2.1.

Primary Antibody Conjugation

Monoclonal antibodies were purchased from AbCam (Cambridge, UK), Thermo Fisher Scientific (Waltham, Massachusetts), Biolegend (San Diego, California), or Cell Signaling Technology (Danvers, Massachusetts) to the following targets: cytokeratin 5 (CK5), cytokeratin 8 (CK8), cytokeratin 19 (CK19), proliferating cell nuclear antigen (PCNA), Ki-67, E-cadherin (E-Cad), human epidermal growth factor receptor 2 (HER2), $\alpha$-smooth muscle antigen ($\alpha$-SMA), CD44, cytochrome c oxidase (CoxIV), CD4, and CD3 (Table 1). A unique dibenzocyclooctyne (DBCO) or amine-terminated single-stranded oligonucleotide (DS, 28 mer in length) was obtained from Integrated DNA Technologies (Coralville, Iowa) to conjugate to each primary antibody. Prior to modification, the immunoglobin G (IgG) antibodies were purified from storage buffers, including an azide reagent using 50 kDa Amicon filters (EMD Millipore, Burlington, Massachusetts). Antibody modification and oligonucleotide (oligo) conjugation were performed using either the SiteClick™ antibody azido modification kit (Thermo Fisher Scientific) or the Solulink™ antibody modification kit (TriLink Biotechnologies, San Diego, California) following the manufacturer’s instructions, explained briefly as follows. The SiteClick kit cleaved the carbohydrate domains on the heavy chains of the monoclonal IgG antibodies using $\beta$-galactosidase, enabling the addition of an azido moiety. The DBCO-terminated DS was mixed with the azido-modified whole IgG and standard copper-free click chemistry reagents, resulting in site-specific labeling of the antibodies with their unique oligo sequence (Ab-oligo). Using the Solulink kit, the antibody was modified at the available lysine residues, where a S-HyNic (part# S-9011-1-01, TriLink Biotechnologies) small molecule was conjugated. Simultaneously, the amine-terminated DS was covalently modified to contain a 4-FB (part# S-9011-1-02, TriLink Biotechnologies) small molecule on the $5^{\prime }$ end. The 4-FB modified DS and S-HyNic conjugated antibody were mixed together at a molecular target ratio of 20 oligos to 1 antibody and the $N$-hydroxysuccinimide (NHS) ester reaction occurred in the presence of the Turbolink™ catalyst (part# S-9011-1-05, TriLink Biotechnologies). Excess oligo was removed from the Ab-oligo conjugates using molecular weight cut-off spin columns. After purification, the absorbance of the Ab-oligo conjugate was measured on a Spectramax M5 (Molecular Devices, San Jose, California). The maximum absorbance of the antibody and conjugated DS was measured at 280 and 260 nm, respectively. Using the estimated extinction coefficient of $\mathrm{210,000}{\mathrm{M}}^{-1}{\mathrm{cm}}^{-1}$ for all antibodies and the manufacturer reported extinction coefficient for each DS, the approximate Ab-oligo conjugation ratios (CR) were calculated to quantify the average number of oligos bound to an antibody.

Table 1

Oligonucleotide conjugated antibodies (Ab-oligos).

An aliquot of the same primary antibody used for oligo conjugation was directly conjugated to AF555 NHS ester (AF555 NHS, Thermo Fisher Scientific). The starting antibody concentration was determined by measuring the absorbance at 280 nm prior to any conjugation chemistry. The primary antibody was again purified of any storage buffers (e.g., sodium azide) and buffer exchanged into $1\times$ PBS, pH 8.3 with a 10 kDa Amicon filter. AF555 NHS was resuspended at 10 mM in anhydrous dimethyl sulfoxide and added to the antibody at a ratio of 10 fluorophore molecules to 1 antibody in a reaction volume of 1 mL. The mixture was rocked gently at 4°C for 3 h protected from light. The antibody conjugate was then purified with a 10 kDa Amicon filter. Absorbance was measured at 280 and 555 nm to calculate the antibody and fluorophore concentrations, respectively, enabling quantification of the CR (FL CR) or the average number of fluorophores bound to an antibody.

2.2.

Formalin Fixed Paraffin Embedded Tissue and Cell Samples for Antibody Validation

Deidentified human FFPE tissue blocks were obtained from the Oregon Health and Science University Knight Biolibrary. The functionality of oligonucleotide and fluorophore conjugated antibodies was validated on FFPE blocks of normal breast tissue, normal tonsil tissue, and FFPE cell buttons generated from MDA-MB-468 or Sk-Br-3 breast cancer cell lines. All cell lines were originally purchased from ATCC (Old Town Manassas, Virginia) and maintained at less than 25 passages for all experiments. The tissue or cell button type known to express the target of interest was used for validation of each primary antibody (Table 1). Five micrometer sections of the selected FFPE block were captured onto $1\times 3\mathrm{in.}$ Superfrost™ plus slides (Thermo Fisher Scientific), which were baked at 65°C in a hybridization oven for times ranging from 30 min to overnight prior to commencing staining.

2.3.

Antibody Staining on FFPE Samples

The baked FFPE slides were deparaffinized at room temperature (RT) in xylenes ($2\times 10\min$ washes). The tissue was then gradually rehydrated in ethanol and water solutions as follows: 100% ethanol ($2\times 10\min$), 95% ethanol (5 min), 70% ethanol (5 min), and then, 50% ethanol (5 min). The tissue was then rinsed in 100% deionized water (${\mathrm{diH}}_2\mathrm{O}$) followed by a wash in $1\times$ PBS, pH 7.4 for 10 min. A two-step antigen retrieval procedure was performed in a Pascal Pressure Cooker (Dako, Santa Clara, California). RT solutions of 10 mM sodium citrate, pH 6, $1\times$ tris hydrochloride (HCl), pH 8, and ${\mathrm{diH}}_2\mathrm{O}$ were placed in the pressure cooker in three individual plastic buckets immersed in 500 mL of ${\mathrm{diH}}_2\mathrm{O}$, which covered the bottom of the chamber. The slides were immersed in the sodium citrate buffer and the pressure cooker lid was secured. The temperature was increased to 125°C over a 15-min period, where the pressure reached $\sim 15\mathrm{psi}$. The pressure cooker was removed from the heat, and the temperature and pressure were allowed to decrease to 90°C and 0 psi, respectively, over $\sim 25\min$. The residual pressure was released as the lid was removed and the slides were rinsed in the hot ${\mathrm{diH}}_2\mathrm{O}$. The slides were immediately placed in the hot tris-HCl buffer, where they were incubated for 10 min with the lid on the pressure cooker. The slides were then transferred back to the hot ${\mathrm{diH}}_2\mathrm{O}$ and brought back to RT by slow addition of RT water to the vessel. After 5 min in RT ${\mathrm{diH}}_2\mathrm{O}$, the slides were washed in PBS, pH 7.4 at RT for an additional 5 min.

Antibody staining studies were completed using four groups for staining pattern validation, including (1) conventional indirect IF with unconjugated primary and fluorophore conjugated secondary antibody, (2) the Ab-oligo conjugate plus the same fluorophore conjugated secondary antibody, (3) the Ab-oligo conjugate plus the complementary fluorophore conjugated IS, and (4) the fluorophore labeled primary. The same overall staining procedure was used for each group, explained as follows. The antigen retrieved slides were blocked at RT for 30 min in an Ab-oligo blocking and dilution buffer, which contained 2% bovine serum albumin (BSA, bioWORLD, Dublin, Ohio), $0.5\mathrm{mg}/\mathrm{mL}$ sheared salmon sperm DNA (Thermo Fisher Scientific), and 0.5% dextran sulfate (Sigma-Aldrich, St. Louis, Missouri) in $1\times$ PBS, pH 7.4. The Ab-oligo conjugate was diluted in the Ab-oligo blocking and dilution buffer to a final protein concentration of $15\mu \mathrm{g}/\mathrm{mL}$. The tissue sections were covered with 40 to $100\mu \mathrm{L}$ of the diluted Ab-oligo conjugate and incubated at 4°C overnight in a humidified chamber. The next day, the sections were washed with a $2\times$ saline-sodium citrate (SSC) buffer, pH 7 (VWR, Radnor, Pennsylvania) for 15 min. The sections were fixed in 2% paraformaldehyde (PFA, Sigma-Aldrich) for 15 min at RT and then washed again in a $2\times$ SSC buffer ($3\times 5\min$).

For groups (1) and (2), AF555 conjugated secondary antibody was added at 350 nM. For group (3), complementary IS labeled with AF546, diluted to 350 nM in an IS dilution buffer containing 2% BSA, $0.5\mathrm{mg}/\mathrm{mL}$ sheared salmon sperm DNA, and 0.5% dextran sulfate in a $2\times$ SSC buffer was added. The sections were incubated with IS at RT protected from light for 45 min. The IS was removed, and the sections were washed in a $2\times$ SSC buffer ($3\times 5\min$). The slides were protected from light for the remainder of the staining procedure. Group (4) did not require any secondary detection reagent. 4',6-diamidino-2-phenylindole (DAPI, Thermo Fisher Scientific) was applied to all stained sections at 300 nM for 10 min at RT, and the slides were then washed in a $2\times$ SSC buffer ($2\times 5\min$). The stained slides were mounted in Fluoromount-G (Southern Biotech, Birmingham, Albama) and coverslipped for imaging.

2.4.

Fluorescence Microscopy, Image Visualization, and Image Analysis

Antibody stained slides were imaged on a Zeiss AxioImager.M2 with motorized XY scanning stage (Carl Zeiss AG, Oberkochen, Germany) equipped with a CoolSNAP HQ2 14-bit CCD camera (Photometrics, Tucson, Arizona). Fluorescence images were collected using three channels for IS visualization. Depending on the fluorophore used for IS labeling, the following Zeiss filter sets were used: 38HE (Cy2/AF488), 43HE (Cy3/AF555), and 50 (Cy5/AF647). Excitation light was filtered using the following bandpass (BP) filters 470/40 (HE), 550/25 (HE), and 640/30, for AF488, AF546/AF555, and AF647, respectively. Emission light was filtered using the following BP filters 525/50 (HE), 605/70 (HE), and 690/50, for AF488, AF546/AF555, and AF647, respectively. DAPI was imaged using Zeiss filter set 49 (UV/DAPI), where excitation was filtered at 365 nm and emission was filtered using a 445/50 BP filter. Single-color, single field-of-view images were collected using an X-Cite 120Q (Excelitas Technologies Corporation, Waltham, Massachusetts) light source at $40\times$ (Plan-Apochromat, 0.95NA) magnification. Images of the entire FFPE section were collected with 10% overlap at $20\times$ (Plan-Apochromat, 0.8NA) magnification, where overlapping images were tiled into a single tissue map. Each fluorescence channel was registered using QiTissue software (Quantitative Imaging Systems, LLC, Pittsburgh, Pennsylvania).

Signal-to-background ratio (SBR) was calculated from collected images using a Python script to extract mean fluorescence intensity of marker-specific fluorescent signal and separate the background from each image (DOI: 10.5281/zenodo.3738745). To quantify the marker specific signal, a threshold was established to create a binary mask to extract mean fluorescent intensity from pixels only in positively stained regions of the image. The threshold was established using ImageJ v1.51 (NIH), where the histogram minimum and maximum were adjusted to only display positive pixels. Positive pixels were displayed white and equal to one in the image array while negative pixels were black and equal to zero. For example, a binary mask for a membrane specific marker in which only tissue membrane pixels were white would be created. The binary image array of ones and zeros was used to filter pixels to be counted if their binary image array location had a value of one. To measure background fluorescence, the binary mask was inverted to measure fluorescent signal from pixels not in positively stained regions of the tissue, where, now, the background pixels previously valued as zero in the binary array were equal to one and included in the background fluorescence quantification. SBRs were calculated for image data sets by dividing the signal fluorescent intensity by the background fluorescent signal intensity. Mean fluorescence intensity was used in lieu of SBR for data reporting signal removal in which signal is only present prior to signal removal, making SBR an inappropriate metric for image analysis after signal removal.

2.5.

Ab-oligo Conjugate Titration

The previously detailed FFPE antibody staining procedure was used to stain normal breast tissue with the CK8 Ab-oligo conjugate after being diluted in an Ab-oligo dilution buffer to final concentrations of 15, 5, 1.5, and $0.15\mu \mathrm{g}/\mathrm{mL}$. Images were collected at 60 ms with varying contrast settings to display positive staining at all Ab-oligo concentrations. The images were then used to calculate mean intensity of CK8 specific positive staining, as previously described, to evaluate the optimal staining concentration.

2.6.

Ab-oligo IS Titration

The previously detailed FFPE antibody staining procedure was used to stain normal breast tissue with E-Cad or $\alpha$-SMA Ab-oligo conjugates and triple negative breast cancer tissue with the Ki67 Ab-oligo conjugate. The complementary IS was diluted in an IS dilution buffer and added to the Ab-oligo stained slides at 100, 250, 350, 500, 750, or 1000 nM. Images were collected at 300 ms for E-Cad, 10 ms for $\alpha$-SMA, and 500 ms for Ki67 Ab-oligo conjugates. The displayed images have varying contrast settings to exhibit positive staining pattern at all IS concentrations. The images were used to calculate SBR, as previously described, which was used to evaluate the optimal staining concentration.

2.7.

Optimal IS Fluorophore Configuration

Sk-Br-3 cell buttons were stained with the HER2 Ab-oligo conjugate, and normal breast tissue was stained with the E-Cad Ab-oligo conjugate using the previously described FFPE staining protocol. IS with fluorophore labeled on either the $5^{\prime }$ (i.e., one fluorophore) or both the $5^{\prime }$ and $3^{\prime }$ ends (i.e., two fluorophores) was used for detection. Ab-oligo stained cell buttons or tissues were subsequently stained with 250 nM of the one fluorophore or two fluorophores labeled IS. Images of each stained tissue were collected at 3000 ms for HER2 and 175 ms for E-Cad and were used to calculate SBR to determine the optimal IS configuration.

2.8.

Signal Removal Using Photocleavable Linkers with Varied IS Lengths

Normal breast tissue was stained with the E-Cad Ab-oligo conjugate using the previously described FFPE staining protocol. ISs of varied lengths (28, 27, and 26 mer) that contained a PCL and AF546 fluorophore on both the $3^{\prime }$ and $5^{\prime }$ ends were used for staining at 350 nM. As a positive control, a 28 mer IS with AF546 on both the $3^{\prime }$ and $5^{\prime }$ ends without PCLs was used. Images of the stained tissues were collected at 600 ms. The mean fluorescence intensity of each image was calculated to determine the optimal IS length. Following imaging, the slides were treated with UV light for 15 min on a UVGL-55 Handheld UV lamp (UVP, Upland, California) through the cover glass. The UV-treated slides were placed vertically in $0.1\times$ SSC for 5 min, allowing for removal of the cover glass. The slides were then washed 10 times with $0.1\times$ SSC and mounted using Fluoromount-G prior to imaging using the same settings. The mean fluorescence intensity of the image was again quantified to determine the amount of retained signal.

2.9.

Multiplexed Ab-oligo Staining and Imaging

$\mathrm{HER}2+$ BC and BC tissue microarray (TMA) FFPE samples were stained with a cocktail of the 14 Ab-oligo conjugates using the previously described staining protocol. The 12 antibodies outlined in Table 1 as well as an oligo conjugated estrogen receptor (ER) and PD-1 antibodies were mixed at a concentration of $15\mu \mathrm{g}/\mathrm{mL}$ per antibody into a single cocktail for staining. The tissues stained with the Ab-oligo conjugates were labeled with IS in rounds, where IS labeled with distinct fluorophores and complementary to three Ab-oligos were stained at 350 nM in each staining round. Serial sections were used for staining with the Ab-oligo cocktail or IS only as a negative control. Images were collected in all three channels (AF488, AF546, and AF647), followed by a separate image of the DAPI channel. All stained slides were treated with UV light for 15 min followed by washing 10 times with $0.1\times$ SSC and remounted with Fluoromount-G. Finally, the slides were imaged with the same settings used prior to UV treatment to quantify any remaining signal. Subsequent rounds of IS addition, imaging, and signal removal were repeated until all Ab-oligo conjugates were imaged.

2.10.

Tissue Damage Quantification of Ab-oligo cyCIF

DAPI images from each round of cyclic staining were analyzed using ilastik²⁸ v1.3.3 (European Molecular Biology Laboratory) to quantify the number of nuclei present in the first round and quantify any tissue loss in subsequent rounds. First, a pixel classifier was developed in the Pixel Classification pipeline. The classifier was trained and applied on the full tissue image for each tissue sample to stratify pixels as either a pixel of a cell nucleus or as a background pixel. A map in which each pixel was identified as either a nuclear or background pixel was then used in the boundary-based segmentation with multicut pipeline to create objects from the classified pixels. A watershed algorithm was applied to plant watershed seeds inside each nucleus. The boundary of each watershed object was then exported as the object segmentation map. The object segmentation map was put into the object classification pipeline, where a classifier was trained to identify each object as either background or a cell. After applying the classifier to the full tissue image, a comma-separated values table identifying each object as either background or cell was exported. The number of cell objects for each image was then counted to quantify tissue loss per round of cyclic staining.

3.1.

Staining Pattern Validation of Ab-oligo Conjugates

Optimal Ab-oligo conjugate staining concentrations were evaluated on tissue positive for CK8 [Fig. 1(a)], where titration of the Ab-oligo conjugate was performed from 0.15 to $15\mu \mathrm{g}/\mathrm{mL}$. Quantification of the CK8-specific fluorescent signal of the same structure in serial FFPE tissue sections showed the highest mean intensity using $15\mu \mathrm{g}/\mathrm{mL}$ of the Ab-oligo conjugate [Fig. 1(b)]. Subsequently, the optimal IS staining concentration was assessed on tissues positive for $\alpha$-SMA, E-Cad, and Ki67 [Fig. 1(c)], where IS was titrated from 100 to 1000 nM. The resulting images were qualitatively and quantitatively assessed, showing increased SBR from 100 to 350 nM IS concentrations, with little SBR increase with further increases in IS concentration [Fig. 1(d)], resulting in the selection of 350 nM IS as the optimal staining concentration. A panel of 12 antibodies, selected to elucidate the cell state, breast cancer epithelial cells, stromal compartment, tissue architecture, and immune infiltrate of $\mathrm{HER}2+$ breast cancers (Table 1), were conjugated to a unique DS. The staining pattern of each Ab-oligo conjugate with its complementary IS was confirmed using qualitative comparison to conventional indirect IF staining with an unconjugated primary, indirect IF staining using the Ab-oligo conjugate as the primary antibody, and direct IF staining using a fluorophore-conjugated primary antibody. Negative control tissue was stained with secondary only, IS only, or without any detection reagent (Fig. 2). The staining patterns for all 12 antibodies were qualitatively similar between conventional indirect IF, Ab-oligo indirect IF, Ab-oligo with IS detection, and direct IF staining, demonstrating that oligonucleotide conjugation did not substantially change the affinity of the primary antibody. Four of the 12 antibodies showed similar SBR between conventional indirect IF and Ab-oligo indirect IF staining [Figs. 2(a), 2(e), 2(l), and 2(m)]. Five of the 12 antibodies resulted in a lower SBR when comparing Ab-oligo indirect IF with conventional indirect IF staining [Figs. 2(b), 2(c), 2(d), 2(h), and 2(i)]. The other three Ab-oligo conjugates stained with secondary antibody had higher SBRs than conventional indirect IF staining [Figs. 2(f), 2(j), and 2(k)]. As expected, Ab-oligo with IS staining required longer exposure times than either indirect IF method, generally resulting in decreased SBRs as compared with indirect IF. This required increase in exposure time can be attributed to the signal gain obtained using secondary antibodies that was not possible using the Ab-oligo with our IS staining strategy. As a more direct comparison, primary antibody directly conjugated to AF555 was used to stain serial sections and SBR was quantified for comparison. Overall, similar SBRs were calculated for Ab-oligo detected with IS and direct IF staining; however, there were still sizable differences depending on the stained antigen. For all 12 selected antibodies, the negative control demonstrated that minimal nonspecific background contributed to the overall signal as only nuclear DAPI fluorescence was visible in the normalized control images (Fig. 2). Thus, all 12 Ab-oligo conjugates provided positive staining for the antigen of interest and were able to generate sufficient signal for visualization by conventional fluorescence microscopy.

Fig. 1

Ab-oligo conjugate and IS titration for optimal staining. (a) The CK8 Ab-oligo conjugate was titrated (0.15 to $15\mu \mathrm{g}/\mathrm{mL}$) onto serial sections of normal breast FFPE tissue and equivalent concentrations of IS were applied to all tissue samples for visualization. Images are displayed with contrast and gain optimized for visualization of the positive CK8 staining pattern generated at each antibody conjugate concentration. (b) Image quantification showed the highest mean fluorescence intensity using $15\mu \mathrm{g}/\mathrm{mL}$ antibody concentration for tissue staining. (c) IS was titrated (100 to 1000 nM) onto FFPE tissue with equivalent Ab-oligo conjugate concentrations present of $\alpha$-SMA (staining completed on normal breast tissue, top), E-Cad (staining completed on normal breast tissue, middle), and Ki-67 (staining completed on breast cancer tissue, bottom). (d) The SBR was calculated for each concentration tested for all biomarkers. The $40\text{-}\mu \mathrm{m}$ scale bars are displayed in all images.

Fig. 2

Ab-oligo conjugate staining validation. The staining pattern of each Ab-oligo conjugate was validated by staining serial sections using conventional indirect IF ($1°+2°$), the Ab-oligo conjugate ($1°$) detected with a conventional fluorophore-labeled secondary antibody ($2°$), the Ab-oligo conjugate detected using the complementary IS ($\mathrm{Ab}\text{-}\mathrm{oligo}+\mathrm{IS}$), and direct IF using fluorophore (FL) conjugated $1°$ antibody. The appropriate negative control image is located below their corresponding antibody stained image. Ab-oligo staining pattern was verified for (a) CK5, (b) CK8, (c) CK19, (d) PCNA, (e) Ki67, (f) E-Cad, (g) HER2, (h) $\alpha$-SMA, (i) CD44, (j) CoxIV, (k) CD4, and (l) CD3. SBR was calculated for (m) each indirect IF and $\mathrm{Ab}\text{-}\mathrm{oligo}+2°$ image and compared with (n) SBR calculated for each fluorophore (FL) conjugated $1°$ antibody and $\mathrm{Ab}\text{-}\mathrm{oligo}+\mathrm{IS}$ image. The $40\text{-}\mu \mathrm{m}$ scale bars are displayed in all images.

3.2.

Optimal IS Design

The measured average CR of the 12 Ab-oligo conjugates was near one (average Ab-oligo $\mathrm{CR}=1.22$, Table 1). Thus, a single IS was expected to hybridize to each Ab-oligo conjugate during staining, accounting for the relatively long exposure times compared with indirect or direct IF, where multiple secondary antibodies or fluorophores were available for signal detection, respectively. In an effort to increase the Ab-oligo staining intensity, the IS was modified to contain a fluorophore on both the $3^{\prime }$ and $5^{\prime }$ ends (i.e., two fluorophore design). The Ab-oligo staining intensity was quantitatively compared using identical IS sequences with the one or two fluorophore design for HER2 [Fig. 3(a)] and E-Cad [Fig. 3(b)]. As expected, the additional fluorophore increased the staining intensity, resulting in improved SBR [SBR improvement with two fluorophores versus one fluorophore design: $\mathrm{E}\text{-}\mathrm{Cad}=27.4\%$ and $\mathrm{HER}2=29.9\%$, Fig. 3(d)]. Due to the SBR increase for both HER2 and E-Cad, the IS design with fluorophore labeled on both the $3^{\prime }$ and $5^{\prime }$ ends (i.e., two fluorophores) was selected for all future studies.

Fig. 3

IS design optimization. An IS labeled with a second fluorophore was compared with the original IS design with a single fluorophore for (a) HER2 and (b) E-Cad. (c) IS length with and without a PCL was also investigated using E-Cad labeling, (d) Images were quantified to calculate the generated SBR using an IS labeled with one or two fluorophores as well as (e) with different oligonucleotide lengths (i.e., 26, 27, or 28 nt). The $40\text{-}\mu \mathrm{m}$ scale bars are displayed in all images.

3.3.

Photocleavable Linkers Enable Complete Signal Removal

PCLs were used as the signal removal strategy, where the size of the PCL and conjugated fluorophore was equivalent to approximately two nucleotides (nt). A study was conducted to determine whether the size of the PCL and fluorophore label created steric hindrance at the $3^{\prime }$ end of the IS binding to the DS using the exact 28 mer IS complement to the antibody conjugated DS. Staining using the E-Cad Ab-oligo with a 28 nt IS sequence without PCL was compared with staining using IS of 28, 27, or 26 nt in length with PCL. Each IS was used to stain a consecutive tissue section as a negative control [Fig. 3(c)]. Quantified mean fluorescence intensity was similar for all IS containing the PCL, regardless of sequence length [Fig. 3(e)]. PCL signal removal was validated after each sample was treated with UV light, resulting in complete signal removal for all PCLs containing IS but maintained staining pattern in the 28 nt IS without PCL [Figs. 3(c) and 3(e)]. Although E-Cad staining intensity was not strongly affected by IS length, the 26 nt IS length was selected for further use to ensure that steric hindrance would not diminish Ab-oligo staining of other antigens.

Serial sections of the same test tissue per antigen (Table 1) were stained with IS containing fluorophore on both the $3^{\prime }$ and $5^{\prime }$ ends without PCL and with PCL (Fig. 4). Images of the staining pattern collected prior to UV light treatment demonstrated variable staining intensity with the addition of the PCL to the two fluorophore IS, with some targets showing similar mean fluorescence intensity with and without the PCL [Figs. 4(a), 4(d), 4(f), and 4(j)], some showing lower mean fluorescence intensity with the PCL [Figs. 4(b), 4(c), 4(e), 4(h), 4(i), 4(k), and 4(m)], and one showing improved mean fluorescence intensity with the PCL [Fig. 4(l)]. Negative control slides stained with the IS only, where either the IS with or without the PCL was used, showed that no appreciable signal from nonspecific staining or tissue autofluorescence contributed to overall staining intensity (Fig. 4). Antibody-specific fluorescence signal was completely removed in the PCL containing IS stained samples, where signal intensity was reduced to levels akin to IS only negative control samples. As expected, the Ab-oligo staining pattern was maintained when the IS without PCL was used for staining, although some signal was lost after UV light treatment, which was likely due to photobleaching of the fluorophore label with the high energy UV light (Fig. 4).

Fig. 4

Ab-oligo signal removal validation using PCLs. Ab-oligo conjugate signal removal using PCLs was validated by staining serial sections using Ab-oligo conjugate ($1°$) and IS without a PCL as well as Ab-oligo conjugate and IS with a PCL. Following image collection to demonstrate the Ab-oligo tissue staining pattern, each sample was treated for 15 min with UV light and images were again collected of each sample. Serial sections were stained with IS only and imaged as a negative control. The Ab-oligo staining pattern before and after UV treatment was verified for (a) CK5, (b) CK8, (c) CK19, (d) PCNA, (e) Ki67, (f) E-Cad, (g) HER2, (h) $\alpha$-SMA, (i) CD44, (j) CoxIV, (k) CD4, and (l) CD3. The mean fluorescence intensity for each image before and after UV light treatment was quantified for samples stained (m) with and (n) without a PCL. The $40\text{-}\mu \mathrm{m}$ scale bars are displayed in all images.

3.4.

Multicolor Cyclic Immunofluorescence Staining and Visualization

The validated set of Ab-oligo conjugates and PCL containing IS were used to generate multiplexed IF images of both a $\mathrm{HER}2+$ breast cancer resection specimen (Fig. 5) and a breast cancer TMA (Fig. 6). The cocktail of Ab-oligo conjugates was applied as a single stain, while IS was applied in groups of three, where each IS group had spectrally distinct fluorophore labeling. DAPI staining was used in each imaging round for registration and to monitor for any potential tissue loss throughout the cyclic staining experiment. Tissue scanning was performed in each IS staining round, permitting image tiling for visualization of the entire specimen [Fig. 5(a)]. A magnified field-of-view demonstrated the spatial resolution of the Ab-oligo cyCIF technique, where individual cell staining patterns were readily visualized [Fig. 5(b)]. Staining of the $\mathrm{HER}2+$ breast cancer tissue by round of IS staining showed variation in biomarker expression across the resection specimen for all interrogated biomarkers. As expected, the breast cancer epithelial cell marker HER2 as well as CK8 and CK19 was highly expressed in the breast cancer tumor nests. Proliferative cells, marked by Ki67 and PCNA, were also largely localized to the breast cancer tumor nests. Immune markers, including CD3 and CD4, were localized to the periphery of the tissue sample, mostly isolated from the breast cancer epithelial cells [Figs. 5(a) and 5(b)]. Even with only 12 biomarkers imaged, the difficulty with simultaneous visualization becomes apparent as overlays of more than four to five colors in any single, static image were challenging to visualize and interpret. Using the QiTissue Software, static visualization was enhanced using additional visualization tools, such as height maps to aid in data display [Fig. 5(c)].

Fig. 5

Multiplexed Ab-oligo cyCIF on $\mathrm{HER}2+$ BC. (a) Tiled and stitched, whole breast cancer tissue images of Ab-oligo cyCIF are displayed by round of IS application. (b) A magnified view of a breast cancer tumor nest within the tissue sample (white box) is displayed for a higher resolution view of the staining pattern of each marker in their respective round of imaging. (c) QiTissue software was used to generate enhanced visualizations of this Ab-oligo cyCIF imaging data set by displaying CK8 as a height map. The height map is colorless, resulting in enhanced visualization of colocalization of markers without increased false coloring.

Fig. 6

Fourteen-color Ab-oligo cyCIF on BC subtypes stained in a BC TMA. The validated Ab-oligo conjugates were used to stain a TMA containing different subtypes of BC. The biomarkers in the collected images were organized into tissue characterization panels of differentiation (CK5, CK8, and CK19), proliferation (CoxIV, PCNA, and Ki67), microenvironmental ($\alpha$-SMA, PD-1, CD4, and CD3), and disease-specific ($\alpha$-SMA, E-Cad, ER, CD44, and HER2) biomarkers, which were then used to characterize (a) $\mathrm{HER}2+$, (b) $\mathrm{HER}2-$ and $\mathrm{ER}-$, and (c) $\mathrm{ER}+\mathrm{BCs}$.

Multicolor cyCIF staining in a breast cancer TMA was completed using the same rounds of three color IS application and UV light treatment. However, to help demonstrate the power of cyCIF, the biomarkers were remixed into biologically relevant groups, including differentiation (CK5, CK8, and CK19), proliferation (CoxIV, PCNA, and Ki-67), microenvironmental ($\alpha$-SMA, PD-1, CD4, and CD3), and breast cancer-specific ($\alpha$-SMA, E-Cad, ER, CD44, and HER2) markers rather than by imaging round (Fig. 6). Visualization of the selected biomarkers on different breast cancer subtypes showed the expected distinct expression patterns for $\mathrm{HER}2+$ [Fig. 6(a)], $\mathrm{HER}2-$, $\mathrm{ER}-$ [Fig. 6(b)], and $\mathrm{ER}+$ [Fig. 6(c)] breast cancers. As expected, the $\mathrm{HER}2+$ breast cancer had the highest HER2 staining intensity, while the $\mathrm{ER}+$ breast cancer had the highest ER staining intensity. CK5 was only expressed in the $\mathrm{ER}+$ breast cancer, with minimal expression in either of the other two breast cancer subtypes. CK8 and CoxIV showed ubiquitous expression across the three breast cancer subtypes, while $\alpha$-SMA was found mainly in benign ductal breast tissues (Fig. 6). Tissue loss per round was quantified for all three TMA cores displayed by automated counting of DAPI labeled nuclei in each round of imaging, resulting in an average tissue loss per round of $0.52\%\pm 0.37$.