunity to test the influence of almost every protein gradient immobilized onto a surface to cells and their behavior. The correlation of migratory cell behavior with the gradient shape or protein concentration and surface modifications can be investigated, as well as changes in cell behavior in the presence of inhibitors. In addition, the influence of soluble chemoattractants on cells following surface-bound gradients can be studied (29, 57, 67). The presented model system of neutrophils responding to CXCL8 is only one example of the wide field of chemokines guiding the migration of immune cells. Because chemokines are not the only determinant of immune cell migration, the role of other chemoattractants like complement factor 5a (68), inflammatory cytokines such as TNF (69), growth factors like CSF-1 (68), or modulating factors such as adhesion molecules like selectins (70) or integrins (64) could be investigated with this system. The directed migration of cells is involved in inflammation, development of the immune system, survival strategies of pathogens, and allergic reactions. Therefore, an easy method to study immobilized gradients of individual proteins in vitro is a useful tool to better understand their individual impact or to confirm hypotheses derived from in vivo observations. In gradients produced by this method, the proteins are immobilized by noncovalent interactions without any modifications so that the method can be performed with recombinant proteins, as well as with native proteins purified from tissues or body fluids. Another interesting aspect of this assay system is the possibility of studying the interplay of surface-bound protein gradients and soluble factors. Only a few years ago, reports have pointed out the different effects triggered by surface-bound protein gradients and soluble gradients (71), and by now this phenomenon has only been investigated for a limited number of model systems. A lot of research still needs to be done, and a straightforward and inexpensive method to produce surface-bound gradients with unmodified proteins will add to the progress in this area.

Finally, the proposed CPI is suitable for quantification of cell migration, chemotaxis, and haptotaxis. It is independent of the application as long as the migration data can be recorded on a single-cell level in a time-resolved manner. It indicates the precision by which cells follow the direction of a given gradient of chemotaxis much better than the standard values used in the literature.

We thank Anke Imrich for protein expression and Marina Joest and Kevin Brahm for scientific discussion.

The authors have no financial conflicts of interest.