Enhanced surface nanoanalytics of transient biomolecular processes

Fundamental knowledge of the physical and chemical properties of biomolecules is key to understanding molecular processes in health and disease. Bulk and single-molecule analytical methods provide rich information about biomolecules but often require high concentrations and sample preparation away from physiologically relevant conditions. Here, we present the development and application of a lab-on-a-chip spray approach that combines rapid sample preparation, mixing, and deposition to integrate with a range of nanoanalytical methods in chemistry and biology, providing enhanced spectroscopic sensitivity and single-molecule spatial resolution. We demonstrate that this method enables multidimensional study of heterogeneous biomolecular systems over multiple length scales by nanoscopy and vibrational spectroscopy. We then illustrate the capabilities of this platform by capturing and analyzing the structural conformations of transient oligomeric species formed at the early stages of the self-assembly of α-synuclein, which are associated with the onset of Parkinson’s disease.

The fixing of bio-organic and biomolecular samples on surface-based methods is often performed by manual deposition. For most analytical methods, conventional manual surface-based sample preparation involves three fundamental steps: 1) pipetting a volume of solution onto a surface and allowing it to adsorb for up to a few minutes, 2) rinsing to remove weakly adsorbed biomolecules and excess salt present in buffer, and 3) removing solvent using gentle drying. Conventionally, steps 1 and 2 are applied to measure in a liquid solution, while step 3 is necessary to measure the sample in an air or vacuum environment.
Supplementary Note 2: Optimisation of microfluidic spray experimental set-up Every aspect of the experimental set-up was carefully considered, such as syringe size and material (Fig. S2). We found low volume glass syringes (250 µl) to be ideal, as this minimises sample clogging and ensures accurate volumes of liquids are flowed through the microfluidic spray device. Glass tubing proved to be incompatible with the experimental set-up due to its lack of flexibility, therefore thin polytetrafluoroethylene (PTFE) tubing was used. A standard inlet tube length of 12 cm minimised dead volumes. This allowed us to achieve a working volume for the operation of the device as low as 20 µl for each experiment, and without any minimum requirement for sample concentration.

Supplementary Note 3: Droplet evaporation & salt crystallisation
In order to understand the droplets sizes evaporation and salt crystallisation in microfluidic spray deposition, AFM images of salt crystals were analysed. The height was measured for numerous salt crystals formed, with an average height being 95 ± 65 nm. The average droplet size was measured using optical images taken using the AFM-equipped camera. Typical droplets formed were ~8-13 µm in diameter. From the droplet size, the evaporation time can be calculated (Table  S1). This is discussed more thoroughly in ref (1).  8   Table S1. Determination of the evaporation time of the droplets generated by microfluidic spray. The average droplet size was measured from optical images (n=132). Note that this value is likely an overestimate, as droplets were measured using the optical camera based on the faint coffee-ring formed of salt crystals. Smaller droplets (those a few µm in diameter) will have smaller salt crystals forming that are below the detection limit of the optical camera.

Droplet diameter (µm)
According to Burton-Cabrera-Frank theory of salt crystallisation (39), from the crystal size and evaporation time, we can calculate the growth rate of salt crystals using the equation Note that height was used as a measure in crystal size instead of length (L). This is due to artefacts in AFM images preventing reliable measurement of the lateral dimensions of crystals. From this, we can say the growth rate of salt crystals is ~1x10 -5 µm/s, which is in agreement with literature values (40)(41)(42).
Next, we applied the calculated crystal growth rate to traditional deposition methods with slow drying times. Assuming a very conservative estimate of a few minutes (180 s) drying time of a large, 10 µl droplet, the resultant crystal would be at least a few µm (1.8 µm) in size.

Supplementary Note 4: In-flight droplet evaporation
In addition to the droplet drying on the surface, discussed above and in previous work, we must also consider the in-flight droplet drying. By means of a high-speed camera, we have previously measured the approximate velocity of the droplets to be approximately 20-30 ms -1 (1,12). From this, we can estimate the time of flight to be ~1-2 ms, based on the known distance between the nozzle and the surface (in this case, 3.5 cm). Taking the median droplet size of 9.7µm (Table  S1), we can calculate the drying time on the surface to be ~8 ms. Therefore, the relative time of flight of the droplet compared to the time as a liquid on the surface is about 12-25%.

Supplementary Note 5: Comparison of size distributions measured from EM
The size distribution of samples deposited via microfluidic spray and manual deposition was compared to assess the preservation of sample heterogeneity of both methods. Comparisons were made to bulk DLS measurements acquired in solution. The size of the Aβ oligomers measured via spray was measured in the 2-20 nm range, compared to manual which displayed a 2-10 nm range.
As bulk DLS measurements are often not compatible with measuring oligomers due to the heterogeneous, large amorphous assemblies which can form, we instead opted to compare these EM measurements with single-molecule AFM studies. AFM is particularly well-suited for the accurate size characterisation of oligomers due to the high resolution of the technique and the lack of staining step. AFM analysis showed a 1-25 nm diameter range, which is in excellent agreement with spray EM results (Fig. 2f).
Next, we sought to assess the size distribution of colloids. An analysis was not possible for manual preparation, as the colloid aggregation upon manual deposition prevented their accurate measurement. It is expected that colloids will cluster in solution to form a few large particles, which will then be over-represented in DLS measurements; this was not an issue with our spray measurements. Despite the presence of large (>1000 nm) particles observed via DLS, an excellent correspondence was observed between spray deposition EM and bulk DLS measurements, with most colloids being between 5-25 nm in diameter (Q1-Q3 11-56 nm) (Fig. S4).

Supplementary Note 6: Sensitivity of FTIR measurements acquired in liquid & air
Measurements of thyroglobulin acquired in liquid (Fig. 3d) resulted in reliable spectra which reflect a conformationally-stable protein down to 12 µg. However, low sample concentration resulted in significant distortions of spectra due to difficulty in compensating for the IR absorption of water. Water predominantly absorbs IR light at ~3500 cm -1 (symmetric and asymmetric O-H stretching) and at ~1600 cm -1 (O-H bending). In particular, the IR absorption of water overlaps with the amide I band of protein, which is key for structural analysis of protein.
Thus, to measure lower concentrations, FTIR-ATR spectra are typically acquired using dried samples (Fig. 3e). However, the drying of salt in the buffer results in significant spectral distortion, which hampers chemometrics and structural analysis. Reliable spectra can be achieved using smaller amounts of protein by the addition of an extra rinsing step to remove excess salt, which allowed us to retrieve spectra with high quality amide band I and II peaks and with a good signal to noise ratio. However, the rinsing step can perturb the heterogeneity and molecular conformation of the sample. Our single-step lab-on-a-chip spray deposition enabled the characterization of extremely small amounts of protein in the presence of salt. The ability to measure spectra of small amounts of protein in salt, and associated improvement in sensitivity that we achieved, can be rationalised by considering the reduced salt crystallisation occurring with fast drying microdroplets, when compared to the formation of large salt crystals during the drying of a macroscopic liquid droplet ( Fig. S6 and Supplementary Note 5).

Supplementary Note 7: Salt crystallisation effect on FTIR-ATR Spectra
In Fig. S5, the effect of salt is described for thyroglobulin samples prepared using microfluidic spray, and those measured in air, with and without rinsing. The difference in sensitivity of ATR-FTIR by manual and spray deposition can be attributed to the formation of salt crystals.
The absorbance, and therefore the signal intensity, of FTIR-ATR measurements is related to the penetration depth (dp) of the evanescent wave and it is proportional because of the Beer-Lambert law to the thickness of the material deposited on the prism. Thus, the absorbance is proportional to the height (thickness) of the material on the surface a thicker film or crystal salt on the surface will produce a stronger IR absorption.
As calculated in the Supplementary Note 3, the slow drying (seconds to minutes) of a macroscopic droplet causes the formation of large micrometre sized salt crystals, which have significant absorption and affect the IR measurement. Based on the growth rate of salt crystals of ~1 x 10 -5 µm/s (Supplementary Note 3), we estimate that the salt crystals can grow as large as a 1.8 µm in the slow drying method. Therefore, it is expected that the IR spectrum would disproportionately reflect the presence of the salt. While the fast evaporation of the sprayed droplets minimises the time for salt crystallisation to occur, thus producing smaller nanometre sized crystals that do not significantly affect the IR measurements. In fact, the average crystal size is 95.11 ± 65.38 nm in height for samples deposited by spraying (Fig. S5). This effect is quantified by analysing the ratio between IR absorption peaks typical of the protein and the salt in solution (Fig. S5). In order to quantify the relative contribution of the salt to the IR absorbance spectra, the peak at ~1050 cm -1 (the C-O bond in the primary alcohol groups of Tris was integrated and compared to that of the amide I peak We observed a ~2-fold increase of the ratio of the IR signal of the amide I and salt peak between microfluidic spray and air which was not rinsed, despite the concentration of protein to salt being the same in solution. Thus, by significantly reducing the size of salt crystals, we thereby enable the acquisition of spectra in buffer with nanogram sensitivity without the need for rinsing (Fig. S5). Figure S1. Schematic of the single and double inlet microfluidic device designs. (a-c) 25 µm high liquid channel (a), 50 µm high gas channel (b), and their combination using a two-step lithography process (c). The same applies for the double-inlet device, which simply has an additional liquid inlet and a longer channel to allow for sufficient mixing. (e) Assembly of the two PDMS layers occurs via plasma bonding. The PDMS pieces from the two respective masters are carefully aligned with respect to one another and left to bond to give the completed device. Figure S2. A cartoon depiction of the standardised experimental set-up. The following parameters were adjusted to enhance deposition reproducibility between experimentalists and to minimise sample usage: sample inlet flow rate, gas inlet pressure, tubing material and length, the time which sample is sprayed, and the distance between the spray device and the surface. A detailed description is included in the Methods section.    (a-f) Cartoon depicting the relative sizes of protein and salt on the FTIR-ATR prism in air: without rinsing (a), with rinsing (b) and sprayed (c) (not to scale), and the corresponding spectra of thyroglobulin (1 mg/ml) measured in buffer (d, e, f, respectively). (g) In order to understand the contributions of salt and protein to the IR spectra, the ratios between the area under the salt peak (~1050 cm -1 ) and the amide I peak (~1650 cm -1 ) were calculated for each preparation method. Error bars represent the SD. Despite containing the same ration between salt and protein, the relative influence of salt on the IR spectra is more important when deposited manually than when deposited via spray. (h) When samples dry slowly (as in traditional deposition methods), large salt crystals form which result in distorted protein spectra. As such, rinsing is required in order to remove salt crystals. In the case of microfluidic spray, the fast timescales of drying limit salt crystallisation and enable the easy measurement of proteins in buffer. Salt crystals were measured using AFM; a representative image is presented. A coffee-ring effect is observed, with salt crystals forming a ring-like shape at the edge of the droplet. The average crystal size is 95 ± 65 nm.