Reduced Non-Specific Binding of Super-Resolution DNA-PAINT Markers by Shielded DNA-PAINT Labeling Protocols

The DNA-based single molecule super-resolution imaging approach, DNA-PAINT, can achieve nanometer resolution of single targets. However, the approach can suffer from significant non-specific background signals originating from non-specifically bound DNA-conjugated DNA-PAINT secondary antibodies as shown here. Using dye-modified oligonucleotides the location of DNA-PAINT secondary antibody probes can easily be observed with widefield imaging prior to beginning a super-resolution measurement. This reveals that a substantial proportion of DNA probes can accumulate, non-specifically, within the nucleus, as well as across the cytoplasm, of cells. Here, Shielded DNA-PAINT labeling is introduced, a method using partially or fully double-stranded docking strand sequences, prior to labeling, in buffers with increased ionic strength to greatly reduce non-specific interactions in the nucleus as well as the cytoplasm. This new labeling approach is evaluated against various conditions and it is shown that applying Shielded DNA-PAINT can reduce non-specific events ∼5 fold within the nucleus. This marked reduction in non-specific binding of probes during the labeling procedure is comparable to results obtained with unnatural left-handed DNA albeit at a fraction of the cost. Shielded DNA-PAINT is a straightforward adaption of current DNA-PAINT protocols and enables nanometer precision imaging of nuclear targets with low non-specific background.


Background
Super-resolu6on microscopy is a powerful technique to achieve beyond diffrac6on limited imaging of biological targets using light alone [1][2][3] .DNA-PAINT belongs to a branch of super-resolu6on imaging called single molecule localiza6on microscopy (SMLM) and has been shown to yield single nanometer localiza6on precision under appropriate condi6ons [4][5][6] .The approach circumvents limita6ons of alterna6ve SMLM approaches, namely photobleaching and the requirement for poorly controllable photoswitchable dyes and buffers.Instead, in DNA-PAINT, markers are conjugated with short oligonucleo6des called docking strands.Complementary sequences with a dye modified end, called imagers, are introduced into the imaging buffer.These imagers are free to stochas6cally, and importantly reversibly, bind to available docking site/s on the docking strand.It is the transient immobiliza6on of the imager to these sites that we detect as a bright single molecule event via a camera.The dura6on of these binding events can be fine-tuned by altering the number of matching base pairs [7][8][9] or adjus6ng the ionic strength of the buffer 10 .DNA binding kine6cs and sequence design are well understood which has led to the design of highly versa6le nanobiotechnological tools [11][12][13] .Despite these advancements there are s6ll limita6ons, including the possibility of non-specific binding interac6ons that may occur to varying levels in DNA-PAINT imaging 7 .This is a serious concern when agemp6ng to image proteins with unknown/variable/sparse distribu6ons in dis6nguishing real from spurious binding events as it may not be possible to iden6fy and remove such events post-hoc by computa6onal discrimina6on.The problem of non-specific events is greatly exacerbated and highly no6ceable when imaging targets within and around the nucleus of cells.Improved methods for super-resolu6on detec6on of nuclear factors will enable new analyses of cellular events.Several nuclear processes (e.g. the synthesis and processing of RNA) take place in defined membrane-less compartments 14,15 .The composi6on of these compartments is unclear, including the arrangement and stoichiometry of their components.Such informa6on is cri6cal for accessing mechanis6c details.Methods like DNA-PAINT are poten6ally powerful for addressing such ques6ons but have historically suffered from high nuclear background signals, limi6ng their applica6on.Recently, an approach using leb-handed DNA (L-DNA) was put forward as a poten6al solu6on to reduce these non-specific interac6ons 16 .L-DNA does not hybridize to natural right-handed DNA (R-DNA) which should reduce non-specific interac6ons resul6ng from imagers or docking strands hybridizing to endogenous DNA in the nucleus.The commercial procurement of L-DNA is significantly more expensive per nucleo6de than conven6onal R-DNA.This makes purchasing longer sequences, which are oben required for more advanced super-resolu6on approaches, research development and experimenta6on, prohibi6ve.In this study, we introduce a new labeling procedure, called Shielded DNA-PAINT, that greatly reduces the levels of non-specific interac6ons within cells when preparing samples for DNA-PAINT imaging using conven6onal and easy to procure right-handed DNA.Shielded DNA-PAINT  involves par6ally, or fully, double stranding DNA-PAINT an6bodies in higher salt containing solu6ons prior to applying them to biological samples leading to a reduc6on in the levels of non-specific uptake caused by those markers.We addi6onally present the capability to shield short 11 nt DNA-PAINT an6bodies using an excess of complementary strands that, due to their size, can be simply washed off prior to imaging.

Results
Reten'on of DNA-PAINT Markers in the Nucleus of Cells Immunohistochemistry labeling using commercially available primary and secondary an6bodies should ideally exhibit negligible non-specific interac6ons around the cytoplasm or nucleus of biological samples.Figure 1a, which shows the mitochondria located signal from an6-Tom20 an6bodies where mitochondria are arranged around the loca6on of the (here invisible) nucleus, demonstrates the expected fluorescent labeling outcome.For DNA-PAINT experiments our laboratory has generally adopted dye modified oligonucleo6des as docking strands in order to ascertain labeling by widefield imaging prior to oben 6me-consuming super-resolu6on imaging 17,18 .Working with COS-7 cells we no6ced a strong nuclear fluorescence signal when we replaced a normal commercial Cy3 secondary an6body with one of our Cy3 dye modified DNA-PAINT secondaries, Figure 1b.This nuclear background signal was also clearly visible when omilng the primary an6body, Figure 1c.By fully hybridizing the docking strand during the labeling procedure the nuclear signal arising from non-specifically bound secondary an6bodies (ABs) was greatly reduced, Figure 1d.However, this fully double-stranded configura6on precludes DNA-PAINT super-resolu6on imaging as the docking site is not accessible to imager strands.Prompted by these observa6ons we inves6gated the sources of DNA-PAINT backgrounds quan6ta6vely.

Imager-only and Marker-based DNA-PAINT Background Signals
Previous work on nuclear DNA-PAINT backgrounds focused on backgrounds resul6ng from imagers non-specifically binding in the nucleus 16 .We replicated such observa6ons by quan6ta6vely inves6ga6ng the non-specific nuclear binding of different imager sequences in fixed COS-7 cells free from any docking site containing an6body markers, Figure 2a.A nominal 0.1 nM imager concentra6on was applied to the sample and imaged under DNA-PAINT super-resolu6on imaging condi6ons, see also Supplementary Table 1.In these experiments the imager binding rate is specified as detected events per µm -2 per 1000 camera frames, with each frame las6ng 100 ms, which we call 1 edu (event density unit, i.e. 1 edu = 0.01 events µm -2 s -1 ).This takes into account that in DNA-PAINT the number of events is propor6onal to the area sampled and the dura6on of imaging as photobleaching is effec6vely eliminated (as long as "docking strand site loss" is small).Photodamage has previously been observed to result in docking strand site loss in DNA-PAINT imaging of DNAorigami 19 .In a primary/secondary an6body system, where there are many markers, the loss is small at typical excita6on powers and rela6vely short (<1 hour) acquisi6on 6me as indicated by a near constant event rate for the dura6on of an experiment.All event densi6es are given as the mean ± standard error of the mean.
The oben used P1 imager (10 nt), first used in the original introduc6on of the DNA-PAINT methodology 5 , exhibited slightly elevated nuclear event densi6es, 2.30 ± 0.13 edu in comparison to D2 imagers (10 nt), 1.58 ± 0.17 edu.The shorter D3 imager (9 nt) sequence had fewer non-specific events at 0.93 ± 0.08 edu.Recently, non-natural leb-handed DNA sequences (L-DNA) have been shown to facilitate the imaging of nuclear targets 16 .We therefore also tested a L-DNA sequence reported to behave favorably, called the LP12 imager (9 nt) 16 .As expected, this strand exhibited the lowest levels of non-specific binding within the nucleus, having an event density of 0.27 ± 0.02 edu, see Table 1 for the full list of sequences.Figure 2b shows a rendered example of localiza6on data for a D3 imager acquisi6on from a nuclear region.Sporadic and sparse event detec6ons can be seen.By contrast, applica6on of a DNA-PAINT secondary an6body conjugated with D3 docking strands exhibits a much larger number of events across an otherwise unlabeled cell, when probed with D3 imager, see Figure 2c.Experiments with a D3 docking strand conjugated secondary AB only (no primary ABs) yielded a nuclear event density of 15.3 ± 0.6 edu, Figure 2d, an ~16 fold increase in nuclear event density compared to using D3 imagers only in cells not incubated with DNA-PAINT an6bodies.These data suggest that the non-specific binding of DNA-PAINT an6bodies in the sample (compare also Figure . 1c) leads to more than an order of magnitude higher background event densi6es than the direct binding of imagers to nuclear cons6tuents.In other words, DNA-PAINT marker-associated backgrounds in prac6ce far dominate over the "imager-only" backgrounds that were previously studied 16 .We therefore inves6gated DNA-PAINT marker-associated backgrounds more systema6cally and sought methods to reduce it.
Figure 2e schema6cally highlights the different origins of specific versus non-specific event localiza6ons in DNA-PAINT super-resolu6on imaging.Conceptually, recorded events arise from three types of interac6ons, (i) imagers binding non-specifically to endogenous cons6tuents of the cell, (ii) imagers binding (as designed) to complementary docking strands on DNA-PAINT an6bodies that themselves are non-specifically bound in the cell and (iii) imagers binding to docking strands on DNA-PAINT an6bodies bound to their intended target, i.e. primary an6bodies of the targeted species.We call the background arising from non-specifically bound DNA-PAINT an6bodies the "marker-based background" versus the "imager-only background" and finally, the desired signal corresponding to the presence of the primary ABs we term the "target signal".In experiments with just imagers (no primary and secondary ABs) we measure solely imager-only background.In experiments with no primaries but using DNA-PAINT secondaries, we measure the sum of marker-based and imager-only backgrounds.In experiments with primary ABs and secondary DNA-PAINT ABs we measure the sum of all 3 signal contribu6ons, but their contribu6ons vary in different regions of the cell.Data interpreta6on is greatly simplified by one of the signals domina6ng over the others.For example, in cell regions where abundant target is concentrated (here mitochondria) the target signal dominates over the backgrounds as we show below.

Shielded DNA-PAINT Labeling
To reduce the compara6vely prominent nuclear marker-based background contribu6ons iden6fied in our experiments we modified the incuba6on step in which secondary an6bodies are applied during the labeling procedure.We hypothesized that since a fully double-stranded sequence greatly reduced the uptake of DNA-PAINT markers into the nucleus (Figure 1d) then a par6ally doublestranded (δds) system may similarly reduce marker-based background.Importantly, a δds docking strand leaves access for an imager to bind to the single-stranded (ss) unshielded por6on of the sequence, Figure 3a, making DNA-PAINT feasible.
With this mo6va6on the labeling procedure was adapted in a process we term Shielded DNA-PAINT labeling.In this technique, we pre-incubate the DNA-PAINT secondary ABs with excess 'shield' strands purposely designed to create a δds system, Figure 3b, prior to adding the shielded secondary ABs in this incuba6on solu6on to the sample.In addi6on, we inves6gated the use of higher ionic strength incuba6on solu6ons, prompted by the use of higher ionic strength to reduce non-specific backgrounds when using DNA-barcoded an6bodies 20 .As argued there, increasing ionic strength should help reduce electrosta6c interac6ons between DNA and charged cons6tuents of the cell.In addi6on, higher ionic strength will also stabilize the hybridiza6on of the shield strands with the DNA-PAINT AB docking strands to form the shielded δds docking strand configura6on.To inves6gate the combined effects of shielding and ionic strength, we evaluated both ss and δds DNA systems at three NaCl concentra6ons: 137 (as in a standard PBS solu6on), 500 and 1000 mM.In the ss configura6on (i.e.omilng the shield strands) each of the widefield images exhibited a clear nuclear signal indica6ve of non-specific uptake of the DNA-PAINT secondary an6body, Figure 3c.When the DNA-PAINT secondary was incubated using the Shielded DNA-PAINT approach, i.e. with docking strands in a δds state, the signal intensity from the nucleus in widefield images was greatly reduced already at the standard NaCl concentra6on of 137 mM.At higher salt concentra6ons (500-1000 mM) the nuclear region appeared virtually free from obvious signal with Shielded DNA-PAINT.To confirm these observa6ons quan6ta6vely in super-resolu6on data, event densi6es of DNA-PAINT localiza6on data were analyzed for each incuba6on condi6on using COS-7 cells labeled with an an6-Tom20 primary an6body and a Cy3-modified D3 docking strand conjugated secondary an6body to show Tom20 distribu6on in mitochondria.
Cells labeled in this way were probed with a D3 imager that binds in the ss domain of the D3 docking strand, see also Supplementary Figure 1.The dissocia6on rate, kOFF, was similar between ss and δds configura6ons indica6ng the presence of the 'shield' strand had minimal effect on imager binding 6mes, Supplementary Figure 2a.To quan6fy the background and signal components, event densi6es were evaluated in 3 different regions of the cell, (1) in the nuclear region, (2) in the mitochondrial target regions where Tom20 is known to be highly expressed and (3) in cytoplasmic regions distal to mitochondria to quan6fy cytoplasmic background (for further details of the analysis see Methods and Supplementary Figure 3, and for super-resolu6on examples see Supplementary Figure 4).Event densi6es in the nuclear region are expected to be dominated by marker-based background, in the target regions (mitochondria) would be dominated by target signal whereas in cytoplasmic regions away from mitochondria marker-based cytoplasmic backgrounds should dominate.
Cells labeled "conven6onally" in a PBS based an6body incuba6on solu6on containing standard 137 mM NaCl and unshielded ss D3 docking strand ABs exhibited the highest levels of non-specific signals within the nucleus with an event density of 23.4 ± 1.6 edu, Figure 3d.Applica6on of Shielded DNA-PAINT in the same buffer saw a significant decrease to 7.9 ± 0.5 edu (p << 0.001).There was no significant difference when incuba6ng ss D3 at 500 mM NaCl compared to ss at 137 mM NaCl (p = 0.483).Nuclear event densi6es were lowest in δds systems containing 500 or 1000 mM NaCl and interes6ngly also in ss at 1000 mM NaCl, 4.9 ± 0.3, 4.3 ± 0.3 and 4.7 ± 0.3 edu, respec6vely (p << 0.001 compared to δds 137 mM NaCl).A smaller ~2.6x reduc6on was observed in cytoplasmic background regions, Figure 3e, when comparing ss at 137 mM to δds 1 M NaCl event densi6es (p << 0.001).
The target event densi6es measured within mitochondria and resul6ng therefore mostly from "target signal" (compare also Fig. 2e) remained similar across the various labeling approaches at ~215 edu, Figure 3f.A general reduc6on in target event densi6es were however observed in ss systems where the NaCl concentra6on was increased from 137 mM, 215.4 ± 14.7 edu, to 1000 mM, 143.1 ± 7.2 edu (p<<0.001).The ra6o between the target signal and nuclear background event densi6es, or the "signal-to-background ra6o" (S/B), exhibited the most favorable rela6onship in δds state at 1 M NaCl, 56.5 ± 3.4, Figure 3g.This corresponds to an approximate 5.7x improvement in signal to background event ra6o compared to control ss in 137 mM NaCl measurements.A slightly lower S/B was achieved with ss D3 at 1 M NaCl, 33.1 ± 2.1 (p << 0.001) as a result of the dip in target localiza6ons under those condi6ons.An ~2.5-fold improvement was also observed with respect to the ra6o between the target signal and cytoplasmic background event densi6es, Supplementary Figure 5a.Comparison of cytoplasmic to nuclear background levels quan6fied by the ra6o between event densi6es for (mitochondria-distal) cytoplasmic and nuclear regions exhibited a general trend towards unity when using the δds state systems, Supplementary Figure 5b.
To evaluate the generality of these findings we addi6onally tested a different strand sequence designed for Shielded DNA-PAINT, the D2 docking sequence (see Supplementary Figure 1) agached to secondary ABs in ss & δds incubated states at 1 M NaCl, Supplementary Figure 6.An approximate 2.5 fold reduc6on in the non-specific nuclear event densi6es occurred when incuba6ng the DNA-PAINT D2 secondary in 1 M NaCl compared to 137 mM NaCl, but without shield strands.In the δds incubated state (i.e. with shield strands) at 1 M NaCl the reduc6on was ~4.4 fold, Supplementary Figure 6a.Similar to D3 ABs, the cytoplasmic background was reduced ~2.4xand the target signal was rela6vely unchanged between control and δds state, Supplementary Figure 6b-c.Overall, signal to background ra6os for nuclear and cytoplasmic regions saw broadly similar improvements as observed with the D3 docking strand and the ra6o between nuclear and (mitochondria distal) cytoplasmic regions got closer to unity with Shielded DNA-PAINT incuba6on, Supplementary Figure 6d-f.For a full list of measurements see Supplementary Table 2.
Our D2 and D3 sequences are longer (28-30 bp) than typical DNA-PAINT markers (9-11 bp) to enable the ability to func6onalize the marker for various applica6ons 7,9,11 that generally require longer DNA docking strands.For example, the D2 docking strand design was previously used to func6onalize markers for repeat docking domains 7 and also formed one half of the base used to create a super-resolu6on proximity sensor 11 .To enable the use of these schemes, post Shielded DNA-PAINT labeling, when the D2 or D3 strands are in δds state one can safely remove them through the process of toehold-mediated strand displacement 21 in order to further func6onalize the marker (as we have previously demonstrated, for example, with Repeat DNA-PAINT, Fig. 1c in 11 ).For completeness, we also tested an6bodies conjugated with a shorter docking strand that is 11 bp long, D3s, and compa6ble with the D3 imager.These secondary ABs also accumulated within the nucleus, Figure 4a.We adapted Shielded DNA-PAINT to the shorter docking strand, employing a fully complementary sequence to shield the DNA-PAINT marker during its incuba6on with the sample, Figure 4b, which reduced the nuclear background, Figure 4c.Due to the shorter sequence the shield strand is expected to hybridize only transiently with the shorter D3s strand (mel6ng temperature ~36 °C at 1 M NaCl) however, when applied in excess creates a system where the D3s docking strands are almost con6nually occupied.Washing then effec6vely removes the shield strand in this scenario, enabling imager D3 to access the DNA-PAINT D3s docking strands.Super-resolu6on measurements of D3s, Figure 4d, in these two configura6ons reported nuclear event densi6es of 10.8 ± 0.6 edu in ss and 2.3 ± 0.3 edu when ds, a ~5x reduc6on, a level similar to that achieved with D2 and D3 docking strands.This equates to an improved S/B rela6onship of ~5.8 6mes, Figure 4e.We did observe a moderate decrease in average photon rates from 3.9k photons/frame with the D3 docking strand versus 3.2k photons/frame using the D3s docking strand (which has docking strand and imager dyes in closer proximity) but note that varia6ons from experiment to experiment were larger than the mean differences observed, Supplementary Figure 2b.
For comparison with L-DNA based DNA-PAINT markers and imagers we also evaluated secondary ABs conjugated with LB12 docking strands (that were modified with a Cy3 dye, see Table 1) and probed these with LP12 imagers.We found that leb-handed oligonucleo6des conjugated to secondary an6bodies also suffered from non-specific uptake during the labeling period, see Supplementary Figure 7a, at a level of 4.8 edu.Consistent with our R-DNA based experiments above, this markerbased background is >10 6mes higher than the imager-only background from LP12 imagers alone (which was 0.27 edu, see Fig. 2i).Comparison with Fig. 2 shows that Shielded DNA-PAINT with 1 M NaCl in the incuba6on buffer and using the (conven6onal R-DNA based) D2 and D3 DNA-PAINT an6bodies reduced the nuclear background levels to similarly low values (<4.5 edu) that are otherwise only achievable with the much more expensive L-DNA.
The LB12 AB associated nuclear event density was reduced further to 1.5 ± 0.2 edu by increasing the incuba6on solu6on NaCl concentra6on to 1 M.This was s6ll ~5x greater than the background levels from LP12 imagers alone, see Figure 2a.Target and cytoplasmic event densi6es hinted at a slight, but sta6s6cally non-significant, reduc6on at 1 M, Supplementary Figure 7b & c.Conven6onal labeling achieved a target/nuclear ra6o of 39.5 ± 2.6, Supplementary Figure 7d, notably ~1.4x worse than shielded δds D3 AB labeling a 1 M NaCl, Figure 3g.With 1 M NaCl incuba6on the S/B of L-DNA measurements was further improved ~2.8 6mes to ~100.
We considered the possibility that the docking strand dye modifica6on could be exacerba6ng the non-specific uptake of our DNA conjugated ABs.To test this hypothesis, we conjugated a blank, not dye modified, D3 sequence to a blank secondary an6body.We visualized where the conjugate ended up by using a ter6ary labeling approach with Alexa Fluor 488, Supplementary Figure 8a.Under conven6onal labeling procedures the Alexa Fluor 488 was strongly absorbed into nuclear regions, Supplementary Figure 8b.By using the Shielded DNA-PAINT labeling method when applying the DNA-PAINT secondary, only the target was fluorescently labeled when the ter6ary an6body was added, Supplementary Figure 8c.Nuclear uptake led to event densi6es for ss incubated at 137 mM NaCl at 14.2 ± 2.2 edu, which was significantly reduced to 1.1 ± 0.2 edu (p << 0.001) for δds 1 M, Supplementary Figure 8d.This corresponds to an approximate 12-fold reduc6on in nuclear nonspecific events, a further 2x improvement over the dye modified version of the same strand, approaching levels of imager-only interac6ons (which always contribute to the measured event densi6es).Tom20 localiza6ons, Supplementary Figure 8e, were rela6vely similar between condi6ons 96.4 ± 9.0 edu for ss 137 mM NaCl compared to 104.1 ± 10.6 edu (p = 0.583) for δds 1 M NaCl.These findings are thus broadly similar to the observa6ons with Cy3 modified docking strands, omission of the Cy3 modifica6on moderately reduces backgrounds albeit at the expense of no intrinsic widefield signal prior to super-resolu6on imaging with DNA-PAINT.

Imaging of Nucleolin with Shielded DNA-PAINT
To illustrate the u6lity of the Shielded DNA-PAINT labeling approach we applied it to image nucleolin in fixed COS-7 cells, Figure 5. Labeling of this nuclear target with ss D3 in standard 137 mM NaCl demonstrates the difficulty of applying conven6onal DNA-PAINT labeling procedures to imaging nuclear targets, Figure 5ai.The widefield and super-resolu6on data, Figure 5aii, exhibit substan6al nuclear signal and notably also considerable cytoplasmic signal.DNA-PAINT markers, using the same D3 conjuga6on, but now incubated in a δds state at 1 M NaCl, Figure 5bi-ii, leave the nucleoli within the nucleus clearly visible above a lower level nucleolin signal typically present across the nucleus.Note that this lower level signal across the nucleus corresponds to nucleolin in the nucleus that is expressed outside of the nucleoli and has been observed previously with conven6onal an6body staining (see, for example.Fig. 1 in Terrier et al 22 ).A cytoplasmic background signal surrounding the cell when using conven6onal staining condi6ons (Figure 5ai,ii) is no longer discernable in widefield images, Figure 5bi, indica6ng the effec6veness of the Shielded DNA-PAINT protocol.This is mirrored in the super-resolu6on image where the nucleoli are well defined and cytoplasmic signal is, as expected, low.

Discussion
While background in the nucleus of mammalian cells due to non-specific imager binding had been observed during DNA-PAINT imaging before 16 , the much larger contribu6on from DNA-PAINT AB reten6on had not been recognized as such.Background from an6bodies exhibi6ng ss DNA had been previously no6ced in widefield images during mul6plexed imaging for many protein targets 23,24 but had not been quan6fied in DNA-PAINT imaging.It is highly relevant in prac6ce since we here show that these (marker-based) backgrounds typically are at least ten 6mes higher than imager backgrounds alone.We essen6ally resolve this problem by introducing a Shielded DNA-PAINT protocol with modified incuba6on solu6on that greatly reduces these backgrounds to a low level similar to that achievable with L-DNA probes but using much more widely available conven6onal right-handed DNA probes.Key to rou6nely visualize and easily detect this non-specific background with basic widefield fluorescence imaging was using dye modified DNA-PAINT markers which revealed substan6al accumula6on of the dye within nuclear regions of cultured cells.DNA-PAINT measurements corroborated the widefield informa6on with event densi6es about an order of magnitude greater than from imagers only.This also implies that maneuvers aimed at reducing imager concentra6on through the use of repeated docking domains 7 have ligle effect on the domina6ng marker-based backgrounds as the reduc6on in imager concentra6on in this method is compensated by the increase in the number of docking domains per secondary AB.
In general, a sequence dependence of non-specific reten6on of DNA-PAINT an6bodies seems difficult to predict.There are likely differences that depend both on strand length and proper6es of specific sequences.For example, the degree to which some of the non-specific labeling was reduced, when increasing the NaCl concentra6on of incuba6on solu6ons, differed for the sequences that we evaluated when docking strands remained single-stranded (no shielding), e.g. when comparing D2 and D3 conjugated ABs.
By contrast, Shielded DNA-PAINT labeling, the applica6on of par6ally or fully complementary oligonucleo6des to DNA-PAINT docking strands prior to being added to the sample, consistently reduced non-specific event densi6es in both nuclear and cytoplasmic regions of cells.Further, with Shielded DNA-PAINT we saw no detectable decrease in specific event densi6es at target loca6ons (as was seen with 1M NaCl and ss docking strands), overall leading to improved signal to background ra6os when used with 1 M NaCl in the incuba6on solu6on.
Experiments with docking strands lacking a dye modifica6on suggested that the presence of the Cy3 dye was not a major contributor to non-specific reten6on of the marker.Nevertheless, the improvement in target/nuclear ra6os (signal to background) was ~2x more favorable with a plain docking strand D3 versus D3 with dye modifica6on.Despite this finding, the ability to observe DNA-PAINT markers prior to super-resolu6on imaging in widefield mode is extremely useful and the dye modified oligos can also assist in the quan6fica6on of the frac6on of oligos successfully conjugated to the marker of choice 11 .
In agreement with a prior study 16 , an L-DNA imager, LP12 imager, performed the best in terms of low levels of non-specific binding within the nucleus.Consistent with ABs conjugated to conven6onal right-handed DNA, elevated nuclear event densi6es were detected following conven6onal incuba6on of the L-DNA marker with LB12 docking strands.The marker-based L-DNA background was 16x higher than the LP12 imager-only background.In absolute terms, the background of <5 edu is sufficiently low that it will not interfere with moderately expressed target protein signals (S/B is ~40 for the Tom20 target signal).
The reduc6on in background achieved with Shielded DNA-PAINT becomes clear by comparison with the LB12 DNA-PAINT AB performance as a reference.Nuclear marker-based backgrounds were reduced broadly from levels of 15-25 edu to ~4 edu, smaller than the level seen with LB12 ABs (at standard NaCl).This reduc6on corresponds to ~5.7x and ~2.5x improved target/nuclear and target/cytoplasm ra6o, respec6vely, as measures of desired target signal to backgrounds.The S/B ra6os are in a range of 30-50, in some cases >100 when related to the (rela6vely abundant) Tom20 target.The remaining AB associated background is in a range that is only 2-4 6mes larger than imager-only backgrounds.In the case of D3 without a dye the remaining background is as small as 1.1 edu and likely mostly reflects remaining imager-only backgrounds.
The reduced non-specific, nuclear accumula6on of (par6ally) double-stranded markers compared to classic single-stranded constructs could be due to a variety of factors.Primarily, we hypothesize that hydrophobic or hydrogen-bonding interac6ons may occur between the unpaired nucleo6des of ssDNA markers and either proteins or other nucleic acids in the nucleus, which would not occur, or would be reduced, in dsDNA constructs.
Coulomb interac6ons may also play a role, for instance with posi6vely charged DNA binding proteins.Regardless of the source of interac6on, it is likely that the very different polymeric proper6es of single-stranded and double-stranded DNA, with the former being highly flexible (persistence length ~2 nm) and the lager much more rigid (persistence length ~50 nm), modulates non-specific interac6ons 25,26 .One may indeed envisage flexible ssDNA being able to fold into configura6ons maximizing affinity with non-specific targets, which may be inaccessible to dsDNA.
Evidence that increasing ionic strength reduces nuclear accumula6on in single-stranded markers supports the hypothesis that Coulomb interac6ons may be involved, which become screened at high salt concentra6on.It is however also possible that a higher ionic strength hinders hydrophobic, or hydrogen bonding interac6ons mediated by the nucleobases by stabilizing secondary structures within the single-stranded docking strand 27 or interac6ons between the docking strand and hydrophobic or hydrogen-bonding residues on the secondary an6body.
It is possible that binding of markers non-specifically in the nucleus (or the cytoplasm) may be increased when mul6ple ssDNA strands are present on a single marker (as may occur with randomlabeling) and that this involves coopera6ve binding in some way.In this scenario the use of sitespecific docking strand agachment, as for example regularly employed with nanobodies 28 , may reduce marker-based backgrounds to the level of imager-based backgrounds.Accordingly, such marker systems should be inves6gated with the methods presented here.
Alterna6ve blocking strategies could be considered, such as the addi6on of charged polymer dextrans 20 or salmon sperm DNA 23 .Exploratory experiments with a charged polymer dextran, >500 kD dextran sulfate, using concentra6ons sufficient to reduce non-specific nuclear labeling, exhibited a concomitant reduc6on in target labeling so that S/B ra6os were substan6ally smaller than achieved with Shielded DNA-PAINT (Supplementary Figure 9).We therefore did not pursue these approaches further given the compara6ve simplicity and effec6veness of Shielded DNA-PAINT protocols presented here.
For highest signal to background ra6os L-DNA markers with addi6onal charge screening by incuba6ng the DNA-PAINT ABs in 1 M NaCl are s6ll a possible choice.In prac6ce Shielded DNA-PAINT with conven6onal DNA docking strands should be suitable for observing all but the most sparsely expressed target proteins within the nucleus.

Conclusion
We have shown that during the labeling procedure with DNA-PAINT ABs there can be significant nonspecific reten6on of these markers when applying them in a conven6onal an6body incuba6on solu6on.By incuba6ng the markers instead in a par6ally double-stranded (δds) or fully doublestranded configura6on and increasing ionic strength of the incuba6on buffer, Shielded DNA-PAINT labeling significantly reduces non-specific uptake of DNA markers into the nucleus and cytoplasm of cells.This approach helps to reduce marker-based backgrounds (DNA-PAINT AB reten6on) to levels similar to the lesser imager-only contribu6ons.In so doing, our approach can be considered by anyone conduc6ng DNA-PAINT based experiments on biological specimens.Notably, the level of marker-based backgrounds should be evaluated rou6nely in DNA-PAINT experiments.Shielded DNA-PAINT labeling performed similarly to conven6onal labeling using L-DNA markers albeit at a frac6on of the cost.This approach is straighcorward to incorporate into exis6ng super-resolu6on labeling protocols and enables the use of conven6onal right-handed DNA.The variable length of docking strands used in this study are compa6ble with the concept of fluorogenic imager use and would further help to reduce non-specific interac6ons for Shielded DNA-PAINT labeling applica6ons 9 .

Oligonucleo'des
The oligonucleo6des used in this study were first checked with NUPACK analysis web applica6on (www.nupack.org) to determine likely folding behaviors and to check complementarity with addi6onal sequences.Docking strands (DS) were ordered from IDT whilst imagers and non-modified strands were ordered from Eurofins.Leb-handed DNA (L-DNA) were ordered from Biomers.net with the same modifica6ons as their right-handed counterparts, see Table 1 for a full list of sequences used.Docking strands had a 5' Azide modifica6on and 3' fluorophore modifica6on of Cyanine 3. Imagers had a 3' ATTO 655 modifica6on which when hybridized were detected as single molecule events.The D2 design was previously used to func6onalize markers for repeat docking domains 7 and also formed one half of the base used to create a super-resolu6on proximity sensor 11 .The D3 imager was previously introduced as P39* in another study 13 and we incorporated the corresponding complementary domain into the D3 docking strand.Table 1.Oligonucleo6de sequences and their modifica6ons (mod).

Experimental Setup
A modified inverted Nikon Eclipse Ti-E microscope (Nikon, Tokyo) was used to acquire data through a 1.49 NA APO, 60x oil immersion TIRF objec6ve lens (Nikon, Tokyo).Focus was controlled with a piezo objec6ve scanner (P-725, Physik Instrumente, Karlsruhe).Signal was collected with an Zyla 4.2 sCMOS camera (Andor, Belfast).ATTO 655 imagers were excited with a 140 mW LuxX 647 nm CW diode laser (Omikron, Rodgau).Widefield images were achieved using a LED light source (p4000 CoolLED, Andover).An auxiliary camera captured transmiged light at a non-interfering wavelength to correct for focal drib 17,30 .Briefly, a calibra6on stack of transmiged light images above and below the focal plane were acquired and ac6vely monitored during any super-resolu6on acquisi6on.With this system, a 50 nm tolerance of drib in z was ac6vely maintained.Red fluorescent beads ~200 nm in diameter, (Thermo Scien6fic, product #F8887), were diluted 1:10k in imaging buffer and added to each sample.The beads were allowed to segle for approximately 15 minutes before excess beads were removed with several light washes with imaging buffer.These fiducial markers were used to correct any lateral drib over the period of an acquisi6on.Super-resolu6on data was acquired with a camera integra6on 6me of 100 ms.The open-source Python Microscopy Environment (PyME) sobware (hgps://github.com/python-microscopy/python-microscopy)and addi6onal func6onal plug-ins (hgps://github.com/csoeller/PYME-extra),were used to both acquire and analyze data.Briefly, for super-resolu6on acquisi6ons the default parameter selngs for DNA-PAINT were used, these included a 'Point Finding Threshold' value of 2.0 and 'Background' selng of 0:0 (i.e.no sliding background average is subtracted).Single molecule events were figed using a 2D Gaussian PSF model 31 (LatGaussFitFR module).

Data Acquisi'on
Event Density Experiments Stock imagers were kept frozen at 100 µM, an intermediate working stock imager solu6on of 100 nM in imaging buffer was used to freshly dilute imagers to 0.1 nM prior to beginning the imaging experiment.Larger volumes were used to reduce pipelng error, nominally 6-10 µl of 100 nM imager were diluted into 6-10 mL total volume imaging buffer.Samples were always imaged with freshly diluted 0.1 nM D2 imager, D3 imager, LP12 imager depending on the experiment.Each acquisi6on captured at least 5 k frames, up to 25 k, at 100 ms integra6on 6me.
Non-specific Imager Binding COS-7 cells were fixed, permeabilized, and blocked as described in the 'Labeling procedure'.These unlabeled samples were then introduced to 0.1 nM of the specific imager being tested (P1 imager, D2 imager, D3 imager or LP12 imager) in imaging buffer.Five thousand frames were captured at 100 ms frame integra6on 6me for each acquisi6on.Images of non-specific binding for Fig. 2 were Gaussian rendered 32 with 20 nm pixel size.Example super-resolu6on images of Tom20 were rendered with 10 nm pixel Jigered Triangula6on.
Nucleolin Imaging with DNA-PAINT COS-7 cells were fixed, permeabilized, and blocked as described in the 'Labeling procedure'.Rabbit polyclonal an6body to nucleolin (Abcam, #ab22758) was incubated for 1H at RT and washed in PBS.Samples were then either incubated in control condi6ons using an6body incuba6on solu6on containing 137 mM NaCl or following the Shielded DNA-PAINT labeling protocol for D3 docking strand using D3 shield (allowing D3 imager access) in 1 M NaCl.Samples were washed with imaging buffer a minimum of three 6mes before introducing the ~200 nm red fiducial markers.Freshly diluted, 0.05 nM, D3 imager was introduced to the sample.Widefield images of the regions imaged were acquired prior to commencing DNA-PAINT super-resolu6on measurements using a camera integra6on 6me of 500 ms.Nucleolin super-resolu6on images were Gaussian rendered 32 with 10 nm pixel size.

Data Analysis
Event Density Measurements Localiza6on events were drib corrected in PyME by using the fiducial markers.Events from the same detec6on window, i.e. from a single binding event, were coalesced into a single localiza6on.Single molecule events were then rendered as 20 nm pixel size histograms 32 and saved as .6ffile format.
The images were imported into Fiji, ImageJ 33 using Bioformats 34 .The rectangle selec6on tool was used to draw a single region of interest over the nuclear area for each dataset.Approximately ten areas were selected for cytoplasmic regions.Experiments labelled with primary an6bodies had approximately ten regions covering Tom20 signal traced around the outlines of mitochondria using the polygon selec6on tool, see also Supplementary Figure 3.The inbuilt 'ROI Manager' tool was used to record the loca6on where each selec6on was made and saved as .zipfile for future reference.Each label was then measured and set parameters were recorded, including area and mean gray values.These measurements were stored in .csvfile format.Custom Jupyter (hgps://jupyter.org)notebooks wrigen in python code were used to create plots and further analyze the data to quan6fy event densi6es.Tom20 and cytoplasmic measurements were averaged per cell and this value was used in the respec6ve boxplots.Event densi6es are specified in units of edu (event density unit) as detected events per µm -2 per 1000 camera frames, with each frame las6ng 100 ms, i.e. 1 edu = 0.01 events µm -2 s -1 .The area and dura6on normaliza6on takes into account that in DNA-PAINT the number of events are propor6onal to the area sampled and the dura6on of imaging.See also data availability statement.

Figure 1 .
Figure 1.Various labeling approaches applied to COS-7 cells.a.A primary monoclonal an<body targe<ng mitochondrial target Tom20 and a commercial Cy3 secondary an<body labels only the target.Note that the nuclear region is clear from fluorescence.b.The commercial secondary an<body is replaced with a Cy3 dye modified DNA-PAINT oligonucleo<de conjugated to a blank secondary an<body.This results in a visibly strong non-specific nuclear signal.c.ALemp<ng to label the cells with the DNA-PAINT secondary an<body only shows the same level of non-specific uptake within the nucleus.d.A fully double-stranded DNA-PAINT secondary significantly reduces non-specific accumula<on of markers within the nucleus.Red dashed lines indicate the nuclear perimeter.Scale bar: 5 µm.Images (b-d) taken with an<bodies conjugated to docking strand D2 (seeTable1).
Figure 1.Various labeling approaches applied to COS-7 cells.a.A primary monoclonal an<body targe<ng mitochondrial target Tom20 and a commercial Cy3 secondary an<body labels only the target.Note that the nuclear region is clear from fluorescence.b.The commercial secondary an<body is replaced with a Cy3 dye modified DNA-PAINT oligonucleo<de conjugated to a blank secondary an<body.This results in a visibly strong non-specific nuclear signal.c.ALemp<ng to label the cells with the DNA-PAINT secondary an<body only shows the same level of non-specific uptake within the nucleus.d.A fully double-stranded DNA-PAINT secondary significantly reduces non-specific accumula<on of markers within the nucleus.Red dashed lines indicate the nuclear perimeter.Scale bar: 5 µm.Images (b-d) taken with an<bodies conjugated to docking strand D2 (seeTable1).

Figure 2 .
Figure 2. Pathways leading to non-specific localiza<on event detec<on of oligonucleo<des.a.A plot of the normalized mean event density (edu) for non-specific localiza<ons measured across nuclear regions of unlabeled cells for various imagers.Three right-handed oligonucleo<de imager sequences P1 imager, D2 imager and D3 imager are shown and have decreasing nuclear presence (2.30 ± 0.13, 1.58 ± 0.17, 0.93 ± 0.08 edu respec<vely).The LP12 imager, a le[-handed non-natural oligonucleo<de, has the least number of detec<ons within the nucleus, 0.27 ± 0.02 edu, (20, 20, 24 & 16 cells, n = 3).b.An example rendered image of a nuclear region in an unlabeled cell exposed to 0.1 nM D3 imager for 5 k frames.c.A COS-7 cell incubated with a D3 DNA-PAINT conjugated secondary an<body (2'AB) only, imaged and rendered as in b. d.A plot of normalized mean event localiza<on density for cells labeled non-specifically in this manner show an increase in detected events by over an order of magnitude, 15.3 ± 0.6 edu (27 cells, n = 3), when compared to corresponding D3 imager-only experiments.e. Schema<c showing possible routes of single molecule event detec<ons in biological samples.Marker-based background is caused by the non-specific aLachment of conjugated oligonucleo<de ABs during labeling.A much smaller, imageronly, background is caused by the non-specific, stochas<c transient trapping of imagers on or inside the cell.All values given as mean ± standard error of the mean.Boxplot points represent the mean value obtained per measured cell, and line indicates the median.Independent two-tailed t-tests indicate the following levels of significance: ** p ≤ 0.01 and *** p ≤ 0.001.Scale bars: 5 µm.

Figure 3 .
Figure 3.Comparison between conven<onal secondary an<body labeling and Shielded DNA-PAINT labeling using adapted incuba<on solu<ons for reduced non-specific interac<ons.a. DNA-PAINT imagers can freely hybridize to docking sites on single-stranded (ss) oligonucleo<des at target loca<ons.When the strand at the target loca<on is double-stranded (ds), imagers can no longer bind.In a par<ally double-stranded (δds) system DNA-PAINT super-resolu<on measurements are s<ll possible.b.Workflow for labeling biological samples.(1) the sample is fixed, permeabilized, blocked and labeled with a primary an<body.(2) The sample is then either labeled with a DNA-PAINT secondary in (a) a normal an<body incuba<on solu<on (PBS based) typically containing 137 mM NaCl or incubated in (b) a modified solu<on containing up to 1M NaCl with complementary 'shield' strands that form a par<ally double-stranded (δds) system aimed at reducing the non-specific binding of the secondary an<body during labeling.(3) The sample is washed in imaging buffer, fiducials are added and the sample finally imaged with appropriate imager sequences.c.Widefield images of COS-7 cells labeled for Tom20 using Cy3 dye modified D3 DNA-PAINT secondary with the various incuba<on solu<ons containing, from le[ to right: ss and δds at increasing NaCl concentra<ons (137, 500, 1000 mM).Normalized mean event densi<es calculated from localiza<on events detected for the different labeling condi<ons in d. nuclear regions & e. cytoplasmic regions show decreased levels of non-specific binding with all δds systems and with 1000 mM NaCl.Target event densi<es of Tom20, f, showed similar levels in ss and δds configura<ons.A slight reduc<on in ss state at 1 M was observed.g.The ra<o of nuclear event densi<es to target showed improved signal to background (S/B) for δds compared to ss, (28, 22, 27, 28, 27 & 24 cells, n = 3).Boxplot points represent the mean value obtained per measured cell, and line indicates the median.Independent two-tailed t-tests indicate the following levels of significance: ns p > 0.05, * p ≤ 0.05, ** p ≤ 0.01 and *** p ≤ 0.001.Scale bars: c) 2 µm, inset in d) 5 µm.

Figure 4 .
Figure 4. Shielded DNA-PAINT labeling with a short 11 nt, D3s, sequence.a.At low 137 mM NaCl, D3s binds to the nucleus.A schema<c diagram of the short oligonucleo<de sequences, b, shows a fully complementary D3s* used to occupy the D3s docking strand whilst labeling the sample.At this size the binding is transient or reversible and easily removed through washing following incuba<on.This then enables D3 imager to bind to D3s docking strands for DNA-PAINT super-resolu<on single molecule experiments.Applica<on of Shielded DNA-PAINT protocol with this shorter D3s sequence protects the marker from non-specifically binding to the nucleus, c, as observed with widefield images.d.Nuclear event densi<es were reduced ~5 fold equa<ng to a ~5 fold improvement in Tom20/nuclear signal, e, with ds 1 M compared to ss 137 mM.Red dashed lines in a & c indicate the nuclear perimeter.(26 & 30 cells for ss 137 mM and ds 1 M NaCl experiments, n = 3. Boxplot points represent the mean value obtained per measured cell, and line indicates the median.Independent two-tailed t-tests indicate the following levels of significance: *** p ≤ 0.001).Scale bars: 5 µm.

Figure 5 .
Figure 5. DNA-PAINT labeling of nucleolin staining in COS-7 cells.a. Schema<c of ss D3 incubated in 137 mM NaCl.As expected, the DNA-marker was observed in widefield images, ai, to non-specifically incorporate itself across the nucleus of the cell and across the cytoplasm.Super-resolu<on measurements, aii, of the same region match the non-specific signal seen in the widefield image.b.Schema<c of δds D3 incubated in 1 M NaCl incuba<on solu<on where the light-blue por<on of the δds indicates where the imager binds.bi.Widefield images show the nucleoli present within the nucleus of cells with much greater contrast.bii.Super-resolu<on detec<on of single molecule events display well-defined nucleoli and minimal cytoplasmic localiza<ons.Scale bars: 2 µm.