Ultra-streamlined single cell proteomics by all-in-one chip and data-independent-acquisition 1 mass spectrometry 2

Single cell proteomics provides the ultimate resolution to reveal cellular phenotypic heterogeneity 20 and functional network underlying biological processes. Here, we present an ultra-streamlined 21 workflow combining an integrated proteomic chip (iProChip) and data-independent-acquisition 22 (DIA) mass spectrometry for sensitive microproteomics analysis down to single cell level. The 23 iProChip offers multiplexed and automated all-in-one station from cell isolation/counting/imaging 24 to complete proteomic processing within a single device. By mapping to project-specific spectra 25 libraries, the iProChip-DIA enables profiling of 1160 protein groups from triplicate analysis of a 26 single mammalian cell. Furthermore, the applicability of iProChip-DIA was demonstrated using 27 both adherent and non-adherent malignant cells, which reveals 5 orders of proteome coverage, 28 highly consistent ~100-fold protein quantification (1-100 cells) and high reproducibility with low 29 missing values (<16%). With the demonstrated all-in-one cell characterization, ultrahigh 30 sensitivity, robustness, and versatility to add other functionalities, the iProChip-DIA is anticipated 31 to offer general utility to realize advanced proteomics applications at single cell level.


33
Rapidly developing single cell omics-based molecular measurements have revolutionized modern 34 biological research 1, 2 . As proteins are functional workhorses of the cell, proteomic profiling 35 provides a direct snapshot of the dynamic biological network and its alteration to complement the 36 genomics and transcriptomics architecture 3 . However, the sensitivity of proteomic profiling is 37 greatly limited due to the wide dynamic range of proteome constituents and the lack of a viable 38 protein amplification strategy 4 . Targeted protein analyses have enabled sensitivity down to the 39 single cell level, but their multiplexity is often limited and depend heavily on the antibody 40 availability and quality [5][6][7][8] . Mass spectrometry (MS)-based proteomics approach, meanwhile, offers 41 label-free analysis with high specificity and deep proteomic coverage, which theoretically extends 42 to the single cell sensitivity [8][9][10][11] . However, multi-step processing in traditional MS workflows often 43 results in significant sample loss, linking trade-offs between the high proteome coverage and the 44 accessible sample size 8, 12 .

45
Microproteomics workflows aiming at handling minute samples were widely developed to expand 46 the horizon of MS-based proteomic analysis towards limited input samples (<1000 cells) 13 .
Typical strategies developed reagents and methods that could integrate the entire or partial demonstrated using both the human adenocarcinoma cell (PC-9) and human chronic B cell 79 leukemia cell (MEC-1), whose size differences were readily quantified using the built-in cell theoretically recover all peptides in the m/z and retention time domains of DIA data (Fig. 1g).

111
Specifically, spectra libraries constructed from the cell lines with different cell numbers have been 112 tested and optimized to maximize the number of protein identification and quantification.

113
In the first step of streamlined proteomics workflow, the cell trapping efficiency was optimized 114 and determined using non-small cell lung cancer (NSCLC) PC-9 cells (method). Using optimal 115 cell density (5.0×10 5 cell/ml), desired numbers of cells (1-100) for each unit can be trapped in 10-116 60 s. The average percentage of cells captured from traps containing a single cell were 100 %, 92 117 ± 3 % and 89 ± 8 % for chambers with 10, 50 and 100 traps respectively. The targeted capture 118 efficiency for all units achieves ~100% after counting traps containing 1 (90%) and 2 or 3 cells 119 (~10%), establishing it as an absolute quantifiable module to perform simple and fast size-based 120 cell isolation (Fig. 2a, b, method and Supplementary Video 2). Compared to external stand-121 alone cell sorting instruments, such a built-in module offers a simple, rapid and efficient cell 122 isolation. Additionally, we also showed that by using lower input cell density of 2.5 ×10 4 cell/ml, 123 such cell chambers allow precise capture of lower numbers of cells at the level of 1 and 5 cells 124 ( Supplementary Fig. 3). Furthermore, although unexplored in this study, such design is amenable 125 to adapt different numbers of traps, as well as alternative cell sorting strategies, such as via the 126 trap functionalization of cell surface markers.

127
Next, we sought to characterize if mixing of input reagents can efficiently occur in the closed 128 vessel during the cell lysis and protein digestion. Three mixing approaches, including vortexing, 129 shaking (by a plate shaker), and passive diffusion, were tested ( Supplementary Fig. 4). Using 130 imaging analysis, relative mixing index (RMI) was calculated to assess the mixing performance 131 (method) 28 . The result showed that it took 11, 16, and 30 min for vortexing, shaking, and diffusion-132 mixing to reach 75% RMI, indicating that all three mixing strategies were sufficient to 133 accommodate reactions within minutes to hours reaction kinetics, which fit the timescale of 134 conducting proteomics workflow (Fig. 2c, d and Supplementary Fig. 4). Although vortex mixing 135 was found to provide faster mixing, mixing by shaking was used in subsequent experiments due 136 to its flexibility in handling and sufficient reaction timescale.
Two sets of 5 μm spaced filters were embedded at both ends of the column to enable desalting 139 materials such as C18 beads packing (Fig 2e and method). The loading capacity and peptide 140 recovery efficiency of the desalting module were assessed through the BCA assay by using tryptic 141 peptides of BSA protein (method). The quantitation result showed a linear correlation from 0.125 142 to 1 μg with ~89 % recovery ( Fig. 2f and method). Assuming that a typical mammalian cell 143 contains 200 pg proteins 14 , the desalting column is thus anticipated to capture peptides from 144 approximately 4000 cells. Furthermore, the concern of a compromised sample retrieval from 145 circular reaction vessel to desalting column due to the preferential flow resulting from (1) circular 146 shape of the chamber and (2) non-negligible flow resistance caused by compactly-packed C18 147 beads was studied by flowing a colored dye through the C18 beads-packed column (method). The 148 result showed that 9 psi was the minimal flow pressure to overcome the preferential flow, and that 149 11 psi was used in our following workflow (Supplementary Fig. 5 and method).

178
To evaluate the performance of DIA-based quantitation, analytical merits in sensitivity, proteome 179 coverage, and reproducibility were systematically investigated by using iProChip to process 13-180 14 PC-9 cells and compared to the conventional DDA method (Fig. 3a). By the conventional DDA

228
Encouraged by the high sensitivity, we further pushed the profiling sensitivity at a single PC-9 cell 229 using 10-cells capture chambers. An average of 976±37 protein groups (3,069 peptides) were 230 identified from a single PC-9 cell (Fig. 4b). Triplicate measurements yielded identification of a 231 total of 1160 protein groups (3,995 peptides) from a single PC-9 cell. To the best of our knowledge, 232 the results present one of the highest reported coverage so far from independent replicates.

233
Comparing the identification coverage showed 69% protein groups and 55% peptides were 234 common between triplicate results. Intriguingly, with the built-in capability to directly image each 235 captured cell, we observed that numbers of identified proteins and peptides are correlated with the 236 approximate size of the individual captured cell (Fig. 4b).

237
To evaluate the analytical reproducibility of our approach for analysis of different numbers of cells,

286
HLA-DRB1 and HLA-DRB5 from as little as a single cell (Fig. 5b). While other key proteins 287 including CD19, CD22, CD47 and CD74 were identified from 14 and 117 cells in addition to the 288 aforementioned list of proteins (Fig. 5b). Functional annotation using UniProt showed that many 289 proteins related to adaptive immunity, innate immunity, kinases, phosphatases and Ig domain were  (Fig. 5d). It was also noted that although 298 MEC-1 cells were smaller than PC-9 cells, protein identification achieved good coverage and 299 overlap (61%-81%) using the iProChip-DIA approach, suggesting the versatility and robustness 300 of our platform for different cell types (Supplementary Fig. 7). antigen-dependent activation of BCR contributes to cell proliferation 34 . Using the iProChip-DIA 306 approach, we were able to map 83% proteins within the BCR pathway and identify the key BCR 307 co-receptors including CD19, CD21, CD22 and CD81 (Supplementary Fig. 14). In addition,  is readily applicable to profile extremely limited numbers of leukemia cells even at single cell level.

318
We reported herein, to the best of our knowledge, the first all-in-one and fully automated device         Supplementary Fig. 1). Triplicate operational units are designated for 10, 50 and 100, cells.

395
The control layer contains 34 hydraulic microvalves to control the flow layer. To account for 396 PDMS shrinkage, the flow layer layout was expanded by 1.5 % relative to the control layer. The  To fabricate master molds for the iProChip, regular photolithographic protocols were followed and 401 performed on silicon wafers using a EVG-620 mask aligner 23 . Briefly, a 4-inch silicon wafer was cleaned thoroughly using acetone, isopropanol, and DI water, followed by dehydration (

428
It was then activated using the oxygen plasma at highest RF power for 1 min (Harrick plasma 429 cleaner PDC-001-HP), before being aligned and bound to the thin control layer using a custom 430 stereo-microscope with independent x-, y-and z-alignment controller (Nikon-SMZ18). After 431 baking in the 80 o C oven overnight, the bounded chip was trimmed, peeled off and holes-punched, before binding to a freshly plasma-treated 75x50x1 mm glass slide. The bounded chip was then 433 placed in the 80 o C oven for at least 48 h before following experimental use.

434
Preparation and characterization of the SPE column using BCA assay

435
The desalting columns were prepared by slurry packing with 5 μm C18 beads in acetone (5 g/ml)

436
with an input pressure of 13 psi, which typically takes 10 to 12 min (Supplementary Video 1), 437 followed by washing with methanol for 4-6 min and activated by buffer A (100% ACN + 0.1%  where, N is the total number of pixels, Ioi is the local pixel intensity in the unmixed state, Ii is the 463 local pixel intensity in mixed state, and <I> is the average intensity.

464
The fluid dynamics, especially the possibility of preferential flow as a result of (1) the circular   Fig. 6). In the third step, Lys C (42 nL, protein/Lysine-C 20:1 w/w) 491 and trypsin (42 nL, protein/trypsin 10:1 w/w) were sequentially infused into individual digestion