Skip to main content
European Commission logo
English English
CORDIS - EU research results
CORDIS
CORDIS Web 30th anniversary CORDIS Web 30th anniversary
Content archived on 2024-06-18

Self-assembled virus-like vectors for stem cell phenotyping

Final Report Summary - SAVVY (Self-assembled virus-like vectors for stem cell phenotyping)

Executive Summary:
The project aimed at integrating a number of nanoscale components via hierarchical self-assembly into Raman active particles and evaluating their usability in cell sorting applications. We have made excellent progress towards the nanoscale self-assembly of a complex array of functional nanoparticles at multiple length scales. The base particle architecture was achieved by an electrohydrodynamic co-jetting process that can produce Janus vectors with distinct hemispheres (i.e. the compartments). In this design, both the interior as well as the surfaces of the vector’s hemispheres can be controlled independently. Anisotropic surface-directed self-assembly of amphiphilic gold NP onto the Janus vectors constitutes another critical component. The remaining surface area of the vector particles is modified with polyethylenglycol to impart particle stealth. The third functional element in the form of plasmonic nanostars ensures high Raman-enhancement (SERS).

We established human pluripotent stem cell cultures and appropriate differentiation protocols in formats that are directly accessible by Raman microspectroscopy.
We successfully established all functional process steps of the microfluidic-based system, such as particle/cell separation, cell singularization, Raman analysis and cell sorting and translated them into independent microfluidic modules. These single modules were successfully characterized and one module for each functional step was selected as preferred design.
Even though partner WITec joined the consortium only during month 19 of the project and the work done by the previous partner OMT had to be repeated, a prototype of the cell sorter with line illumination has been successfully built by WITec. Our initial Raman results support the overall feasibility of the SAVVY cell sorting concept: (i) It could be shown show that a very low laser power is sufficient to obtain reasonable good Raman signals, even for short integration times. (ii) We confirmed enhancement of the Raman signal of 1-2 orders of magnitude as compared to unlabeled particles. (iii) Although the final step of a fully integrated hardware system comprised of the microfluidic chip AND the microscope in one unit was not realized, we still could distinguish two different cell types based on their Raman signature inside of a microchannel.
While this project has validated the fundamental approach taken by the SAVVY consortium, further engineering work would be necessary to establish a integrated hardware solution that would be sufficiently robust for larger-scale cell sorting.

Summary description of project context and objectives
The planned work of the SAVVY project was divides into 6 scientific workpackages (WP1 through WP6), one workpackage that was concerned with training of students and dissemination (WP7) and WP8 that dealt with all administrative questions and general management of the project.
WP1: WP1 was concerned with all questions regarding the hierarchical self-assembly of SAVVY reporters. The main objectives for this workpackage were the establishment of a hierarchical self-assembly co-jetting of SAVVY reporter components and the assembly of SAVVY reporters with three configurations. Comparative analysis of the hierarchical assembly efficiency and reliability were an important part of the project as was the comparative analysis of SAVVY reporters’ attachment efficiency to model membranes and of their Raman signals. SAVVY aimed at the synthesis of the next generation of membrane fusing nanostars and the generation, characterization and efficiency analysis of simplified SAVVY reporters. After the critical choice of optimal SAVVY design the SAVVY reporters were used to fully analyze a cell line in correlation with the full proteomic a genomics analysis. The final goal was to design striped NP that fuse with cell membranes with optimization of their size to achieve highest interaction strength which was measured by AFM.
WP2: The main goal of WP2 was the SAVVY-enabled cell phenotyping. In order to reach this goal the consortium established the following objectives: Fully reproducible stem cell culture platforms were established and the cell phenotype was characterized using stage specific markers. An experimental protocol for SERS-based cell phenotyping using SAVVY reporters was developed. In addition an optimized classification system that enables high throughput phenotypic classification was developed. An important part of this workpackage was the establishment of a phenotype classification software that can control the SAVVY process.
WP3: Workpackage 3 (Automation 1: Microfluidics for SAVVY sorter) centered around the establishment of a microfluidic device that can be used as an interface between the fluidic control station and the Raman microscope. In order to reach this goal the the following objectives were put in place: Individual functional modules were developed to perform the various assay steps from NP manipulation to the Raman detection and cell handling in microfluidic devices. Establishment of a fluidic control station which contains all necessary elements to drive the microfluidic modules and development of an instrument interface of the fluidic control station to the Raman detection unit to combine the microfluidic assay with the detection unit. The microfluidic modules were then combined into an integrated microfluidic cartridge. The SERS detection conditions and integration within microfluidic chips were optimized.
WP4: This workpackage (Automation 2: High performance Raman microscopy for SAVVY sorter) was engaged in the development of a high performance Raman microscope. The main objectives to reach this ambitious goal were as follows: It was planned to optimize the Raman measurements on a test system of individual polystyrene (PS) beads (with/without
SERS enhancement) and then fully develop the high performance Raman microscope for SERS measurements on cells. As next step the integration and test of the “phenotype classification software (SW)” developed in WP2 and the control of the microfluidic system in the RAMAN instrument controlled by the SW for the SAVVY sorter were established.
WP5: The goal of Workpackage 5 was the process engineering of the SAVVY reporters. In order to reach this goal the consortium planned to first optimize the fabrication process of the reporters (fine-tuning of size, shape, surface patches, surface chemistry, and loading) to reach a high uniformity of the particles. Additionally it was planned to scale-up the Janus vector preparation based on electrohydrodynamic co-jetting. The synthesis and purification of striped NP was automated in order to achieve a polydispersity of below 10%. Fine-tuning of fabrication processes was planned to ensure size, shape, and concentration of plasmonic NP with optimal amplification of the SERS signal. In a last step the optimization of the surface chemistry of all constituent particles was ensured to maximize the SERS effect and interaction with the membrane molecules.
WP6: The last scientific workpackage of the SAVVY project dealt with the integration of all components developed in the other 5 workpackages to eventually develop the SAVVY instrument allowing for Raman analysis of SAVVY labeled cells. The workpackage had three major objectives: (i) On chip validation of cell labeling with SAVVY reporters, (ii) on chip validation of Raman based cell classification and (iii) on chip validation of ultra-fast cell sorting.
WP7: This workpackage, called Dissemination, Training & Exploitation had the objective to establish potential applications and markets for the results of the project, to coordinate dissemination, to carry out technology transfer, and to prepare the dissemination and the exploitation plan. In more detail the consortium planned to disseminate the project results to the general public as well as to scientific and research communities; to raise public awareness of the SAVVY project and its new technology; to develop an exploitation plan of project results; to train young scientists in using the new technology; and to enhance young people’s interest in science.
WP8: This workpackage was concerned with the overall management of the project.

Project Context and Objectives:
1. Goal of Project

The Self-Assembled Virus-like Vectors for Stem Cell Phenotyping (SAVVY) project used hierarchical, multi-scale assembly of intrinsically dissimilar nanoparticles to develop novel types of multifunctional Raman probes for analysis and phenotyping of heterogeneous stem cell populations.
Human stem cells have great potential for a broad range of therapeutic and biotechnological applications, such as regenerative medicine or pharmaceutical testing, but the characterization or sorting of stem cell populations has been extremely challenging and is generally addressed using flow cytometry. However this approach is vitally hampered by: (i) lack of specific antibodies, (ii) need for fluorescence markers that limit multiplexing, (ii) low concentration of stem cells in the abundance of background cells, (iv) low efficiencies and high costs due to need for antibodies. The SAVVY approach uses intrinsic differences in the composition of cell membranes to distinguish and ultimately sort stem cell populations.
Compared to the current practice (e.g. fluorescence-based multiplexing), SAVVY uses a fundamentally different approach, which (i) does not require antibodies, aptamers, or any other biomarker, (ii) is fluorescence-label free, and (iii) is thus scalable at acceptable cost. Because of the fundamental differences to the currently used antibody-based detection schemes, SAVVY provides a paradigm shift for stem cell phenotyping with applications in medical diagnostics.
The project integrated live cell sorting system incorporates nanoparticles-based signal enhancers along with Raman micro-spectroscopy in an integrated microfluidic cell sorter (‘SAVVY sorter’). This required preparation of fundamentally novel Raman probes (‘SAVVY reporter’) based on the design and hierarchical self-assembly of constituent nanoparticles.
This project integrates a number of nanoscale components via hierarchical self-assembly into Raman active particles and evaluated their usability in cell sorting applications. Most importantly, we made excellent progress towards the nanoscale self-assembly of a complex array of functional nanoparticles at multiple length scales. The base particle architecture was achieved by an electrohydrodynamic co-jetting process that can produce Janus vectors with distinct hemispheres (i.e. the compartments). In this design, both the interior as well as the surfaces of the vector’s hemispheres can be controlled independently. Anisotropic surface-directed self-assembly of amphiphilic gold NP onto the Janus vectors constituted another critical component. The remaining surface area of the vector particles is modified with polyethylene-glycol to impart particle stealth. The third functional element in the form of plasmonic nanostars ensures high Raman-enhancement (SERS).
In parallel, excellent progress was made towards establishing human pluripotent stem cell cultures and appropriate differentiation protocols in formats that are directly accessible by Raman microspectroscopy. We successfully established all functional process steps of the microfluidic-based system, such as particle/cell separation, cell singularization, Raman analysis and cell sorting and translated them into independent microfluidic modules. These single modules were successfully characterized and one module for each functional step was selected as the preferred design.
Even though WITec joined the consortium only during month 19 of the project, a prototype of the cell sorter with line illumination was successfully built by WITec. The initial Raman results support the overall feasibility of the SAVVY cell sorting concept: (i) It could be shown that a very low laser power is sufficient to obtain reasonably good Raman signals, even for short integration times. (ii) We confirmed enhancement of the Raman signal of 1-2 orders of magnitude as compared to unlabeled particles. (iii) Although the final step of a fully integrated hardware system comprised of the microfluidic chip AND the microscope in one unit was not realized, we still could distinguish two different cell types based on their Raman signature inside of a microchannel.
While this project has validated the fundamental approach taken by the SAVVY consortium, further engineering work would be necessary to establish an integrated hardware solution that would be sufficiently robust for larger-scale cell sorting.

2. Component Fabrications for the SAVVY Reporter
The design of the SAVVY reporters is comprised of a core 500 nm particle immobilized with rippled nanoparticles for cell attachment and nanostars for SERS enhancement. As such, below are the three individual components that were designed, fabricated, and characterized for the final assembly: (i) rippled nanoparticles, (ii) nanostars, (iii) and Janus particles.

2.1 Synthesis of Rippled Nanoparticles
To synthesize rippled nanoparticles, an alcohol-based process has been established that works at room-temperature and allows for mechanical mixing. After developing the synthetic protocol that gives consistent characteristics of nanoparticles in terms of surface monolayer and size, we tried to scale up the synthesis to get 1 g per synthesis. One of the key parameter in the synthesis is the mixing of the solution prior to addition of reducing agent. Vigorous mixing with big stirring bars in round bottom flasks could easily overcome that barrier to get homogeneous distribution of the precursors. The characterization of the striped nanoparticles at a small-scale and a large-scale synthesis has been compared in terms of size and ligand ratio. The comparison, revealed that there is no significant difference between the two synthetic approaches. The approach of large batch synthesis and parallel reactors is therefore viable.
The amount of rippled particles is not the limiting step. Less than a milligram could be enough for testing a couple of milligrams of polymer beads and nanostars. The small size of the rippled nanoparticles implies higher concentration (higher number of nanoparticles for a given volume), even for a small amount of mass. The current protocol that yields 1 g of nanoparticles per day would be more than enough for massive production of SAVVY reporters for the future. Details regarding the synthesis of the rippled nanoparticles are below.
Striped nanoparticles have been produced with 11-mercaptoundecane sulfonate and 1-octanethiol. To this mixture, gold salt (HAuCl4) in ethanol and the thiol ligands in methanol were added, followed by a saturated solution of sodium borohydride (NaBH4) in ethanol. After sedimentation overnight, the nanoparticles were washed by centrifugation several times with ethanol, methanol, and acetone consecutively. Finally the precipitate was left under vacuum to dry completely.
Rippled nanoparticle could also be specifically synthesized to contain desired ligands (such as N-hydroxysuccinimide, amine carboxylic acid or azide ended thiol ligands), which allowed the application of different chemistries for the attachment to nanostars and polymeric particles.
During synthesis, gold nanoparticles are formed as heterogeneous mixtures of different sizes and ligand densities. This leads to part of the batch being poorly covered with water-soluble ligands and thus, agglomerating and being insoluble in water. Therefore, it is sometimes necessary to purify the nanoparticles from poorly soluble aggregates and even nanoparticles that have very big or very small sizes. To achieve this goal, we applied a density gradient ultracentrifugation method also known as fractionation. The technique is widely used among protein biochemists and cell biologists to purify proteins or separate cellular compartments from each other. Analytical ultracentrifugation (AUC) was used to separate the gold nanoaparticles by sizes. Transmission electronic microscopy (TEM) analysis confirmed the monodispersed nanoparticle fractions obtained.
A mixture of hydrophilic and hydrophobic organic molecules is used to cover the surface of small gold nanoparticles through gold – sulfur bonds. During functionalization of gold nanoparticles the same chemistry is used binding gold and sulfur. This implies a similar kinetics for functional ligands and MUS/OT ligands that are mainly covering the surface. Therefore, it can be safely assumed that storage quality of the nanoparticles does not alter upon functionalization. The characterization of functional groups on gold nanoparticles is conducted by NMR. The peaks are assigned to each corresponding molecule and quantitative estimation is conducted based on the integrations of the peaks. 10:1 molar ratio (MUSOT:N3 linker) is consistently found over three replications of functionalization experiments.

2.2 Synthesis of Nanostars
The approach here was to work on a 100 mg batch synthesis of gold nanostars (AuNSs) and use an increase in reaction volume as a scale-up-strategy. 30 mg of gold nanostars can be obtained using 500 mL synthesis. The performance of several syntheses per day or the increase of the volume per synthesis, are two viable methods for obtaining the desired amount of gold nanoparticles. Fabrication details are provided below.
AuNSs were prepared by a modified seed-mediated growth method. Briefly, the seed solution was prepared by adding a citrate solution to HAuCl4 solution under vigorous stirring. After a brief boiling period, the solution was cooled down to room temperature and kept at 4 °C for long-term storage. The as-synthesized Au nanoparticle seeds had an LSPR maximum at 519 nm. For AuNSs synthesis, citrate-stabilized seed solution was added to HAuCl4 solution containing HCl in a glass vial at room temperature under moderate stirring. Quickly, AgNO3 and ascorbic acid were added simultaneously to the above solution. The solution rapidly turns from light red to green indicating the formation of AuNSs. Immediately after synthesis, the solution was stirred with PEG-SH, washed by centrifugation, and redispersed in water. The synthesized AuNSs present a main Localized Surface Plasmon Resonance (LSPR) located around 750 nm and a less intense shoulder at lower wavelengths, corresponding to the contributions of the tips and the nanostar core, respectively. Gold nanoparticles, which feature multiple sharp tips and present core sizes around 35 nm, were successfully obtained.
To functionalize the AuNSs with azide groups, the AuNS were incubated overnight with N3-PEG-NH2 after the washes, in order to target a theoretical surface modification of 95 molecules per nm2. This amount of ligand per nm2 should ensure surface saturation, as previously reported.1 Typically, gold nanoparticle functionalization requires the removal of free ligands after reaction. While more sophisticated separation methods are needed for clusters or small nanoparticles, for larger gold nanoparticles such as gold nanostars, washing by centrifugation is a well known procedure to eliminate the unreacted ligands.2 After functionalization the particles were stable, as confirmed by the UV-Vis spectra and the TEM characterization. Since further steps, which mainly involve click reaction with PLGA nanoparticles, has not been affected by the possible lack of reproducibility of the AuNS functionalization, we consider this a good strategy to properly modify the gold surface of the plasmonic nanoparticles.

2.3 Fabrication of Janus Particles
The goal in this section of the project was to fabricate Janus particles with distinct patches that could be used for the immobilization of gold nanostars and rippled nanoparticles. The particles were fabricated using the Electrohydrodynamic (EHD) Co-Jetting procedure developed and were analyzed with several different characterization tools, including SEM, confocal microscopy, NTA, and DLS before being used to create the SAVVY assemblies. The details of the fabrication and characterization of the Janus vectors are below.
Janus PLGA particles were fabricated through the EHD co-jetting procedure, as previously established in the Lahann group.3-5 Briefly, during EHD co-jetting polymer solutions are pumped through syringes tipped with metal needles in a laminar regime, while an electric field is applied to the needles. The solutions form a Taylor cone at the tip of the needles and from this droplet a polymeric jet is ejected towards the collecting electrode. The jet splits into a spray of droplets, where the solvents evaporate rapidly, leaving behind polymeric particles on the counter electrode. In order to fabricate particles on a gram basis, a roll-to-roll system for the EHD co-jetting was designed that is capable continuously synthesizing nanoparticles (Figure 11). Based on this system, we are able to fabricate 0.5 g of nanoparticles per hour, for a total of 50 grams per day, which easily supersedes the requirements of this project. To collect and purify these particles for further studies, the particles were collected in sterile DI water with tween 20 and fractionated via centrifugation into more monodispersed populations.
Through various experiments, it has been determined that the best methods for Quality Control (QC) of the Janus particles before/after surface modification/immobilization is to test each batch with SEM/TEM and DLS to determine their size distribution, shape, and level of nanoparticle immobilization.
In order to quantify the ligand density on the surface of the particles, a procedure established by the Lahann lab was used (Rahmani, et al, 2015 Journal of Drug Targeting).6 Nanoparticles with an alkyne functional group on the surface were fabricated using the EHD co-jetting technique. After centrifugation, the particles were characterized with Nanosight to determine their size distribution as a function of their concentration and were then reacted with different amounts of azide-PEG-FITC via copper catalyzed click chemistry overnight. The nanoparticles were centrifuged to separate the particles from the unreacted material and the unreacted azide-PEG-FITC was isolated and measured via UV Visual Spectroscopy to determine the amount of unreacted material based on a previously established calibration curve. Based on this information, as well as on the size distribution and concentration measurements, the total number of PEG molecules reacted per surface area of the nanoparticles was determined. Here, two sets of particles, one set with functional groups on both hemispheres (i.e. monocompartmental) and one set with functional groups just on one side (i.e. bicompartmental) were used. In the case of the monophasic particles, the ligand density increased with increasing PEG concentrations. The same trend was observed for the bicompartmental nanoparticles, with the distinction that at higher concentrations the ligand density of the bicompartmental nanoparticles is significantly lower (3.67 ligands per squared nanometer) due to the fact that only one hemisphere is covered with the ligand.
To determine the best methods for the handling and storage of the Janus particles, several approaches have been examined. For short-term storage, the particles can be kept in a 4oC storage unit for up to 1 month before use in cellular experiments. Based on SEM imaging and DLS, such particles retain their shape, morphology, immobilization, and size distribution for this period. It is recommended that the particles be briefly tip sonicated and tested with DLS before each use. For long-term storage, the Janus particles need to be freeze-dried after the various immobilization steps and stored in the freezer. To use, the particles can simply be resuspended in DI water and tween 20, briefly tip sonicated, and used. As a precautionary step, the particles should be analysed via DLS to ensure their successful resuspension. Based on recent analysis, the freeze-drying of particles does not harm the particles and results in well dispersed particles upon resuspension. The particles retained their average size and concentration after the freeze-drying step.

3. Assembly of SAVVY Reporters
Once the individual components of the SAVVY reporter were fabricated, their assembly into the SAVVY reporters were investigated. Two approaches were pursued, where one involved using polystyrene (PS) beads as the particle to immobilize the rippled nanoparticles & nanostars on, and the second used the Janus particles discussed above for the immobilization. Both approaches are discussed in detail below.
3.1 PS Assemblies
Amino-functionalized polystyrene particles can react with gold nanostars and rippled gold nanoparticles. The high affinity of gold and the amino groups led to a particularly stable structure. The non-detachment of the gold nanoparticles from the polymer surface in the presence of cell culture medium confirmed the strong interaction between the particles and the nanostars.
The development of a strategy to modify the loading of nanostars onto the microparticles was also carried out. For PS beads, by keeping constant the amount of gold nanostars and decreasing the final concentration of the polymeric particles, we were able to obtain polymer particles covered with gold nanostars.
In conclusion, we have developed several strategies to obtain different coverage of gold nanostars onto the polymer surface. Cellular uptake experiments have been conducted using the hierarchical assembly of SAVVY reporters with the combination of amounts that give the best coverage and stability. HeLa cell lines were used for these experiments and the assembly was incubated with the cells at 37 °C for 1 hour. Confocal microscopy images showed that the assembly incorporated successfully within the cell membranes. While PS beads - nanostars assembly samples showed only background fluorescence, the addition of striped nanoparticles clearly proved the presence of SAVVY assembly on the cells.
3.2 Janus Assemblies
To fabricate Janus particles for covalent immobilization of the rippled NP and nanostars, one side of the particles needed to contain alkyne groups, while the second side needed to contain groups to reduce the non-specific association. As such, the particles were used to immobilize a hydroxyl star-PEG to the hemisphere containing the carboxyl groups via EDC/sulfo-NHS chemistry to reduce the non-specific attachment of the SAVVY reporters, while the second hemisphere containing the PLGA-alkyne groups were left unmodified.
The particles were analyzed with DLS to determine their size distribution before and after the surface modification, which demonstrated a consistent size of about 500 nm (the final size of particles was at 580 nm). As can be seen, the size distribution narrows after the surface modification as the particles tend to aggregate less after PEG-ylation. Copper catalyzed click chemistry was then used to immobilize nanostars and rippled nanoparticles on the second hemisphere.
For the immobilization of the gold nanostars, the Janus particles containing PEG on one hemisphere were incubated with nanostars functionalized with an azide-PEG, washed, and imaged via SEM and STEM to determine the selective immobilization of the stars on one hemisphere. While the concentration of the Janus particles and the nanostars, as well as the washing steps, were optimized, a formulation was developed for the selective immobilization of the gold nanostars on one patch.
The same strategies as with the nanostar attachment were used for the immobilization of rippled nanoparticles on the Janus particles. Janus particles containing alkyne groups on one side and carboxyl groups on the second side were fabricated using EHD co-jetting. The particles were fractionated into 500 nm particles via serial centrifugation and analyzed via dynamic light scattering. The particles were PEG-ylated on the carboxyl side via EDC/sulfo-NHS chemistry with an eight-arm amine PEG. For the immobilization of amphiphilic nanoparticles, PEG-ylated Janus vectors were incubated with the amphiphilic NPs functionalized with azide groups, washed multiple times to remove all unreacted material, then imaged via TEM imaging to determine their selective immobilization.
Finally, to demonstrate the selective immobilization of nanostars and amphiphilic nanoparticles on one hemisphere of Janus particles, a sequential set of chemistries was used. First, the same route described above for the immobilization of the gold nanostars was used, followed by several washes to remove all unreacted material. The Janus particles, now immobilized with nanostars, were then incubated with the amphiphilic nanoparticles for the second copper catalyzed click reaction (same procedure as above). The particles were then washed numerous times to remove the unreacted material. The size of the Janus particles increases slightly after PEG-ylation, and significantly more after the nanostar and rippled nanoparticle attachment. Additional TEM images confirming the attachment of the nanostars and the amphiphilic nanoparticles to one side of the Janus vectors

4. Fabrication of the Microfluidic Device
In the course of developing the whole SAVVY platform, each functional module as well as the functional interaction between two or more modules were optimized in a continuous way of process. This procedure is tightly associated with the stepwise integration of the modules forming a more and more complete SAVVY platform including the fluidic control station and the necessary interface. The cartridge has slide dimensions (75.5 mm x 25.5 mm) and contains Luer interfaces for liquid supply and connection of several microfluidic modules or separate functional elements on chip. All interfaces except the first inlet have a special geometry with a big hole to reduce the risk of losing cells. The first inlet is also a Luer but with a standard changeover hole. This hole is smaller to support a pipette.
Three membranes were assembled on the polymer chip. A bondfoil from top and below seals the channels except the holes for the quartz glass unit. The quartz glass unit is made out of two quartz glasses which are assembled with a double adhesive tape. This double sided adhesive tape creates and seals the channel as well. An additional double sided adhesive tape is necessary to mount the quartz glass unit to the main cartridge. This cartridge system was used for the Raman spectroscopy analysis of various cell types.

5. Application of SAVVY Assemblies for SERS Detection
The assembled SAVVY vectors were tested to determine their compatibility with cells and their ability to be used for SERS signaling. Here, both the PS assemblies and the Janus PLGA-based assemblies were tested and both demonstrated the ability to produce SERS signaling with SERS active molecules and with HeLa cells.
For both systems, the particle characterization with cells produced heat maps of relevant peaks corresponding to the positions of SERS events. For our Raman spectroscopic studies, we refined our Raman spectroscopic imaging protocols, which have also been aided by the optimisation of the particle- and cell-based work. Raman spectroscopy imaging experiments were performed on a confocal Raman microscope from WITec alpha300R+. The Control FOUR 4.0 software was used for measurement and WITec Project FOUR 4.0 Plus for spectra processing. The Raman spectroscopy images were 30x30 μm with a 750 nm spatial resolution and a 1 sec integration time. Furthermore an in-house library in Python was used to estimate the pixel effects assuming that each image follows an additive model of the form where the intensity γwp for wave number w and pixel p can be decomposed (apart from the constant Y) into a component Aw specific to the wave number w (and common to all spectra), and a component Bp particular for a spectrum p but shared across wave numbers. The remaining intensity unexplained by the model is represented by the residuals ewp. Such a model was applied to each Raman image through the robust Median polish algorithm, where we can disentangle the effects of the image location (equivalent to the “pixel” or “spectral” effects) from the effects of the wave number dimension. The resulting residuals can be compared across images, and furthermore this same model can be applied to concatenated images.

5.1 PS Assemblies
To begin, the SERS capability of PS-SAVVY particles (immobilized with gold nanostars and rippled nanoparticles) were tested. The SERS spectra of PS-SAVVY particles were measured with Rhodamine 6G (R6G) in various concentrations and SERS signals down to 10-3 M using 0.1 mW laser power on the samples with 10 sec integration time could be detected.
Once the SERS activity of the particles was established, the use of the PS-SAVVY particles for SERS detection in various cell types was explored. We have performed Raman spectroscopic analyses of HeLa cells, C2C12 cells and neural stem/progenitor cells (NSCs) in the presence of only nanostars, PS beads with nanostars, and PS beads with nanostars and striped nanoparticles (PS-SAVVY).
Therefore, we have characterized the PS-SAVVY nanomaterials and their constituent components in the presence of two cells lines (i.e. HeLa and C2C12) as well as two human stem cell-derived neural cell types. In each of these cases, we successfully obtained cellular-nanoparticle attachment and SERS for each cell type. These studies embodied a characterisation of the interaction between the PS-SAVVY reporters and NSCs to better understand the adhesion process and optimise the coverage of the SERS reporter on the cells for the final application of SAVVY-enabled sorting using a live-imaging approach, incubating the reporters with the plated cells and acquiring fluorescent and bright-field images over the course of the experiment. Once the detection of SERS signalling was established for several different cell types and the optimum particle characterization for cell adhesion was determined, we moved on to using the developed system for the differentiation of distinct cells based on SERS signaling and downstream linear discriminant analysis.

5.2 Janus Assemblies
Similar to the PS assemblies, the Janus assemblies were also first tested with a Raman reporter to determine their SERS activity, before testing them with cells. The Raman reporter 4-MBA at a concentration of 10.10-5 M under a 785 nm laser illumination was used with Janus particles. The nanostar-immobilized Janus particles were SERS active and contained the expected Raman signatures.
Next, the particles were used to detect SERS signaling with HeLa cells. Here, the particles were incubated for 5 hours, after which they were analyzed with a confocal Raman spectroscope. While a distinct difference can be observed between the Janus particles with the nanostars versus the controls without any nanostars, the SERS activity was not as strong as those of the PS beads. In this case, this was to be expected since the particles did not contain the rippled nanoparticles and hence did not attach at the same concentrations. As such, these preliminary results were deemed to be promising and further testing for the differentiation of distinct cells based on SERS signaling and downstream linear discriminant analysis was done.
6. Use of SAVVY Assemblies for Cell Phenotyping
We have been able to create a comprehensive software package for phenotypic prediction/classification of SERS signals. The developed Python module can be used by our partners within SAVVY to analyze data from any Raman spectrum and the data outputs from SAVVY. With input from WITec, the phenotype classification software was integrated into the system through the WITec Control software. This software enables SERS spectra to be measured and classified in real-time through a hit quality index (HQI) between the sample and a set of reference spectra. To further extend the classification capability associated with the complex SERS signals, sophisticated algorithms were developed, which were incorporated into a general hyperspectral analysis toolbox to be employed in a high-performance computing environment. This toolbox is comprised of the Python module and iPython Notebooks to ensure reproducibility of all relevant analyses and allowed for the construction of a comprehensive library of signature spectra that can be exported to the phenotype classification software in order to ensure optimal classification performance.
The developed Raman spectroscopic framework has been implemented as a graphical user interface (GUI) under the WITec Control software. The software has been integrated into the SAVVY Raman microscope prototype that can classify SERS spectra according to the HQI similarity index. Interaction and data exchange between the Raman control software and the phenotype classification software have thus been achieved.
The optimization of the classification capability associated with the complex SERS signals has been achieved by developing more sophisticated preprocessing and deconvolution algorithms, which can be analyzed through the Python module and then used to inform the phenotype classification software. Besides the usual baseline and background correction algorithms, a background removal algorithm has been created to further improve the usual baseline and background correction algorithms. After preprocessing, all images from the same cell type can be analyzed together.
Using these pooled images an endmember extraction algorithm (implemented in the Python module) can be applied to explore possible reference spectra. First both NFINDR and MCR-ALS on each cell type were applied separately to estimate a set of cell type-specific endmembers, and then we extracted one component from each cell type using all endmembers together, such that the dissimilarity is maximized. The resulting three spectra compose our signature library that can already be fed into the phenotype classification software. Using these reference spectra we are able to simulate the behavior of the actual HQI-based phenotype classification sorter.
Once the phenotyping programing was well established, the SAVVY sorter was validated by performing detection of SERS signatures within the SAVVY chips of two different cell lines (HeLa cells and C2C12), using the polystyrene based reporters with rippled nanoparticles and nanostars (PS-SAVVY), the fully assembled Janus carriers based reporters (J-SAVVY). A positive interaction was observed for both PS-SAVVY and J-SAVVY with both cell lines as previously established. Each cell line was associated with a unique SERS spectral signature that can be attributed to the various proteins and lipids comprising the cell membrane. We performed a time-series real-time acquisition that provided insights into the dynamics of the cell sorting. These results showed that SERS spectra could be measured over a longer time period, up to 100 sec.
A two-component principal component analysis (PCA) integrated with linear discriminant analysis (LDA) was then performed on this data. The two first components from the PCA model accounted for most of the variance and were able to discriminate the two cell lines. A supervised LDA classification further confirmed the differences between spectra. LDA scores of the C2C12 cell line were consistent for both the PS-SAVVY and J-SAVVY suggesting that the SERS signals have a degree of reproducibility. Interestingly the PC1 and PC2 scores for HeLa cells showed larger variability suggesting that the PS-SAVVY and J-SAVVY interact differently with the HeLa cells.

Conclusions
There are still no Raman spectroscopy-based techniques available on the market for cell sorting and signature-based characterization. However this project has validated and assessed the feasibility of such an approach. The SAVVY reporter components have been optimized by KIT and CICbiomaGUNE in 2015, which resulted in the validation of a prototype that was validated under investigations using the SAVVY instrument (assembled at Imperial, with input from WITec and ChipShop) which comprises a Raman spectrometer that can enable microfluidic characterization integrated with a phenotype classification software. The progress achieved within this project demonstrates that this approach is feasible for cell sorting and phenotyping and can undergo further optimization and studies to realize its translational potential.

References:

1. Serrano-Montes, A. B.; de Aberasturi, D. J.; Langer, J.; Giner-Casares, J. J.; Scarabelli, L.; Herrero, A.; Liz-Marzan, L. M., A General Method for Solvent Exchange of Plasmonic Nanoparticles and Self-Assembly into SERS-Active Monolayers. Langmuir 2015, 31 (33), 9205-9213.
2. Sperling, R. A.; Parak, W. J., Surface modification, functionalization and bioconjugation of colloidal inorganic nanoparticles. Philos T R Soc A 2010, 368 (1915), 1333-1383.
3. Roh, K. H.; Martin, D. C.; Lahann, J., Biphasic Janus particles with nanoscale anisotropy. Nat Mater 2005, 4 (10), 759-763.
4. Bhaskar, S.; Pollock, K. M.; Yoshida, M.; Lahann, J., Towards Designer Microparticles: Simultaneous Control of Anisotropy, Shape, and Size. Small 2010, 6 (3), 404-411.
5. Bhaskar, S.; Roh, K. H.; Jiang, X. W.; Baker, G. L.; Lahann, J., Spatioselective Modification of Bicompartmental Polymer Particles and Fibers via Huisgen 1,3-Dipolar Cycloaddition. Macromol Rapid Comm 2008, 29 (20), 1655-1660.
6. Rahmani, S.; Villa, C. H.; Dishman, A. F.; Grabowski, M. E.; Pan, D. C.; Durmaz, H.; Misra, A. C.; Colon-Melendez, L.; Solomon, M. J.; Muzykantov, V. R.; Lahann, J., Long-circulating Janus nanoparticles made by electrohydrodynamic co-jetting for systemic drug delivery applications. J Drug Target 2015, 23 (7-8), 750-8.

Project Results:
The SAVVY project provided a proof-of-concept for the assumption that various stem cell populations can be identified / analyzed and eventually sorted using multifunctional Raman
Probes built via a hierarchical, multi-scale assembly of intrinsically dissimilar nanoparticles. The finding of this project will help overcome problems currently hampering stem cell research. Current approaches depend on the very limited number of available antibodies or biomarkers and suitable fluorescent markers.
Stem cells have great potential for a broad range of therapeutic and biotechnological applications. Characterization and sorting of heterogeneous stem cell populations is therefore an important step towards novel stem cell-centered technologies. Once validated for stem cell phenotyping, the SAVVY technology may lead to novel diagnostics tools and devices that are applicable to a number of diseases, such as cancer (as outlined in the S&T part for oesophageal adenocarcinoma), cardiovascular, or neurological diseases. An especially promising area could be diseases linked with ageing, such as Alzheimer's and Parkinson's diseases, where there is a strong need for suitable diagnostics tools in an ever aging European society. As such, SAVVY may lead to improvements in health, personalized
medicine and ultimately to behavioral changes among EU citizens.
A main objective of SAVVY was to create novel materials with exactly tailored properties based on a detailed understanding of multi-scale, hierarchical self-assembly processes of dissimilar components. Further research will result in new generations of products (Raman probes) and services (particle diagnostics) that are in principle applicable to a broad range of applications potentiating the potential health impact.
The creation of databases of cell signatures based on SERS-enhanced signals will enable rapid identification of different cell phenotypes. The newly developed tools will be used to characterize ever growing number cell types.
The major source of innovation will be new knowledge about the bottom-up fabrication of nanoparticle assemblies and their applications in biotechnology and health diagnostics.
The newly developed tools and methods will influence future research into diagnostic tools for cell phenotyping.

The main routes taken for dissemination of the SAVVY results were via presenting results at conferences and workshops (over 120 oral or poster presentations by SAVVY scientists), the organization of three SAVVY workshops (two of which were open to all interested scientists), and the presentation at trade fairs.

Potential Impact:
http://savvyproject.eu/

List of Websites:
Attached: Executive Report including figures,logo, contact details of WP leaders and main PIs.
final1-savvy-executive-report-full-version.pdf