Immunohistochemistry microarrays

was used per protein per tissue slice, equivalent to approximately $0.2 USD ... the Canadian Cancers Society Research Institute (CCSRI), the Canad...
0 downloads 0 Views 1MB Size
Subscriber access provided by UNIVERSITY OF THE SUNSHINE COAST

Technical Note

Immunohistochemistry microarrays Huiyan Li, Gabrielle Brewer, Grant Ongo, FRÉDÉRIC NORMANDEAU, Atilla Omeroglu, and David Juncker Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.7b00807 • Publication Date (Web): 01 Aug 2017 Downloaded from http://pubs.acs.org on August 2, 2017

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Immunohistochemistry microarrays Huiyan Lia,b, Gabrielle Brewera,b, Grant Ongoa,b, Frederic Normandeaua,b, Atilla Omerogluc, David Junckera,b,d

a

Biomedical Engineering Department, bMcGill University and Genome Quebec Innovation Centre, c McGill University Health Centre, Department of Pathology, dDepartment of Neurology and Neurosurgery, McGill University, Montreal, QC, H3A0G1, Canada

*

To whom correspondence should be addressed. Email: [email protected]

Abstract Immunohistochemistry (IHC) on tissue sections is widely used for quantifying the expression patterns of proteins and is part of the standard of care for cancer diagnosis and prognosis, but is limited to staining a single protein per tissue. Tissue microarray and microfluidics staining methods have emerged as powerful high throughput techniques, but either only permit the analysis of a single protein per slide, or require complex instrumentation and expertise while only staining isolated areas. Here, we introduce IHC microarrays (IHCµA) for multiplexed staining of intact tissues with preserved histological and spatial information. Droplets of a dextran solution containing antibodies were pre-spotted on a slide and snapped onto a pre-processed formalin-fixed, paraffin-embedded (FFPE) tissue section soaked in a polyethylene glycol solution. The antibodies are confined within the dextran droplets, and locally stain the tissue below with a contrast similar to the one obtained by conventional IHC. The microarray of antibody droplets can be pre-spotted on a slide and stored, thus neither the preparation of the antibody solutions nor a sophisticated microarray spotter is needed. Sampling considerations with IHCµA were evaluated by taking three tissues with varying levels of cancer cells. A multiplex IHCµA with 180 spots targeting 8 cancer proteins was performed on a breast cancer tissue section to illustrate the potential of this method. This work opens the avenue of applying microarray technologies for conducting IHC on intact tissue slices, and has great potential to be used in the discovery and validation of tissue biomarkers in human tumors.

1 ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 11

Introduction Analysis of protein biomarkers in tumor tissues can help in cancer diagnosis, classification, prognosis, and treatment.1-3 Compared to single protein analysis, a panel of multiple proteins can improve diagnostic accuracy.4-5 To date, immunohistochemistry (IHC) is the most routinely used technology for the detection of proteins in tissue samples.6 However, conventional IHC only detects one protein at a time, limiting its application for high-throughput discovery and validation of protein biomarkers in tumor tissues. Array-based technologies have been developed for the detection of proteins in tissues by arraying patient samples on slides. Tissue microarrays (TMA) comprise an array of hundreds of paraffin embedded tissue disks,7-8 with typical tissue core size ranging from 0.6 mm to 2 mm.9 Likewise, reverse-phase protein arrays (RPPA) 10-11 are made of proteins extracted from either FFPE or fresh frozen tissues. However, histological spatial information is lost not only for RPPA as only a homogenate is measured, but also for TMA as the original sample is destroyed. Hence efforts have been made to directly quantify multiple proteins in a single tissue slice which has the benefit of providing multiplexed and contextual information. An early demonstration is the multipleximmunostain chip formed in a silicone rubber comprising 50 wells, each 2 mm wide and 3 mm apart,12 allowing 50 proteins to be analyzed simultaneously, but requiring tissue sections that are at least 30 × 15 mm2 large, much larger than many human tumors that could be only 5-10 mm in diameter.13 Recently, multiplexed IHC was realized using an aqueous two-phase system with microliter droplets of a dextran solution containing different antibodies applied to a slide soaked within a polyethylene glycol (PEG) solution.14 However the center-to-center spacing was even larger at ~ 5 mm. Miniaturized, multiplexed IHC was first proposed by Kim et al. using microfluidic chips with 800 µm wide and 5 mm long channels for staining 4 proteins simultaneously.15 105 human breast tumor tissues were studied, and the results correlated well with those from conventional IHC. Others stained 10 proteins in parallel,16 automated assays17 and used quantum dots to improve assay performance.18 Lovchik et al. used a microfluidic probe to locally stain tissues with ~ 100 µm diameter tear drop-shaped patterns.19 Here, contact was avoided, and sample consumption was much reduced, but serial processing is required for staining with multiple antibodies. Collectively, these microfluidic technologies demonstrate that multiplex, microscale IHC is possible on a single tissue slice. However, the microfluidic systems required microfabricated chips, specialized equipment such as syringe pumps and pressure controllers, which is not available to typical biology and pathology laboratories. Here, we introduce immunohistochemistry microarrays (IHCµA), a microarray-based multiplexed IHC that allows for (i) probing multiple proteins on a single tissue section without cross-reactivity; (ii) distributing multiple micro-spots (200 µm) at high density across the whole tissue for each protein, instead of macroscopic staining of one or a few sub-sections (e.g. TMA sampling method) of a tissue slice for each protein, to ensure different locations of the tissue are sampled thus better tackling the tissue heterogeneity issues and minimizing false-negative scenarios; (iii) saving reagents and reducing the cost of the assay; (iv) storing the pre-spotted antibody microarrays so that no sophisticated microarray spotter is required by “end-users”. As a proof-of-concept a FFPE breast tumor tissue slice was stained with an 8-plex IHCµA with 180 spots illustrating the potential of this method.

Experimental 2 ACS Paragon Plus Environment

Page 3 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

The assay procedure of the IHCµA is shown in Figure 1. Here, we apply the snap chip technology20-21 to transfer a microarray of primary antibodies simultaneously onto a tissue slide, resulting in ~ 200 µm antibody spots on the tissue slice. All the antibodies are prepared, spotted, and stored in advance to disassociate the reagent preparation and chip fabrication with assay procedure, so that neither antibody preparation nor a microarray spotter is required on site. The IHCµA is compatible with the most widely used FFPE tissue slides, and all the commonly used protocols for tissue pre-treatment (deparaffinization, rehydration, antigen retrieval), blocking, and downstream processing after incubation of primary antibodies can be easily adapted in this method. Both fluorescence and chromogenic detection can be implemented in the downstream processing steps.

Figure 1. Schematic illustrating the procedure of IHCµA. (a-b) The FFPE tissue slide is pre-treated (deparaffinization, rehydration, and antigen retrieval), and then blocked with BSA. (c) Primary antibody droplets are spotted on a slide with an inkjet spotter, and the slide is stored. (d) To perform multiplexed IHCµA, the Ab slide is retrieved, and mounted in a snap apparatus and snapped onto the tissue slide. (e-f) After incubation, the tissue slide is stained with either fluorescence or chromogenic methods, followed by imaging and quantifying the staining.

Results and Discussion To demonstrate the transfer of antibody droplets homogeneously onto a tissue slice without crosscontamination between adjacent spots, antibody microarrays were transferred onto 5 µm-thick FFPE tissue slides from a human breast tumor for staining TIMP122 and estrogen receptor (ER) proteins. Although chromogenic detection is widely used in IHC, the results are semi-quantitative, and therefore immunofluorescence was used to quantify the signal intensities.23-24 We first spotted antibodies in PBS with 20% glycerol and transferred them onto the tissue in a humidified environment, Figure S1a. The fluorescence image of the stained spots showed an obvious ring effect (Figure S1a-b), might be due to the Marangoni effect23-25 or diffusion limitations when a large amount of antibodies reached the tissue from the sides of each spot.26 To overcome the ring effect, we implemented the PEG-dextran two-phase system14, 27 in the IHCµA, with antibodies mixed in dextran solutions and the tissue slice covered with a PEG aqueous solution (see Figure S1c). To adapt the dextran droplets to nanoliter-scale inkjet spotting, we decreased the concentration of the dextran to 5%, and also added 10% glycerol in the spotting solutions to avoid 3 ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 11

evaporation. The fluorescence image in Figure S1c, the quantitative analysis in Figure S1d, as well as the images in Figure 2 and 4 show that antibodies were homogeneously transferred onto the tissue or cell monolayer slices, without ring effect observed in most images, and individual spots were well separated. In Figure S1d, some spots showed higher intensity on the edge, which reflected a lack of cells inside the spots rather than the ring effect. We quantified the signal intensity of 49 replicated spots on a slide with a cell monolayer as shown in Figure S1c, and normalized the data to the number of cells in each spot. The coefficient of variation (CV) of the 49 replicated spots stained using IHCµA was 17%. A close-up image of a microspot stained with the anti-TIMP1 antibody is shown in Figure S2a, using DAPI as the counter-stain for the visualization of cells in the image. Further, we optimized antibody concentration, and found that for example, for TIMP1, 400 µg/mL of the antibodies gave the best s/n ratio of 9.2 (Figure S2b). Also, the results of the storage of the antibody slides are shown in Figure S3. The histological architecture of a monolayer MCF-7 breast cancer cells, known to be positive for ER was revealed by chromogenic staining using IHCµA or conventional IHC/ICC and found to be comparable, Figure 2. IHCµA required higher concentrations of antibodies compared to conventional staining, probably because the aqueous two-phase system with glycerol decreased antibody-antigen binding efficiency, similar to the snap-chip-immunoassays where much higher antibody concentrations (but less antibody overall) were needed compared to large-volume conventional assays.21 In Figure 2a, nuclei outside of the spots were sometimes stained as well, probably due to the diffusion of the antibodies out of the droplet. The center-to-center spacing of 500 µm used here appears adequate to prevent cross contamination between spots. The slightly variation in spot size was also attributed to the variations in inkjet spotting due to the relatively viscous antibody solutions. These results indicate that the IHCµA can provide comparable staining patterns than conventional IHC/ICC, and can be used to reveal spatial protein expression patterns in tissues.

4 ACS Paragon Plus Environment

Page 5 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Figure 2. Images of stained cell monolayer with anti-ER antibody using (a) IHCµA method and (b) conventional ICC (antibodies were used at 10 µg/mL). Dashed circles in (a) delineate the boundary of the stained spots. (c, d) Close-up of the spot within the dashed box in (a) and (b), respectively. Microscale staining of a single area may not be representative of the whole tissue considering the heterogeneity of tissue and the irregular shape of tumors. IHCµA can address this concern by using multiple replicate spots distributed across the whole tissue to stain for the same protein. In this work, we used the percentage of stained cells to help evaluate the variations among microspots and the differences between microspot-based sampling and the whole-slice staining. As a proof-of-concept, three breast tumor tissue sections with varying degrees of ER and PR staining representing three positive, but quantitatively distinct percentage of stained cell scenarios in the widely used H-score system were evaluated,28 Figure 3a-f. Conventional whole-slice staining was used to generate the images, and a digital grid was used to define the 81 spots in silico resembling an IHCµA. The distribution of the percentage of stained cells (Figure 3g) show that the number of positive cells within individual spots varies significantly, but when increasing the number of spots (up to 81) that are sampled, the average number of positive cell converges to the actual average of the whole slice. Also, we observed that (Figure 3h-j), for example, in Figure 3i, the percentage of stained cells from spots on the left side had higher counts than that of the spots in the middle. These observations imply that the microspots should be sampled across the whole area of interest on the tissue, instead of taking all from one sub-area, to address the tissue heterogeneity issues. Therefore, to distribute the spots across the tissue, we divided each tissue into 9 sub-sections, with 9 spots in each sub-section, as shown in Figure 3a-c. For clinical diagnosis of cancer, our first criterion would be to identify whether there are cancer cells or not. Subsampling might increase the risk of false-negative diagnosis. Considering the heterogeneity of tissues, taking one or multiple spots from each sub-section would help minimize false-negative results. Next, we investigated the effect of additional sub-sampling on the percentage of stained cells when using between 1 and 81 spots. For scenarios with (i) up to 9 spots, each spot was selected from a different sub-section and (ii) for more than 9 spots, the same amount of spots was first attributed to each of the 9 sub-sections, and the remainders were distributed according to (i). In each sub-section the spot position was randomly attributed. The percentage of stained cells for each of the 81 scenarios was calculated from all spots. Each of the scenarios from 1 to 81 spots was run 100 times. The distribution of each coverage is shown in Figure 3k. For low number of spots, there was a broad range in the number of stained cells which converges to the 81-spot average as more and more spots are considered. The graphs suggest that between 10 and 25 spots may be needed to obtain staining results that converge towards the 81-spot average. With 10 spots, the CV of the percentage of stained cells, for randomly sampling 100 times, are 12%, 15%, and 21% for the tissues with score 3, 2, 1, respectively. With 25 spots, the CVs are 6.6%, 8.3%, and 10%, respectively. These results indicate that the accuracy of the IHCµA is dependent on many parameters, including the percentage of stained cells of the whole tissue and the tissue heterogeneity. One should keep in mind that conventional IHC also suffers from sampling artifacts as a tissue section constitutes only a small fraction of the tumor tissue. In IHCµA, the more spots are used, the closer the results will converge to conventional whole tissue staining, but at the cost of multiplexing capability. Clinically, the H-score is widely used, and it will be important to ensure that the H-score is conserved between methods, which could be achieved by deriving a confidence metric so that counts falling close to the threshold between two categories are identified as ambiguous. In the case shown here, we found that by using at least 8, 4, 17 spots for each of the tissues, respectively, the same H-score was obtained than by conventional IHC. The sub-sampling method could be enhanced using artificial intelligence and machine learning,29 allowing to extract additional information. If a result is inferred to 5 ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 11

be ambiguous with possible clinical consequences, then staining on additional slides could be performed. In any case, rigorous clinical studies and statistical analysis will be required before therapeutic choices can be made based on the results of IHCµA.

6 ACS Paragon Plus Environment

Page 7 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Figure 3. The effect of microarray-based sampling on the percentage of stained cells on three different tissue slices with varying levels of cancerous cells. A 4.2 × 4.2 mm2 square tissue section that could 7 ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 11

accommodate 81 microspots was defined. Each spot was 200 µm in diameter, and the center-to-center spacing between spots was 500 µm, same as the IHCµA layout used in this work. (a-c) Layout of 81 microspots on three 4.2 × 4.2 mm2 tissue sections, divided into 9 sub-sections. The surface coverage of the 81 spots represents 45% of the square. The percentage of stained cells for the whole tissue is 77% (score 3) for (a) targeting ER protein, 24% (score 2) for (b) and 6.6% (score 1) for (c) targeting PR protein. The close-up image of the spots on the row 2, column 9 of each tissue is shown in (d-f). The nuclei are revealed with diaminobenzidine staining and are brown, and the counter-stain conducted with hematoxylin is blue. Scale bars: 30 µm. (g) Box plots showing the percentage of stained cells for each of the 81 microspots on three tissues. (h-j) 3D column charts demonstrating the percentage of stained cells for each of the 81 spots on the 3 tissues. (k) Box plots showing percentage of stained cells for 100 times sampling of 1-81 spots on each of the three tissues. To demonstrate the multiplexing capability of the IHCµA, we performed an 8-plex assay on a commercial FFPE breast tumor slide, with subtype of ER-, PR+, HER2-. The 8 proteins include three markers for breast cancer subtyping (ER, PR, HER2) and 5 cancer-related proteins. The layout of the microarray and the fluorescence image of IHCµA staining are shown in Figure 4a-b. As evaluated in Figure 3, we used 18 spots for each protein, by randomly taking 2 spots from each of the 9 sub-sections as shown in Figure 3a-c. The distributions of the s/n ratios from 18 spots for each protein are shown in Figure 4c. PR was over-expressed, and the average signal levels from HER2 and ER were similar to those from negative controls. These results are consistent with the tumor subtype as determined by conventional IHC. We found P53 and PTEN were highly over-expressed in this tissue. This high expression of P53 and PTEN in PR+ breast tumor is consistent with previous findings in the same subtype.30-31 TIMP1 was also detected in the tissue while CDK7 and CK7 were not highly expressed. The expression of some proteins, such as PR and TIMP1, was relatively heterogeneous in this tissue, showing a wide dynamic range in s/n ratios, while other proteins such as ER and HER2 were more homogeneous. To validate the multiplex IHCµA, these 8 proteins were also stained using whole-tissue fluorescence assays on adjacent tissue slices from the same tumor, and the s/n ratios against the staining with normal IgG were calculated for each protein (Figure S4), and found to be consistent across both methods. Our results suggested that IHCµA may be a useful tool for quantitative, multiplexed protein detection on FFPE tissue slides.

8 ACS Paragon Plus Environment

Page 9 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Figure 4. IHCµA detecting 8 proteins on a human breast tumor tissue slide with the tumor subtype of ER-, PR+, HER2-. (a) Layout of the antibodies targeting 8 cancer-related proteins. Normal mouse IgG (M IgG) and normal rabbit IgG (R IgG) were used as negative controls. (b) Fluorescence image of the staining results on the tissue slide. (c) Box plots of s/n ratios (the ratio of the net signal intensity to the noise level of the antibody) for each of the 8 proteins and 2 IgG controls.

Conclusions In this work, we developed a novel multiplexed immunohistochemistry using a snap chip. The microarray-based sampling considerations to tackle tissue heterogeneity issues and to minimize falsenegative scenarios were discussed. An 8-plex assay with two negative controls using a total of 180 micro-spots was performed on a breast tumor tissue slide, which is the largest number of isolated stains on a single slice to date to the best of our knowledge, indicating the multiplexing potential of this technology. Compared to previous array-based tissue staining,12, 14 our approach increased the array density by two orders of magnitude, and thus a small tumor tissue of 5 × 5 mm2 could be stained with 100 micro-spots, and used for most tumors that reach this size at diagnosis. In this multiplex assay, 108 nL of antibody solution was used per protein per tissue slice, equivalent to approximately $0.2 USD per target, thus saving reagent and reducing the cost of the assay. Similar to TMA, IHCµA is based on the concept of sampling, which needs to be carefully optimized to ensure representative staining results, but which allows obtaining more information from a given sample. The results obtained with IHCµA were consistent with whole-slice staining, and in case of ambiguous results, could be resolved by staining additional sections from the same tissue. A statistical analysis with a sufficient large number of tissue sections will be necessary to develop theoretical and experimental models that will help choose the optimal number of spots to obtain staining results within a predetermined confidence interval as compared to conventional IHC on a full slice. Further miniaturization might open the door to array-on-array staining by applying a high density antibody microarray on a lower density tissue microarray using a snap chip. For example, with 250 µm spot-tospot spacing, 12 or 27 spots could be accommodated on typical core sizes of TMA with 1 mm or 1.5 mm diameter. We believe that the IHCµA has the potential to be used in large-scale multiplexed protein profiling in archived FFPE tumor tissue samples.

Acknowledgments We thank Rob Sladek for the use of the Nanoplotter inkjet spotter. We thank the financial support from the Canadian Cancers Society Research Institute (CCSRI), the Canadian Institutes for Health Research (CIHR), the Natural Science and Engineering Research Council of Canada (NSERC), and the Canada Foundation for Innovation (CFI). G.B. acknowledges a scholarship from the NSERC-CREATE Integrated Sensor Systems program. D.J. acknowledges support from a Canada Research Chair.

References (1) (2) (3) (4) (5) (6) (7)

Parise, C. A.; Caggiano, V. J Cancer Epidemiol. 2014, 2014, 11. Muller, Patricia A. J.; Vousden, Karen H. Cancer Cell. 2014, 25, 304-317. Thomas, D. S.; Fourkala, E. O.; Apostolidou, S. et al. Br. J. Cancer. 2015, 113, 268-274. Xu, T.; Fang, Y.; Rong, A.; Wang, J. BMC Med Res Methodol. 2015, 15, 94. Kang, L.; Xiong, C.; Crane, P.; Tian, L. Stat. Med. 2013, 32, 631-643. Hornick, J. L. Mod. Pathol. 2014, 27, S47-S63. Kallioniemi, O.-P.; Wagner, U.; Kononen, J.; Sauter, G. Hum. Mol. Genet. 2001, 10, 657-662. 9 ACS Paragon Plus Environment

Analytical Chemistry

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 11

(8) Rimm, D. L.; Camp, R. L.; Charette, L. A.et al. Cancer journal (Sudbury, Mass.). 2001, 7, 24-31. (9) Eckel-Passow, J. E.; Lohse, C. M.; Sheinin, Y.; Crispen, P. L.; Krco, C. J.; Kwon, E. D. Diagn Pathol. 2010, 5, 1-10. (10) Wilson, B.; Liotta, L. A.; Petricoin III, E. Dis. Markers. 2010, 28. (11) Assadi, M.; Lamerz, J.; Jarutat, T. et al. Mol. Cell. Proteomics. 2013, 12, 2615-2622. (12) Furuya, T.; Ikemoto, K.; Kawauchi, S.; Oga, A.; Tsunoda, S. i.; Hirano, T.; Sasaki, K. J. Histochem. Cytochem. 2004, 52, 205-210. (13) Pitteri, S. J.; Kelly-Spratt, K. S.; Gurley, K. E. et al. J. Cancer Res. 2011, 71, 5090-5100. (14) Frampton, J. P.; Tsuei, M.; White, J. B.; Abraham, A. T.; Takayama, S. Biotechnol J. 2015, 10, 121-125. (15) Minseok S. Kim, T. K., Sun-Young Kong, Soim Kwon, et al. PLoS ONE. 2010, 5. (16) Kim, M. S.; Kwon, S.; Kim, T.; Lee, E. S.; Park, J.-K. Biomaterials. 2011, 32, 1396-1403. (17) Ciftlik, A. T.; Lehr, H.-A.; Gijs, M. A. M. Proc. Natl. Acad. Sci. USA. 2013, 110, 5363-5368. (18) Kwon, S.; Cho, C. H.; Lee, E. S.; Park, J.-K. Anal. Chem. 2015, 87, 4177-4183. (19) Lovchik, R. D.; Kaigala, G. V.; Georgiadis, M.; Delamarche, E. Lab Chip. 2012, 12, 1040-1043. (20) Li, H.; Bergeron, S.; Juncker, D. Anal. Chem. 2012, 84, 4776-4783. (21) Li, H.; Munzar, J. D.; Ng, A.; Juncker, D. Sci. Rep. 2015, 5, 11688. (22) Würtz, S. Ø.; Würtz, S. Ø.; Schrohl, A.-S.; Mouridsen, H.; Brünner, N. Acta Oncol. 2008, 47, 580-590. (23) Dupouy, D. G.; Ciftlik, A. T.; Fiche, M. et al. Sci. Rep. 2016, 6, 20277. (24) McCabe, A.; Dolled-Filhart, M.; Camp, R. L.; Rimm, D. L. J. Natl. Cancer Inst. 2005, 97, 1808-1815. (25) Rao, A. N.; Vandencasteele, N.; Gamble, L. J.; Grainger, D. W. Anal. Chem. 2012, 84, 10628-10636. (26) Kris Pappaert, H. O., Hugo Thienpont, Paul Van Hummelen, and Gert Desmet. Biotechniques. 2006, 41, 609616. (27) Frampton, J. P.; White, J. B.; Simon, A. B.; Tsuei, M.; Paczesny, S.; Takayama, S. Sci. Rep. 2014, 4, 4878. (28) Rizzardi, A. E.; Johnson, A. T.; Vogel, R. I. et al. Diagn Pathol. 2012, 7, 1-10. (29) Madabhushi, A.; Lee, G. Med. Image Anal. 2016, 33, 170-175. (30) Torhorst, J.; Bucher, C.; Kononen, J. et al. Am J Pathol. 159, 2249-2256. (31) Lebok, P.; Kopperschmidt, V.; Kluth, M.et al. BMC Cancer. 2015, 15, 963.

10 ACS Paragon Plus Environment

Page 11 of 11

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

TOC graphic 44x31mm (300 x 300 DPI)

ACS Paragon Plus Environment