Click Chemistry-Based DNA Labeling of Cells for Barcoding

Jul 22, 2018 - Stefan D. Gentile , Megan E. Griebel , Erik W. Anderson , and Gregory H. Underhill*. Department of Bioengineering, University of Illino...
1 downloads 0 Views 4MB Size
Download PDF
Article pubs.acs.org/bc

Cite This: Bioconjugate Chem. 2018, 29, 2846−2854

Click Chemistry-Based DNA Labeling of Cells for Barcoding Applications Stefan D. Gentile, Megan E. Griebel, Erik W. Anderson, and Gregory H. Underhill* Department of Bioengineering, University of Illinois at Urbana−Champaign, Urbana, Illinois 61801, United States

Downloaded via UNIV OF TEXAS AT EL PASO on November 4, 2018 at 08:08:47 (UTC). See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

S Supporting Information *

ABSTRACT: Cell labeling and tracking methodologies can play an important role in experiments aimed at understanding biological systems. However, many current cell labeling and tracking techniques have limitations that preclude their use in a variety of multiplexed and high-throughput applications that could best represent the heterogeneity and combinatorial complexity present in physiologic contexts. Here, we demonstrate an approach for labeling, tracking, and quantifying cells using double-stranded DNA barcodes. These barcodes are introduced to the outside of the cell membrane, giving the labeled cells a unique identifier. This approach is compatible with flow cytometric and PCR-based identification and relative quantification of the presence of barcode-labeled cells. Further, utilizing this strategy, we demonstrate the capacity for sorting and enrichment of barcoded cells from a bulk population. In addition, we illustrate the design and utility of a range of orthogonal barcode sequences, which can enable the use of multiple independent barcodes to track, sort, and enrich multiple cell types and/or cells receiving distinct treatments from a pooled sample. Overall, this method of labeling cells has the potential to track multiple populations of cells in both high-throughput in vitro and physiologic in vivo settings.

■

antibodies.19−21 However, there are still several restrictions related to the applicability of these techniques. First, the sensitivity of detecting fluorescent DNA can be limited in some approaches. In addition, similar to traditional fluorescent probes, multiplexing capabilities can be limited when leveraging fluorescent DNA oligo-based techniques. Further, the addition of a zinc-finger probe to the cell surface requires the cell to be genetically modified prior to the initiation of the experiment, and further, there is potential cross reactivity between zinc-fingers and nonspecific DNA oligos that they are not designed to capture. To mitigate some of these issues, recent efforts have begun to explore alternative methods for introducing oligos, as well as a variety of other probes, to the cell membrane. This approach is based on the modification of cell surface glycoproteins with artificial sugars that can serve as attachment points through bio-orthogonal click chemistry.10,11,18 In our studies, we illustrate a straightforward method for labeling cells with DNA oligo barcodes that is broadly compatible with both flow cytometric and PCR-based detection and quantification (Figure 1). This approach enables the independent cell labeling, and subsequent, relative cell frequency quantification in mixed populations. In addition, we demonstrate the multiplexing capabilities of this approach as

INTRODUCTION The ability to label and track cells is imperative to observing and understanding cell fate processes. There are several methods often used to both label and track cells in vivo, including but not limited to fluorescent dyes,1,2 nanoparticles (such as quantum dots),3−5 viral constructs,1,6,7 and metal and radiotracers.6,8 However, these techniques have limitations and drawbacks when trying to increase throughput and multiplexing. In particular, using fluorescence to differentiate between cells of interest is limited by the number of nonoverlapping independent fluorescent signals, and further, many of these methods require extensive cellular modification. The attachment of DNA oligonucleotides (oligos) to the cellular membrane has been used for almost a decade to pattern cells onto substrates and microarrays through DNA hybridization.9−18 In this process, labeled cells interact with a surface previously conjugated with a complementary DNA oligo. In addition, the DNA oligos can be used to label or image the cells by modifying the DNA backbone to incorporate or bind to a fluorescent probe. In the past, Nhydroxysuccinimide (NHS) ester-modified DNA oligos have been used to attach to primary amine groups on the cell surface, leaving a DNA oligo anchored to the cell membrane.9,18 Other methods for attaching DNA to the cell membrane include the expression of zinc-finger proteins on the cell membrane that specifically attach to DNA oligos, the use of artificial lipids conjugated to DNA oligos to integrate directly into the membrane, or the use of DNA tagged © 2018 American Chemical Society

Received: June 19, 2018 Revised: July 20, 2018 Published: July 22, 2018 2846

DOI: 10.1021/acs.bioconjchem.8b00435 Bioconjugate Chem. 2018, 29, 2846−2854

Article

Bioconjugate Chemistry

Figure 1. Method for incorporating DNA oligo barcodes: (A) The addition of ManNAz to the cells under culturing conditions adds azides to the outside of the cell for 48 h. (B) To prepare the DNA barcodes, the cell bind DNA oligo is reacted with succinimidyl ester DIBO alkyne, and subsequently, the complement DNA oligo is added and allowed to hybridize. (C) The complete double-stranded DNA structure is introduced to the ManNAz-modified cells allowing for the reaction of the DIBO alkyne with the membrane-presented azide leading to the attachment of the oligo barcode to the cell membrane. Purple, biotin; green, streptavidin; red, Alexa-Fluor 647.

fluorescence microscopy. The unmodified cells (Figure 2A) showed no detectable Alexa-Fluor 647 signal, while the ManNAz-modified cells (Figure 2B) exhibited a strong Alexa-Fluor 647 signal. Similarly, flow cytometric analysis demonstrated that ManNAz-modified cells exhibited a substantially higher signal (approximately 3 orders of magnitude) than unmodified cells. Next, we sought to examine the duration of the detectable azide following the ManNAz treatment. ManNAz-modified cells were harvested after 2 days in the ManNAz containing medium, then replated in growth medium under standard conditions without ManNAz. Every 2 days, the cells were harvested, and a subset was separated out to stain with DIBO Alexa Fluor 647 for flow cytometry, while the rest were replated under standard growth conditions. These data demonstrated that the introduced azide moieties were still detectable up to 4 days post-ManNAz treatment using flow cytometry (Figure 2C). This process for modifying cells with ManNAz is advantageous because it can be utilized independent of cell type and does not require any genetic modification in order to add attachment points to the surface of the cell. In addition, the azide−alkyne reaction employed is bio-orthogonal to natural cell processes and can be performed under aqueous and standard growth conditions.23−25 Live Cell Labeling with Double-Stranded DNA. After determining that the ManNAz successfully integrated into the glycoproteins in the cell membrane, we tested the capability to incorporate specific DNA barcodes through attachment to the ManNAz-modified cells. A549 cells were seeded onto glass coverslips and once again treated with culture media supplemented with ManNAz. The cells were then treated

well as the sensitive detection of a small number of cells (as low as 0.01% of total) within a larger bulk population.

■

RESULTS AND DISCUSSION Overview of Barcoding Approach. A schematic overview of the experimental methodology is provided in Figure 1. In particular, the key steps in the barcoding procedure are the following. First, cells of interest are cultured in medium containing ManNAz to introduce cell surface azides as attachment points. In parallel, a 5′ amine-modified singlestranded DNA oligo (the “cell-bind” DNA or the single oligo of DNA that will be directly bound to the cell surface) is reacted with succinimidyl ester DIBO alkyne to introduce the cyclocotyne needed for later reaction with the cell surface azide. Next, the complementary DNA oligo is added and allowed to hybridize in a sequence-specific manner. The double-stranded DNA complex is then added to azidepresenting cells to covalently attach the complex to the cell membrane surface. Here, in our presented studies, we experimentally validated and optimized these steps as well as demonstrated the utility of this barcoding approach. Further, additional methodological details are provided in Materials and Methods. Confirmation of Cell Surface Azide Addition. Initially, we aimed to confirm that the ManNAz artificial sugar was sufficiently integrating into the glycoproteins containing sialic acid on the cell membrane.22−27 To verify this, we cultured A549 cells on glass coverslips in a 12 well plate for 3 days, as further outlined in the Materials and Methods. Unmodified A549 cells were cultured in parallel with the cells treated with ManNAz. Both modified and unmodified cells were labeled with DIBO Alexa-Fluor 647, fixed, and imaged using 2847

DOI: 10.1021/acs.bioconjchem.8b00435 Bioconjugate Chem. 2018, 29, 2846−2854

Article

Bioconjugate Chemistry

Figure 2. Confirmation of the presence of cell-surface azide by ManNAz addition: (A) Unmodified cells stained with DIBO Alexa-Fluor 647 and DAPI. (B) Cells modified by adding 50 μM ManNAz to cells under standard growth conditions stained with DIBO Alexa-Fluor 647 and DAPI. (C) Cells modified with ManNAz for 48 h (under standard conditions), passaged, and replated every 48 h. At the 48 h increments, cells were stained with DIBO Alexa-Fluor 647 in solution and analyzed via flow cytometry. Azides were detectable on the cell surface 96 h after treatment.

with sequence 1 double-stranded DNA (dsDNA) that had previously been prepared. Following dsDNA incubation, the cells were treated with Alexa-Fluor 647 streptavidin, fixed, mounted, and imaged. Fluorescence imaging assessment of cells pretreated with ManNAz showed that the barcodes successfully attached to the azides expressed on the cells surface (Figure 3A). Barcode sequence 1 was prepared and later added to two populations of cells: untreated A549 cells and ManNAz treated A549 cells. Alexa-Fluor 647 streptavidin was introduced, for binding to the biotinylated 5′ end of sequence 1 complement, and both populations were analyzed using flow cytometry. ManNAz-modified cells demonstrated an approximately 10fold higher amount of labeling compared to unmodified cells. The flow cytometric signal obtained for the unmodified cells tagged with dsDNA is higher than that of unmodified cells that did not receive DNA treatment, suggesting that there was some nonspecific attachment of the DNA to the cells during the labeling process. The addition of the salmon sperm DNA to both the dsDNA and to the cells before the addition of the DNA helped to block the unspecific attachment of DNA, modestly lowering the florescent signal of the unmodified cells. Consequently, the co-treatment with nonspecific salmon sperm DNA was included in all of the labeling studies presented here. Further, the viability of cells post-ManNAz and ManNAz plus DNA treatments was confirmed using propidium iodide labeling and flow cytometry. In particular, neither of these treatments significantly influenced cell viability, with viable cells representing >98% of the analyzed cells for each of these

labeled populations, which was also nearly identical to unlabeled cells analyzed in parallel (Supplemental Figure 1). Next, we aimed to quantify the amount of DNA labeling using PCR. Specifically, after the labeling process and using a defined number of cells, the relevant cell suspension was heated to dissociate the complementary oligo, and the supernatant was subsequently collected for PCR analysis. Using this method of quantification, we determined that on average a 9.3-fold greater amount of DNA (p < 0.01) was associated with the membrane of ManNAz-modified cells compared to the unmodified control cells (Figure 3C). This is similar to previously reported efforts to introduce DNA to cells, including studies by Vogel et. al, in which they demonstrated between a 12.7 and 1.2 signal-to-noise ratio for metabolic labeling with an initial 50 μM ManNAz incubation.18 DNA Label Detection during Cell Culture. We additionally sought to examine the duration of detectable DNA barcodes following the labeling process. Specifically, to determine the length of time that the DNA remained on the cell, cells were labeled with the DNA barcode oligos, a cell sample was removed and labeled with Alexa-Fluor 647 streptavidin and analyzed with flow cytometry, and the remainder of cells were replated into different 6-well plates. No effects on cell viability, cell attachment, or cell proliferation rates were observed. At 2, 8, 24, and 48 h post-seeding, the cells were removed, stained with Alexa-Fluor 647 streptavidin, and analyzed via flow cytometry (Figure 4A). Flow cytometric analysis demonstrated that after 8 h, the amount of detectable 2848

DOI: 10.1021/acs.bioconjchem.8b00435 Bioconjugate Chem. 2018, 29, 2846−2854

Article

Bioconjugate Chemistry

Figure 4. DNA barcodes detected via flow cytometry and PCR after replating labeled cells. Cells were harvested and labeled with DNA oligo sequence 1 and then re-introduced to standard growing conditions. At various intervals, the cells were harvested, stained with Alexa-Fluor 647 streptavidin, and analyzed with flow cytometry and PCR. (A) After 2 h under growth conditions, harvested cells had much lower fluorescence than at 0 h. The signal continued to drop over the next 48 h until it was undistinguishable from the control. (B) The same samples from (A) were analyzed using PCR. After 48 h, DNA barcode signal remains detectable and is several orders of magnitude above the unlabeled control condition. *:p < 0.05, compared to 0 h time point.

Figure 3. ManNAz-modified cells are successfully labeled with DNAdsDNA successfully attached to cell surface azides on modified cells using previously described methods: (A) A549 cells barcoded with biotinylated dsDNA attached to streptavidin Alexa-Fluor 647. (B) ManNAz-modified cells showed close to a 10-fold increase in fluorescence compared to unmodified cells and a 1000-fold increase over control cells using flow cytometry. (C) Using our PCR-based method for quantifying the number of DNA oligos attached to the cell surface, ManNAz-modified cell had over 9-fold more DNA oligos than unmodified cells. *:p < 0.01, **:p < 0.00001, related to no ManNAZ.

amount of DNA remaining on the cells is 0.191% (±0.066%). At each time point, we removed and collected the complementary DNA oligo from 10,000 cells and performed PCR with the appropriate primers to determine the relative amount of DNA remaining compared to the 0 h time point (Figure 4B). From 8 h post-replate to 48 h post-replate, the amount of DNA on the cell continues to decrease, albeit at a much slower rate than during the initial 8 h post-labeling. Of the DNA remaining after 8 h, 31.1% (± 4.27%) remains detectable after 24 h. At the 48 h time point, there was approximately 26.4% (±4.29%) of the amount detected 24 h. Overall, these measurements determined that approximately 0.0126% of the original DNA introduced to the cells remained detectable after 48 h under culturing conditions, and this DNA is still clearly detectable with PCR. Specifically, these samples demonstrated a sufficient PCR-based measurement that was >3 orders of magnitude higher than unlabeled control cells (Figure 4B). Taken together, these data demonstrate the capability to use both flow cytometry and PCR for detection

DNA label decreased substantially, by over 2 orders of magnitude. After 24 and 48 h, the fluorescent signal from the DNA barcodes further decreased, but not nearly as sharply as between the 0 and 8 h time points. However, the fluorescent peaks for 8, 24, and 48 h post-replating are very similar and cannot therefore be used to differentiate between the different time points (Figure 4A). For this reason, PCR was necessary to quantify the amount of DNA remaining at longer time intervals, while the flow cytometry was sufficient at shorter time points. Using PCR, we determined that after 8 h, the 2849

DOI: 10.1021/acs.bioconjchem.8b00435 Bioconjugate Chem. 2018, 29, 2846−2854

Article

Bioconjugate Chemistry

Figure 5. Quantitative analysis of labeled cells unlabeled bulk populations. (A) Schematic of experimental design. Various ratios of labeled cells to unlabeled cells were mixed together and analyzed using flow cytometry and PCR. (B) Several ratios of barcoded to unlabeled cells were mixed to define distinct populations. Flow cytometric quantification of the number relative amount of barcoded cells to unlabeled cells in the sample. (C) Relative amount of DNA oligo present in each sample (relative to 100% barcoded cells), as determined with PCR. Note: * denotes theoretical value based on initial fraction.

Figure 6. Assessment pre- and post-enrichment. (A) 2% labeled cells (sequence 1 with attached fluorescent streptavidin) were mixed with unlabeled cells and then sorted using FACS utilizing Alexa-Fluor 647-conjugated sequence 1. (B) Flow cytometric dot plot of pre-enrichment population. (C) Post-sorting, the amount of DNA in the samples was quantitatively compared to both the 100% barcoded control and the 2% presort using PCR. *: p < 0.001, **:p < 0.0001, relative to 2% labeled condition.

2850

DOI: 10.1021/acs.bioconjchem.8b00435 Bioconjugate Chem. 2018, 29, 2846−2854

Article

Bioconjugate Chemistry

Figure 7. Orthogonality of multiple DNA barcodes. (A) Equal amounts of cells with barcodes 1−4 together and separated via FACS. Sequence 1(Seq 1), with an attached fluorophore concentrated in the post-sort (+) group and with sequences 2−4 sorted into post-sort (−). (B) Integration of multiple DNA barcodes with cell phenotype. Instead of sorting based on a DNA sequence, sort was based on GFP of cell. Cells expressing GFP labeled with sequence 1, while normal cells labeled with sequences 2−4 (Seq 2−4). GFP+ cells sorted into post-sort (+) and sequence 1 is concentrated in that sample. *: p < 0.05, **:p < 0.005, relative to pre-sort condition.

and quantification at early time points. Further, at later points, following culture and cell proliferation, the measurement of cell associated DNA oligo barcodes with PCR was determined to be possible and quantitatively robust. Quantitative Analysis within a Population. Building on the capability to quantify the amount of DNA barcode labeling, we aimed to determine if we could identify barcoded cells within a population of unlabeled cells. Barcoded cells were mixed together with unlabeled A549 cells in various ratios, then stained with Alexa-Fluor 647, and analyzed with flow cytometry (Figure 5A). Down to ratios as low as 10% barcoded cells, the relative presence of barcoded/unbarcoded ratio is distinguishable by flow cytometry (Figure 5B). Utilizing the PCR-based method, we were able to distinguish barcoded cells among unbarcoded cells at a 0.01% ratio, much lower than using flow cytometry. Specifically, with this approach, we were able to determine the fold difference between the theoretical ratios and the actual values of barcoded cells to unbarcoded cells. For these defined ratios, the measured value was between 1.04- and 3.03-fold of the theoretical amount (all but one below 2.3 fold difference). Further, the detection of the relative presence of labeled cells is independent of the presence of cells labeled with a different barcode sequence (Supplemental Figure 2). In addition, we utilized fluorescence activated cell sorting (FACS) to separate out barcoded cells from a bulk population

of unlabeled cells. First, to generate a predetermined mixed population, 2% sequence 1 (with biotin) barcoded cells were mixed with 98% unmodified cells and then stained with AlexaFluor 647 streptavidin. This mixed population was analyzed using FACS, in which the cells barcoded with DNA were separated from the unmodified cells (Figure 6A). The amount of DNA in the pre- and post-FACS samples was quantified with PCR (Figure 6B,C). These experiments demonstrated that following FACS-based separation, there was 36.0 times the DNA in the positive sort (± 5.71) than the initial sample and a 0.094 fraction of DNA (± 0.00265) in the negative sort. Overall, this demonstrates that the DNA oligo-based barcoding approach is compatible with FACS-based cell enrichment strategies and associated PCR quantification of the fold enrichment. Orthogonality of Multiple DNA Barcodes. Using PCR and the associated specific primers for four distinct DNA barcode sequences, we aimed to demonstrate the multiplexing capability of this approach. First, we demonstrated that the PCR-based quantification of the barcode sequences was specific and orthogonal. In particular, the primers independently amplified the associated barcode sequences in PCR (Supplemental Figure 3). It is therefore potentially possible to mix several different populations of barcoded cells together and then determine the overall percentage of each of the cells in the mixture with PCR. 2851

DOI: 10.1021/acs.bioconjchem.8b00435 Bioconjugate Chem. 2018, 29, 2846−2854

Article

Bioconjugate Chemistry

barcoded cells in pooled populations. Lastly, this barcoding method can support a variety of cell enrichment strategies. Ultimately, this method may provide utility for tracking cells in vivo, as it would only require the recovery of a few barcoded cells to measure with PCR. In addition, this approach could be applied toward barcoding distinct cell types in a co-culture context, for analyzing cell−cell interactions12,28,29 and for tracking cell fates based on prevalence of the initial barcode.16,30,31

After confirming the orthogonality of each barcode, we examined the multiplexing potential with a series of experiments utilizing predetermined mixed populations. First, four different sets of cells were barcoded with different sequences (sequences 1−4), and subsequently the cells were mixed in equal ratios (∼25% each). Only sequence 1, with its biotinylated end is detectable with fluorescence, and FACS was used to separate these cells from the distinctly labeled other cell sets based on the Alexa-Fluor labeling of the sequence 1 DNA barcode (Figure 7A). After FACS-based enrichment, the cell samples were analyzed using the PCR quantification protocol to determine the relative amount of each sequence in the original sample (pre-FACS) and the two post-FACS samples (samples positive for sequence 1/AlexaFluor 647 and samples negative for sequence 1/Alexa-Fluor 647). In the pre-FACS sample, all of the sequences were represented nearly equally, with each represented at approximately 25%. However, in the post-sort samples, these ratios were substantially changed, as expected. Specifically, the negative Alexa-Fluor 647 post-sort showed a significant (p < 0.05) decrease in the amount of sequence 1 from 25.11% to 3.97%, while the percentage of the other three sequences (2− 4) increased. The opposite was true of the positive post-sort. There, only the percentage of sequence 1 significantly increased (p < 0.005) to 91.5%, while the others all significantly decreased to no higher than 2.8% (p < 0.05 for sequence 2, p < 0.005 for sequences 3 and 4) (Figure 7A). In order to demonstrate that the DNA oligo barcodes remain attached to the cells they are designed to detect, we reran the previous experiment with an important difference. Instead of utilizing FACS to sort out cells based on the fluorescent DNA barcode sequence, cells were instead sorted based on a cell-based fluorescence signal (cell expression of GFP) (Figure 7B). GFP-modified cells (with ManNAz treatment) were barcoded with one sequence, while nonGFP-modified cells (also with ManNAz treatment) were barcoded with one of the remaining three sequences (sequences 2−4). As before, the samples were mixed in equal ratios and then analyzed and sorted using FACS, with the pre- and both post-samples saved for PCR analysis. In the negative GFP sorted sample, there was a significant decrease in the amount of sequence 1 (from 24.8% to 3.71%, p < 0.05), while the other three sequences increased (from 23.8% to 27.2% for sequence 2, 25.7% to 34.7% for sequence 3, and 25.6% to 31.5% for sequence 4). In the positive GFP sorted sample, there was a significant decrease in the presence of sequence 2−4 (