in-situ Sensing of Reactive Oxygen Species on Dye-Stained Single DNA Molecules under Illumination

Abstract

Reactive oxygen species (ROS) are a necessary evil in many biological systems and have been measured with fluorescent probes at the ensemble levels both in vitro and in vivo. Measuring ROS generated from a single molecule is important for mechanistic studies, yet measuring ROS near a dye-labeled single-molecule under illumination has been challenging. Here we use CellROX, a group of ROS probes, to sense ROS near dye-stained DNA that has been flow-stretched and immobilized on a surface. ROS is responsible for the photodamage of DNA molecules under this circumstance. In this report, we confirmed the ROS sensing reaction in bulk solutions and optimized the conditions for single-molecule experiments including the selection of substrates, dye concentrations, probes in the CellROX series, excitation lasers, and emission filter-sets. We observed a correlation between ROS and the dye-labeled DNA and localized the ROS-activated CellROX probe molecules at both the ensemble level and the single-molecule level.

Graphical Abstract

Introduction

Reactive oxygen species (ROS) are a class of oxygen-containing molecules that are highly reactive and have been the focus of many studies. Common ROS, such as superoxide, singlet oxygen, hydroxyl radicals, and hydrogen peroxide, are necessary evils in biological systems.¹ ROS are important in cell signaling,^1–4 but at large levels, they can also cause cell damage, commonly known as oxidative stress.^3,5 ROS are believed as a cause of ailments^6–9 and cancer.^10,11 Interestingly, ROS have also been used as a method of cancer treatment.^12–14 Monitoring the production of ROS is important for many applications. For example, common intercalating DNA stains, such as the YOYO-POPO-TOTO series, are known to cause DNA damage through the production of ROS.^15–17 ROS cut DNAs into small pieces (photocleavage), which is problematic for DNA imaging particularly single-DNA imaging. Thus, monitoring the generation of ROS and measuring its ability for photocleavage is an important step towards finding a solution to this problem.

Monitoring ROS at the cellular level and above has already been commonly reported.^18,19 For example, A genetically encoded yellow fluorescent protein (YFP)‐based H₂O₂ sensor, named HyPer, has been used to probe H₂O₂ concentrations in live cells.¹ A set of commercially available dyes to detect ROS is the CellROX® series from Thermo Fisher. CellROX dyes have been used to monitor ROS in living cells as well.^20,21 CellROX dyes are nonfluorescent before activation. After they are activated by the ROS, they are fluorescent with different colors. Thus, monitoring the growth of the CellROX fluorescent signal indicates the production of ROS in the solution.

Probing interactions at the single-molecule level can elucidate new insights indiscernible at the bulk level and gain quantitative information about ROS generation and suppression mechanisms such as using oxygen scavengers. This is particularly important for single-DNA measurements. Yet to the best of our knowledge, there are no reports of directly monitoring ROS at the single-molecule level on dye-labeled DNA. In this work, we used CellROX dyes to probe ROS generated in-situ from the intercalating DNA stains YOYO-1 and POPO-1 (structures shown in Fig. 1).

Figure 1.

Molecular structure of YOYO-1 and POPO-1.²² The difference is that YOYO-1 has the two rings shown in green, while POPO-1 does not.

Experimental

All DNA used is λ-DNA (hereby referred to as DNA) from Fisher Scientific. Unless otherwise specified, the buffer used is an aqueous solution of 10 mM HEPES and 10 mM NaCl at pH 7.0. All DNA is dissolved in this buffer.

UV/vis. UV/vis spectra were collected on an Agilent 8453 spectrometer for aqueous solutions of (1) 100 μM CellROX Red (Fisher), (2) 100 μM CellROX Red and 20 μM YOYO-1 (Fisher), and (3) 100 μM CellROX Red, 20 μM YOYO-1, and 0.15 μg/μL DNA. For solution (3) CellROX Red was added last to ensure YOYO-1 was intercalated into DNA without interference. Spectra were collected after 0, 5, 20, and 40 min irradiation times of each solution with a 473 nm laser with a beam diameter large enough to cover the whole cuvette. The laser power density was 8.1×10⁻³ W/cm² measured using a Coherent Fieldmax II power meter equipped with an op-2-vis sensor. One spectrum was also taken of activated CellROX Red in an aqueous solution of 100 μM CellROX Red, 10 μM H₂O₂, and 100 μM CuSO₄. In this case, CellROX Red could be activated by hydrogen peroxide, hydroperoxyl radicals, or hydroxyl radicals produced by the reactions shown below. The reduction of copper by hydrogen peroxide followed by a Fenton like reaction has been reported in the literature.^23–25 The CellROX Red peak was selected for each spectrum and background corrected by fitting a linear line to the background and subtracting it from each spectrum. This was accomplished by using a home-made MATLAB code.²⁶

$$C{u}^{\mathrm{2+}}+H_2{O}_2\to C{u}^{\mathrm{1+}}+H{O}_2^{•}2+H^{+}$$ (1)

$$H_2{O}_2+C{u}^{\mathrm{1+}}\to C{u}^{2+}+O{H}^{-}+O{H}^{•}$$ (2)

Fluorescence spectra for solutions 1-3 described above were taken using a PTI fluorimeter (Photon Technology International) equipped with an 814 photomultiplier detection system and LPS-220B lamp power supply. Two spectra were taken for each solution; one using 473 nm excitation light and the second using 640 nm excitation light. The 473 nm laser was used to excite YOYO-1 and the 640 nm laser was used to excite the activated CellROX Red that was produced. Solution 3 was diluted by a factor of 100 for the spectrum taken using 473 nm excitation light only to avoid saturating the detector.

Sample Substrate Preparation.

Amine modified glass surfaces were created similar to previously reported.²⁷ Briefly, base piranha clean glass coverslips were immersed in a solution of 0.5% 3-aminopropyltriethoxy silane (TCI America) in HPLC grade acetone (Fisher) for 20 minutes. A less-efficiently modified amine surface was functionalized using 0.01% of 3-aminopropyltriethoxy silane in acetone for 20 min. Then they were washed with ethanol and water and dried with nitrogen gas. Polymethyl methacrylate (PMMA, Sigma-Aldrich, average MW 350,000 g/mol) surfaces were fabricated by spin coating a solution of 15 mg/mL PMMA in chlorobenzene onto clean coverslips at 3000 rpm for 30 s. Afterwards, the thin films were annealed in the oven at 160° C for 30 min. Polydimethylsiloxane (PDMS) blocks were made as previously reported.²⁷ Briefly, Sylgard 184 silicone elastomer base (Dow Corning) was mixed with the provided Sylgard 184 elastomer curing agent (10:1 by mass). Then it was put in a petri dish, vacuumed to remove air bubbles, and left overnight to cure. Microfluidic channels were constructed by inserting syringe tips into the cured PDMS and adhering the modified coverslips by sandwiching parafilm between the coverslip and PDMS followed by heating at 50° C to glue them together. Rectangles were cut out of the parafilm to make the channel with dimensions of roughly 20 mm × 1 mm (length, width). The height was ~100 μm which is defined by the thickness of the parafilm.

DNA Photocleavage measurements.

All single-DNA imaging experiments were carried out on the same home-built fluorescence microscope described in previous work.^27,28 DNA was flow- stretched on the amine substrates²⁷ by flowing a solution of 300 pg/μL YOYO-1 labeled DNA (dye : DNA bp ratio of 1:40) at 300 μL/min and then washing with ~0.5 mL of buffer. DNA was stretched onto a PMMA surface by a receding water meniscus as the 1 μL droplets of DNA solution evaporates from the surface while heating. Identical conditions were used to image both samples consisting of 50 μL/min buffer flow, 5.7 W/cm² laser power density, and the Andor iXon Ultra 897 EMCCD camera with 300× EM gain and 50 ms exposure time.

CellROX Generation from Single POPODNA by fluorescence microscopy.

POPO-1 labeled DNA (POPODNA) was stretched onto the amine modified surface by flowing at 300 μL/min and then washed with ~0.5 mL of buffer. A volume of 100 μL of 2.5 μM CellROX Green was injected into the microfluidic channel at 50 μL/min and then washed with 1.5 mL of buffer. The flow was then stopped. Any videos of POPODNA were recorded using a Chroma TRF49901 filter-set with the 460/50m band pass removed to collect the full spectrum of POPO-1. The emission filter-set used is shown in Fig. S1a and is henceforth referred to as the purple filter-set. The POPODNA was excited using a 405 nm laser (Dragon lasers, China). Videos were recorded a few seconds before the 405 nm and 473 nm lasers were un-blocked and continued for about 100 seconds after. The 405 nm laser was used to excite POPO-1 and the 473 nm laser was used to excite the activated CellROX Green. A chroma TRF49904 filter-set was used with an added external filter (542 nm long pass ET542lp chroma) to the detection pathway to block the fluorescence from POPO-1. Emission filter-set is shown in Fig. S1b and is henceforth referred to as the blue filter-set. An EM gain of 300 was used for all videos.

Results and Discussion

ROS sensing in the bulk.

Before moving to the single-molecule level, we tested if irradiating YOYO-1 in bulk solution will activate CellROX Red. If YOYO-1 or POPO-1 generates radicals upon light illumination, these radicals should activate CellROX Red. CellROX Red is not fluorescent before activation but is fluorescent after activation, whose emission can be excited by a 640 nm light.

Control experiments of inactivated and activated CellROX Red with no DNA staining dyes were carried out first. The control solution of activated CellROX is obtained chemically by a copper catalyzed hydrogen peroxide reaction explained in the experimental section. The UV-Vis absorption spectra of activated and inactivated CellROX Red are significantly different (Fig. 2a). The the inactivated species has absorption peak at ~350 nm with a tail to 800 nm. It is stable under 473 nm light for the tested 40 min with a small amount of activation, where no significant decrease of the 350 nm peak (Fig. 2a), and an insignificant growth of the 640 nm peak is observed (Fig. 2d). The activated curve shows an absorption peak at ~640 nm (highlighted with a dashed box in Fig. 2a).

Figure 2.

Absorbance and fluorescence emission spectra for solutions of CellROX Red (left column), CellROX Red and YOYO-1 (middle column), and CellROX Red and YOYODNA (right column). (a-c) Absorption spectra. (d-f) Baseline-treated absorption peak in (a-c). A Gaussian function was fitted to each peak as a guide for the eye. The green spectrum in (a) is from a solution of CellROX Red activated by the chemical reactions (see the experimental section). (g-i) Fluorescence emission spectrum with excitation at 473 nm (blue) or 640 nm (red). The counts due to the excitation light were removed for clarity resulting in the gap in the red spectra. The solution was diluted by a factor of 100 for the blue curve in (i) to avoid detector saturation.

When solutions of inactivated CellROX Red were mixed with YOYO-1 or YOYO-1 labeled DNA, 473 nm light is able to photo-activate the CellROX Red because YOYO-1 generates ROS under this light. This is more obvious by comparison with the control experiments from the growth of the 640 nm absorption peak and decease of the 350 nm peak over the irradiation time of 0, 5, 20, and 40 minutes (Fig. 2a–c).

The fluorescence emission spectra of each solution after 40 min of 473 nm irradiation are shown in Fig. 2g–i. The 473 nm laser is used to excite the YOYO-1 in solution to measure the emission of the residual YOYO-1 (blue curve), and the 640 nm laser is used to excite the activated CellROX Red emission (red curve). The gap in the spectra is due to scattering of the excitation light and was removed for clarity. The raw spectra are shown in Fig S2. The fluorescence emission of CellROX Red is negligible when there is no YOYO-1 (Fig. 2g) and significant when there is YOYO-1 (Fig. 2h, i) as expected from the UV/Vis absorption data. The fluorescence emission of YOYO-1 is ~1000× greater with DNA present (Fig. 2i) than without (Fig. 2h) which agrees well with the literature.^29–31

The shifts of the peak at ~350 nm in the YOYO-1 absorption spectra in the solution of YOYO-1 and CellROX Red suggest complex formation between YOYO-1 and CellROX Red (Fig. 2b). Additional evidence of this is seen in the peak for YOYO-1. Usually, YOYO-1 in water shows a peak at ~460 nm and a shoulder at ~485 nm due to dye aggregation^31,32 but after adding CellROX Red the whole peak is mostly flat indicating complex formation between YOYO-1 and CellROX. Complex formation between these two dyes is likely due to hydrophobic interactions considering both YOYO-1 and CellROX Red are positively charged. The positive charge is shown by much lower adsorption of CellROX Red onto an amine modified glass surface (Fig. S3a) than a bare glass surface (Fig. S3b). Bare glass is negatively charged while amine modified glass is positively charged under normal conditions. CellROX Red adsorbs onto the hydrophobic and relatively neutral PMMA surface suggesting that the dye is hydrophobic (Fig. S3c). There is no peak shift for the solution containing CellROX Red and YOYODNA indicating that there is no complex formation between these two dyes (Fig. 2c). YOYO-1 intercalated in DNA also shows the expected peak at ~490 nm and shoulder at ~465 nm.^31,32 Under this condition, YOYO-1 is protected from the solvent and other dissolved molecules when intercalated in DNA thus no complex is expected.

Photocleavage of dye-labeled single-DNA.

Before monitoring the production of ROS with CellROX on the surface it was necessary to ensure two things: (1) the YOYO-1 or POPO-1 labeled DNA cleaves under our experimental conditions and (2) find a surface to stretch the DNA where photocleavage does not desorb DNA from the surface that interferes with the data analysis. We examined amine and PMMA surfaces. When a DNA is flow-stretched and immobilized on a substrate it is fixed like a spring on the surface. If photocleavage occurs, we expect to see gaps of DNA breaks on the PMMA surface due to the relatively weaker interaction between the DNA and the surface (Fig. 3). We start recording a video with the laser turned off and once turned on we see a uniform signal across the DNA strand in frame 1. Then at a later point in time, such as frame n in Fig. 3 cartoons, the DNA has been photocleaved. On the PMMA surface, when one point on the DNA string is broken, the ends will retract giving a higher signal at one or both ends similar to when the DNA is cleaved by enzymes during optical mapping,^33–35 or single DNA measurements.^15–17 This retraction can be clearly seen in the images of Fig. 4a. The breaks are also visible in the kymograph in Fig. 4a (right) where the intensity in a certain area suddenly drops. For a DNA on the amine surface, no photocleavage is visible under the same conditions (Fig. 4b). This is due to a large number of positively charged amine groups on the surface such that even if the DNA cleaves it will not retract enough to form a visible gap because every part of the DNA backbone is anchored to the amine surface.³⁶ Another DNA stain, POPO-1, also effectively photocleaves DNA under illumination which can be observed on a less-efficiently amine-functionalized surface by reducing the amine concentration. A low amine coverage surface is used to further support that photocleavage does happen on the amine surface. On a surface with a weak affinity to the DNA, it is also possible the signal drops entirely because the small fragments of the cleaved DNA diffuse away from the surface (Fig. S4). Therefore, moving forward, we used the efficiently functionalized amine surface to avoid losing DNA due to photocleavage, which would result in a false loss of signal from CellROX in the gap area.

Figure 3.

Scheme of visualizing photocleavage of a single YOYODNA. The red curves represent the fluorescence emission intensities, spatially, along the DNA molecule.

Figure 4.

Images of Dye-DNA complex (left) and the associated kymograph (right) for (a) YOYODNA on a PMMA surface, (b) YOYODNA on an amine surface, and (c) POPODNA on a less efficiently modified amine surface. The color bars represent the photocounts for YOYO-1 (a, b) or POPO-1 (c).

Single-molecule ROS sensing.

For single-molecule ROS sensing, CellROX Green was selected as the probe. CellROX Green binds to DNA better than CellROX Red that has a poor affinity to DNA and diffuses away too quickly after activation (data not shown). POPO-1 was selected in the pair whose emission is well separated from that of CellROX Green.

Control experiments containing DNA with only POPO-1 (excited by 405 nm laser) and DNA with only CellROX Green (excited by 473 nm laser) were performed (Fig. 5) using both purple and blue filter-sets described in the experimental section and Fig. S1. POPODNA excited by the 405 nm laser under the purple filter-set showed bright fluorescence as expected (Fig. 5a). When the same DNA was excited under the blue filter-set the POPODNA is no longer visible (Fig. 5b). This proves our filters successfully cut the fluorescence of POPO-1 allowing us to probe the fluorescence of only CellROX Green under these conditions. DNA with CellROX Green without POPO-1 shows no fluorescence when excited at 405 nm as expected (Fig. 5d). When excited with 473 nm light, dim fluorescence is visible due to some CellROX Green in the stock solution that has been pre-activated (Fig. 5e). For this experiment, the DNA was viewed first under the blue filter/irradiation and secondly under purple filter/irradiation. The order was reversed in Fig. S5 to ensure nothing is visible under the purple filter regardless of the order to confirm that both channels are not affected by photobleaching of either POPO-1 or CellROX Green. When both POPO-1 and CellROX Green are present the fluorescence of both dyes are visible under their appropriate filter sets and excitation lasers (Fig. 5g, h). The overlay clearly shows that the dyes are located within the same DNA strands (Fig. 5i).

Figure 5.

(a-c) Images of DNA stained with POPO-1 (dye:DNA bp ratio 1:40) or (e-f) CellROX Green or (g-i) both POPO-1 and CellROX Green under 405 nm irradiation and purple filter-set (left column) or both 405 and 473 nm irradiation and blue filter-set (middle column). The right column shows an overlay of the left and middle columns (using ImageJ).

The time-trace of the CellROX activation is shown in Fig. 6. Fig. 6a shows the total fluorescence emission signal of the CellROX Green near the POPODNA complex (blue curve) and the non-labeled DNA (red curve, control experiment), both normalized to the length of DNA in microns. The DNA was labeled with POPO-1, flow-stretched on the amine substrate, then washed with buffer before introducing CellROX solution. The emission signal from the CellROX Green on POPODNA increases exponentially and then decays over time. The decay of the emission signal comes from the photobleaching of the activated CellROX Green because the decay lifetime of 14 ± 7 s is consistent with the photobleaching lifetime of activated CellROX Green in the control experiment (15 ± 2 s in the red curve). The control experiment using DNA without POPO-1 (Fig. 6a red curve) shows no enhancement and the only decay comes from the photobleaching of a small amount of pre-activated CellROX Green that has been already chemically-activated in the stock solution. Figure 6b and 6c show example images of POPODNA and control DNA respectively. Note that the signal of activated CellROX has a high intensity at the center of the DNA and is weaker further away from the DNA longer than our imaging resolution (Fig. 6b). This image blurring is because once generated, the ROS diffuses into the bulk solution that activates the CellROX and only a portion of activated CellROX binds back to the DNA. DNA-bonded CellROX and a small portion of diffused CellROX are detected in these images.

Figure 6.

(a) Curves for the photocounts of CellROX Green normalized per micron length of DNA for POPODNA (blue) and control DNA (red). (b) Images of activated CellROX Green binding to POPODNA and (c) pre-activated CellROX Green bleaching on control DNA without POPO-1. Exposure time was 0.5 s per frame, and power densities were 5.7 W/cm² for 473 nm laser and 1.6×10⁻² W/cm² for the 405 nm laser. The color bar represents the photocounts for CellROX Green in (b) and (c).

Single-molecule CellROX localization.

Reducing the CellROX Green concentration, we were able to observe single activated CellROX molecules using a super-resolution technique reported before, mbPAINT (Fig. 7).^27,37 The image shows some overlap of the molecules to the DNA, (Fig. 7b) but more CellROX molecules are found randomly spread out on the surface which is consistent with the ensemble results shown in Fig. 5i. Obtaining a super-resolution image with CellROX more localized near DNA is challenging because, at lower CellROX concentrations, the radicals must diffuse longer to find a CellROX to activate, which is now farther away from the DNA. Assuming molecular weight of 400 g/mol for activated CellROX green and density of 1 g/mL, the diffusion constant can be estimated²⁷ using Stokes-Einstein equation to be D = 5×10⁻¹⁰ m² s⁻¹, that is, a molecule travels on average $\sqrt{2Dt}$ = 7 μm during each exposure time t = 0.05 s. These diffusing molecules are too fast to be caught by mbPAINT. The molecules observed in Fig. 7 are those immobilized either on DNA or on the surface that have been activated locally near the DNA. In Fig. 5, the average distance between CellROX molecules in the solution is 90 nm (2.5 μM). In Fig. 7, the average distance increases to 440 nm (20 nM). Thus, the average distance of activated CellROX to the DNA has significantly increased which leads to a larger spread of the immobilized activated molecules and a more blurred image in Fig. 7 than in Fig. 5. One possible approach to try in the future is to slow down the diffusion of molecules by increasing the viscosity of the solution.

Figure 7.

(a) Ensemble images of CellROX green activated by illuminating POPO-1 labeled DNA (color scale bar 900-1300 photocounts from blue to red) and (b) super-resolution image of CellROX molecules on the surface (color scale bar 0-4).

Bulk vs Single-Molecule Activation time.

An obvious difference between the bulk (Fig. 2) and single DNA experiments (Fig. 5–7) is that the activation of CellROX is on the order of minutes for the bulk reactions and seconds for the single DNA experiments. This difference is due to the larger laser power density for the latter experiments. We can find the number of photons absorbed by each dye molecule using the equation Absorbed Photons = σP/E/_ph, where σ is the absorption cross-section, P is the power density, and E_ph is the energy of a photon. In the bulk measurements, the power density is 8.1×10⁻³ W/cm² at 473 nm wavelength, the absorption cross-section for YOYO-1 at 490 nm is 1.64×10⁻¹⁶ cm²,³⁰ the absorption cross-section at 473 nm is 74% of that at 490 nm (1.2×10⁻¹⁶ cm²) (Fig. S6a), and the energy of a 473 nm photon is 4.20×10⁻¹⁹ J. We can calculate the photons absorbed by YOYO-1 in the bulk experiment to be 2.3 photons/s.

For the single DNA experiments with POPO-1, we get 43.6 photons absorbed/s, about 20 times larger with both 473 nm and 405 nm lasers ON. The absorption cross-section of POPO-1 at 434 nm is 1.53×10⁻¹⁶ cm².³⁰ The absorption cross-section at 473 nm is 2% of that at 434 nm (3.06×10⁻¹⁸ cm²), and at 405 nm is 46% (7.04×10⁻¹⁷ cm²) (Fig. S6b). The power density of 473 nm laser is 5.7 W/cm², and the 405 nm laser is 1.6×10⁻² W/cm². Putting these values in the equation, we calculate 41.3 photons/s absorbed at 473 nm and 2.3 photons/s absorbed for 405. The sum of both is 43.6 photons/s.

Conclusion

We have developed a proof-of-concept method to measure reactive oxygen species (ROS) on single-DNA molecules by immobilizing the dye-labeled DNA molecules on a surface. When the dyes are illuminated and excited, ROS radicals are generated near the DNA molecules which are sensed by fluorescent probes. We confirmed this sensing reaction in the bulk solution. In single-molecule experiments, when a dye labeled DNA is immobilized on the substrate and excited by the laser, the generated ROS diffuse into the bulk solution forming a concentration gradient near the DNA and subsequently a concentration gradient of activated probes. This gradient challenges the single-probe localization. As such, we optimized the conditions for single-molecule experiments including the selection of substrates, dye concentrations, probes (CellROX series), excitation lasers, and emission filter-sets. By carefully selecting a ROS probe from the CellROX series and a DNA staining dye in the YOYO-POPO-TOTO family, we are able to sense ROS generation from a single dye-labeled DNA molecule. Since ROS continues to be an important biomarker in many studies, measuring their concentration near a DNA molecule is potentially relevant in single DNA damaging studies.