3.1 LoCs design and functions
This work aimed to realize a low-cost and easy-to-use device for cancer cells or EVs trapping and characterization. As a proof of concept, we selected oral squamous cell carcinoma (OSCC) screening by early detection of carcinogenic cells from saliva, a biological fluid readily available for a non-invasive analysis, and EVs samples isolated from neuroblastoma cell lines. In this view, we realized three devices with three different configurations specifically designed to optimize the functionalization and sample preparation required before reaching the final goal of cancer cell detection by microfluidics. Polymethylmethacrylate (PMMA) substrates were used for device fabrication designed by Solidworks CAD software (SolidWorks Corporation, 300 Baker Avenue, Concord, MA, USA) and realized by a Mini-Mill/GX micromilling machine (Minitech Machinery, Norcross, GA, USA). A computer-aided manufacturing (CAM) software was used to transfer the information from the Solidworks CAD software in machine code file for the micromilling control. The first multi-channel testing device was designed to optimize the functionalization process described below. This device is equipped with nine different channels on the same substrate with dimensions of 400 µm and inlet/outlet holes for ElveFlow micropumping system connection. This device was designed to fit in a single substrate and compatible to be simultaneously checked within a single field of view while performing the experiment (as shown in Fig. 2). In this way, it was possible to test several experimental conditions on the same device and optically monitoring them at the same moment (Fig. 3a).
While single cells can be easily identified once captured in the transparent microchannels, even in brightfield mode, the capture of EVs requires a different way of acquiring images. Indeed, this can be achieved by obtaining fluorescent spots, organized as an array. To easily identify them, we designed and realized the geometry of poker suits pitches, connected by microfluidic channels, each of them functionalized with a different antibody against EVs membrane antigens (Fig. 3b).
As a final setup, we realized the sample-in/answer-out device in a multiple-function platforms. It was characterized by an oval-shaped reservoir for fluids mixing (spiked artificial saliva and buffer for dilution) connected to a 150 µm serpentine channel for cancer cells detection (Fig. 3c). To enhance the mixing effect in the oval mixer, we added a pillar in the centre of the chamber (Supplementary Movie 1). In this way, by creating turbulence, fluids were forced to mix before reaching the serpentine area of the device where binding probes will allow the capture of cancer cells.
The serpentine channel, emerging from the mixing chamber, is designed to optimise the catching process of tumor cells. For this purpose, we implemented a device with a serpentine-shaped channel with 20 half loops 150 µm wide that allowed cells to have a long path and related functionalized area to maximize the possibilities of being captured by immobilized probes. This is of crucial importance in real samples, in which rare or few cells are expected to be present in specimens to be analysed.
All the devices were assembled using a recently developed solvent- and thermal-assisted protocol 17, which allow for robust sealing and a plug-n-play connection to a micropumping system through capillary tubes.
3.2 Devices functionalization
Cancer cells and EVs detection was based on the ability of antibodies against membrane antigens to selectively capture this class of biological entities. In particular, the anti-EpCAM antibody can recognize and bind the membrane antigen expressed by OECM-1 cells. EpCAM is a cell adhesion molecule recently discovered to be an oral carcinoma cell-surface marker protein and a therapeutic target 19. By immobilizing antibodies against this protein in the channels device, we were able to selectively capture OECM-1 cell lines in the biological fluid tested, specifically artificial saliva.
The same principle was used for the selective identification of EVs. In this case, the areas of the poker suit device were functionalized with an antibody against Flotillin-1, an integral membrane protein expressed on the membranes of both large- and small-EVs 20,21. However, the binding of antibodies to PMMA substrate is possible only by overcoming the high hydrophobicity of the plastic material, therefore a functionalization step is necessary before biomolecule immobilization. In particular, we tested two functionalization methods with polydopamine and APTES/glutaraldehyde in the multi-channel testing device.
We first tested different polydopamine concentrations in Tris-buffer solution at pH 8.5, at different concentrations and incubation time-points at room temperature as listed in Table 1. The reaction involves catechol groups allowing the polydopamine molecule to adhere to PMMA surface.
Table 1
Polydopamine concentration expressed in mg/ml used for the device functionalization in relationship with incubation time.
| Channel | Polydopamine concentration mg/ml | Incubation time |
| 1 | 1 | 1 h |
| 2 | 1 | 30 min |
| 3 | 1 | 15 min |
| 4 | 0.5 | 1 h |
| 5 | 0.5 | 30 min |
| 6 | 0.5 | 15 min |
| 7 | 0.2 | 1 h |
| 8 | 0.2 | 30 min |
| 9 | 0.2 | 15 min |
The polydopamine working conditions (concentration and time) that generated the most efficient capture surface turned out to be 0.5 mg/ml for 1 hour of incubation. However, polydopamine is characterized by quick polymerization process that resulted in the formation of many debris within the channels (Supplementary Fig. 2). This motivated our choice to investigate another method of functionalization.
The second functionalization we tested includes the subsequent incubation of APTES (0.01%) and glutaraldehyde (0.05%). APTES (0.01%, 1 hour at room temperature) allowed the establishment of covalent bonding between PMMA substrate and APTES ammino-groups, preparing the reactive groups for glutaraldehyde binding. After a wash with absolute ethanol, the incubation is performed with glutaraldehyde (0.05%, 40 minutes at room temperature). This reaction implies that the glutaraldehyde exposes the aldehyde groups for antibody binding. The last functionalization step was the immobilization of the primary antibody anti-EpCAM (Sigma-Aldrich, 1:500 in PBS) overnight at 4°C, followed by the anti-mouse secondary antibody FITC-conjugated (Sigma-Aldrich, 1:500 in PBS for 1 hour at room temperature). The APTES/glutaraldehyde functionalization resulted to be the most suitable functionalization for antibody immobilization as reported in Supplementary Fig. 3, the channels were not affected by the presence of debris and residues of polymer after functionalization and the FITC signal of the secondary antibody was revealed by fluorescence microscope after cell capturing. The presence of immobilized cells was confirmed by DAPI staining, identifying the nucleus of cells (Supplementary Fig. 3).
3.3 Artificial saliva sample preparation
We also exploited the advantage of multiple channels on the same chip in the multi-channel testing device to determine the optimal artificial saliva dilution to flush into the final device.
To improve the surface chemistry of the PMMA, the 9 channels were functionalized with APTES and glutaraldehyde as previously described. Then, we tested artificial saliva and related dilutions. The problem with undiluted artificial saliva was the disturbing presence of several bubbles and debris. This set of experiments aimed to find the most suitable concentration not affecting biorecognition between suspended cells and immobilized antibodies. The NeutraSal artificial saliva was reconstituted as mentioned in Material and Methods and diluted at the following concentrations: 75%, 50%, 20%, 10%, and 5% in PBS. Each solution was flushed in a different channel of the multi-channel testing device and analysed under optical microscope. 75% of artificial saliva solution in PBS resulted to be the best sample dilution as it limits bubble formation and did not release excess of debris into the channels.
Subsequently, we exploited the multi-channel testing device also to understand the limit of presence of cells in 75%-artificial saliva, to find the best dilution able to mimic the real sample manipulation, identifying cells singularly trapped on the surface of channels, without any interference from bubbles or debris. We tested five different OECM-1 cell concentrations: 500,000; 250,000; 100,000; 10,000 and 1000 cells/ml and we found that the highest concentrations (500,000; 250,000 and 100,000) were unsuitable for microchannels injection and detection, since the suspensions tend to produce many bubbles and release high amounts of debris in the device, hindering both biorecognition and optical detection. On the other side, the 10,000 and 1000 cells/ml concentrations, allowed single-cell passage into the channel, without generating debris inside the device (Fig. 4). For our final purpose of cancer cell detection from saliva samples, which is supposed to be very few cancer cells/sample, we proceeded in the next experiment with the lowest cancer cell concentration 1000 cells/ml of 75%-artificial saliva by flushing 100 µl at the same time within the device.
3.4 Extracellular vesicles detection in Poker suits device
The flexibility of our trapping approach was then tested to detect and analyze diverse and complex biological samples. In particular, we tailored this system to the capture and analysis of EVs. The EV samples were isolated from a second cancer cell line SH-SY5Y, to show the possibility to extend the technique to different cell lines. EVs samples from SH-SY5Y were a mixture containing both EV subclasses as characterized by Western blot (Fig. 5a), showing the differential expression of EV markers common to both EV types (Flotillin-1) or specific to large (GRP94) and small (CD63) EVs. We tested the ability to detect the different EV types using the Poker Suits device. In this case, the APTES (0.01%) and glutaraldehyde (0.05%) functionalization of the device was followed by the immobilization of the primary antibody anti-Flotillin-1 which allows the capture of all the different EV populations. The specific detection of the different EV classes was achieved by flowing different fluorescence-labelled antibodies against the particular markers of the various EV types, namely FITC-conjugated IB4 and APC-conjugated Annexin V for large EVs and PE-conjugated anti-human CD63 for small EVs, into the different poker-suit chambers. As shown in Fig. 5b, we were able to capture the EVs inside the device and specifically detect the different EV populations (lEVs and sEVs) by the immunofluorescent signals imaged in each poker suit.
3.5 Sample-in/answer-out device
To demonstrate the ability of our device to work as a system for the on-chip sample manipulation and analysis, we included all the optimized functionalities into a final setup which was designed including two parts: (i) an oval-shaped reservoir equipped with two different inlet holes for cells in artificial saliva solution and diluent fluids which converge in the oval reservoir for mixing; (ii) a serpentine with multiple loops with 150 µm dimension and 130 mm of length, for cancer cells detection directly connected to the reservoir. An additional channel directly connected to the serpentine (as indicated in Fig. 6) allows the preliminary functionalization of the serpentine, by isolating the rest of the device by simply tightly sealing the two inlets of oval mixer (Fig. 6).
To establish flowrates of fluid mixing to obtain the dilution of saliva as assessed in the formerly described protocol for sample preparation, a blue and a red solution were used, simulating saliva and PBS respectively and monitoring the mixing under an optical microscope. To enhance the mixing effect and to avoid the production of laminar flow which can underline the efforts to obtain the preparation of the sample directly on-chip 14, a pillar in the oval shape was placed, able to induce turbulence in the flow. As can be seen in the attached pictures and movie (Supplementary video 1), the two solutions can be efficiently mixed by tuning the flow rates and the general speed of flow is afterward slowed down by the presence of a serpentine, allowing the capture of cancer cells.
By tuning the flowrates, we identified the fluid velocity and pressures allowing the sample dilution into the oval mixer. To obtain the previously optimised dilution of artificial saliva, we injected the red solution (mimicking artificial saliva) at a flow rate of 50 mbar and the blue solution (mimicking PBS) at a flow rate of 150 mbar. The fluids mixed in the reservoir up to the desired concentration (around 70% based on the ratio of the flow rates) as shown in Fig. 7 by fluid colors.
The serpentine portion of the device was instead independently functionalized by simply sealing the other holes. Using only inlet/outlet holes connected to the serpentine, we injected firstly 0.01% APTES/ 0.05% glutaraldehyde; subsequently, the primary EpCAM antibody was added as previously described. The long serpentine with multiple loops that characterize this device allowed us to have a large surface area coated with the EpCAM antibody against the membrane antigen expressed by OECM-1 cells, thus amplifying the possibility of cancer cell detection in the device.
Once the serpentine channel was functionalized and the flow rates established, we prepared a spike suspension of 1000 OECM-1 cells in artificial saliva. We injected then the sample mimicking real saliva and PBS at the defined flow rates to obtain a dilution of the specimen at 75% based of flow rate ratio. After mixing the fluids in the oval mixer, the flow proceeded to the serpentine previously functionalized, lowering its velocity thus maximizing the possibilities for biorecognition events to happen. The final step was the immunofluorescence staining as previously described to be able to observe OECM-1 cells trapped on the channels’ walls: green fluorescence signal shows FITC-conjugated antibody binging OECM-1 cells trapped by EpCAM antibody functionalization of the device and blue fluorescence signal shows cells nuclei stained with DAPI (Fig. 8).
This procedure can actually be useful in the detection of cells from different biological samples and even of different cancer cells by different membrane antigen immobilization, enabling the device to be considered a point-of-care tool for the detection of cancer cells by optical analysis.