A major part of this work was the development of a microfluidic chamber that allowed us to study the traction forces exerted by the platelets under flow condi-tions. Here, we chose to embed a protein-coated PAA gel into a PDMS channel using a plasma binding technique. In particular, the PAA substrate was coated with fibrinogen before it was treated in the plasma cleaner. To avoid damage to the protein and gel, during plasma exposure, the substrate was covered by a thin PDMS stripe. After bonding, the channel and the connecting tubings were filled with water for storage. Depending on the storage technique, we were able to store the devices up to one week.
While the measurement of forces under flow conditions is still an emerging subject in research, some microfluidic chambers have already been described in literature. None of them were constructed similar to the one we used here al-though some come closer to our construction than others. Let us here have a look at other devices and illustrate why we chose the approach we did. The most in-teresting questions to ask are concerning the combination of substrate and PDMS channel as well as the choice of method for the protein-coating. We are also shortly discussing the usage of an external mixing device versus alternative approaches.
For combining the substrate and the channel device, we chose a method where we first produced the PAA substrate similar to the static experiments and
sep-Design of Microuidic Device 8.2 arately cast a PDMS device as often done for different microfluidic devices and
measurements. By employing a similar substrate than we had used prior for the static experiments, we were able to use already derived analysis methods as well as directly compare the results from both experiments to each other. The PDMS casting from a silicon waver we utilised to create the channel had the advantage that we were able to produce several copies of the channel with identical geom-etry by reusing the corresponding waver. These were fabricated in-house in a clean room. Here, the height of the wavers was controlled by different production protocols thus rendering the set-up potentially quite flexible. As an alternative, as done by Steward et al. [112], we could have incorporated the substrate into a commercial available microfluidic platform. However, this diminishes the vari-ability of the geometry we achieved by fabricating the devices ourselves and the idea was thus discarded. Devices constructed by binding a glass slide to a self-made PDMS cast have been employed in other studies by Perraultet al.[92], Das et al.[17], Huret al.[47], Myerset al.[81] and Lembonget al.[63]. For the binding, they either only oxidised the PDMS cast and press bonded the device [17], used customised holders to hold the chamber together [47, 92], silicon adhesives [81] or bound the glass slide and PDMS before the PAA substrate was polymerised [63].
For a precursor of our chamber, we also tried to just activate the PDMS by plasma treatment [96]. However, these chambers proved to not be entirely leak-proof and generally did not sustain for the entire preparation and experimental time of sev-eral hours. Our alternative to prepare all components beforehand and oxidising both the glass and the PDMS cast while protecting critical regions proved to be much more reliable. Indeed, as long as no air pockets remained between PDMS and glass and no part of the substrate was underneath the PDMS, no problem with leakage was observed. At the same time, this procedure did not require any additional tools to facilitate binding. Moreover, an extra layer of adhesive would inadvisedly change the height of the whole channel, changing the flow conditions within.
The microfluidic device that came closest to our device is the one designed in terms of the assembly process by Lamet al.[57]. Here, they incorporated a micro-post array in a PDMS device. In particular, they coated the array with adhesion proteins before plasma treatment of both the glass containing the array and the PDMS cast. During plasma treatment, similar to us, the proteins on the posts were protected by a piece of PDMS which was removed before binding. The prominent difference between their technique and our method was the use of the
Chapter 8 DISCUSSION
force measurement method. As we mentioned previously, the disadvantage of a micro-post array is that data may only be collected at discrete points in space contrary to continuous substrates. While this works well for larger cells which still adhere to a sufficient number of posts, this does not hold true for the smaller platelets. An additional effect to considered in the set-up used by Lam et al. is the distortion of the flow field. As they simulated for their device, it was shown that the flow field along the posts was not constant in space in contrast to our device. They demonstrated that the shear stress was considerably higher on the posts than in between them. For our device, the flow field along the gel was constant, guaranteeing that attached platelet experienced the same shear stress along its entire circumference only distorted by its own geometry. Hence, we argue that our flow regime resembles more the actual conditions within the body than a shear stress system of larger unevenness as seen in micro-post arrays. For comparison, their chamber had the same height as our chamber (100 µm) while their posts were 9µm high.
The next question to consider is why we chose to add the protein before the plasma treatment instead of after the binding of the chamber. Indeed, for the binding of fibrinogen to the substrate, it did not mattered whether the gel under-went plasma treatment first or not. In literature, both approaches exist of either first coating the substrate with protein or first assembling the chamber. The former method was mostly used in the chambers that we described previously containing some type of substrate or micro-contact patterning [18, 39] while the latter method was mostly utilised when handling plain glass slides [34, 82]. The most important reason to not coat the substrate with fibrinogen after the binding was that fibrino-gen reacts to high shear forces. When exposed to high forces, fibrinofibrino-gen elongates and changes its elastic properties [4, 68], an effect which does not reverse directly after force reduction. After binding, to fill the chamber with protein solution, the solution had to be injected via a thin tubing, considerably thinner than the actual chamber. Hence, the shear stresses within the tubing are higher than what we simulated for the chamber. We could thus not exclude that the fibinogen would react to the high shear stress within the tubing. To avoid any unnecessary stresses, all fibrinogen solutions were handled with utmost care. This included the usage of as large pipettes as possible, no vortexing during mixing and no flow through thin tubes. Hence, coating before the plasma treatment was chosen. This also resulted in an easier handling of the coating step.
When we investigated the adhesion of the platelets to the substrates, we
ob-Design of Microuidic Device 8.2 served that the platelets did not attach when choosing a flow rate of 1000 µL/h,
corresponding to a shear rate of about 50 s−1. The same effect was reported by Ranke [96], where the flow had to be stopped to facilitate adhesion. Once the platelets were attached, the flow did not detach the platelets again. The same observation was done by Myerset al. [81], where a shear rate of 100 s−1 did not effect the platelets as long as they were already attached before the start of the flow. According to previous experiments [34, 103], at this flow rate, the reaction to fibrinogen predominates the adhesion. Indeed, Van de Walleet al.[18] studied the attachment to spaced fibrinogen patches under a shear rate of 100 s−1. This raises the question why the platelets did not attach when exposed to a lower shear rate.
Several aspects need to be considered here which all influences the experiments.
First, the reaction between fibrinogen and the integrinαIIbβ3 is a comparably slow reaction (in the order of ms to s for stable binding [67]). Hence, to facilitate attach-ment, sufficient time has to be given for the reaction in terms of dwelling time. In the human body, it has been shown by experiment [1] and simulation [118] that the attachment is aided by the red blood cells actively pressing the platelets to the vessel wall and arresting them there. In our experiments, no such particles were included. Previous studies to the adhesion under flow, apart from Ranke [96], all used whole blood samples instead of purified blood plasma as we did here.
Furthermore, all devices used in these works employed hard glass surfaces in-stead of softer substrates. As has been shown in other studies [93], platelets react much stronger to glass surfaces than softer substrates in terms of adhesion rate and spreading area. Softer substrates may well reduce the chances of attachment.
While the natural environment in the human body is softer by nature than glass, other mechanisms help the attachment along here. As the platelets still attached to the substrates when we used a lower flow rate or stopped the flow, the problem did not lay in the protein coating.
Before continuing to our results, let us briefly remark on our choice of an ex-ternal mixing device to combine the thrombin and platelets. We chose to activate the platelets by introducing thrombin into our experiments. To be able to compare our two different experiments, the concentrations between the components were kept constant. Hence, we had to inject the thrombin at the correct concentration into the platelet solution. This could be done by either using one syringe with the complete mixture, mixing before the measuring chamber or mixing within the device itself. The first approach was used by Ranke [96] and seemed to work quite well. However, the platelets were essentially already activated within the syringe,
Chapter 8 DISCUSSION
making it difficult to approximate the exact time between activation and attach-ment or start of the experiattach-ment. Additionally, the reaction time between the cells and trigger substance statistically increased with increasing recording time, i.e.
platelets entering the chamber at a later time point were exposed longer to throm-bin. For our static experiments, this time could actually be set to about 2 min plus some time for the diffusion of thrombin and platelet within the liquid. Mixing within the measuring chamber lead to the effect that platelets become activated and directly flushed out of the system. Hence, an external mixing device was cho-sen. This had the advantage that the reaction time between the two components was well approximated and controllable by the length of the connecting tubing between the mixing device and the chamber.