DOM: Dual Optical Mapping Combining Sequence-specific Markers and A/T Frequency-dependent Profiles

doi:10.21203/rs.3.rs-3407129/v1

Download PDF

Article

DOM: Dual Optical Mapping Combining Sequence-specific Markers and A/T Frequency-dependent Profiles

https://doi.org/10.21203/rs.3.rs-3407129/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

We introduce Dual Optical Mapping (DOM), a method combining sequence-specific barcodes and AT frequency-dependent profiles. Conventional optical mapping primarily depends on barcodes generated by sequence-specific enzymes. There are also several AT frequency-dependent profiling methods. However, these methods often face challenges, including reduced confidence due to large distances between labels, the occurrence of false positives and negatives, and issues of single-molecule heterogeneity. Addressing these limitations, we devised DOM, which integrates sparse sequence-specific labeling with dense AT-specific staining, and developed an alignment program, DOM.py. Leveraging the experimental characterization of AT-specific staining, we generated an in silico reference map through computer simulation. As a demonstration of its efficacy, we applied DOM to the E. coli genome. Impressively, 178 out of 182 single-molecule DNA images were accurately mapped to the genome, yielding a 98% alignment rate. In conclusion, this study underscores the potential of DOM to confidently map single-molecule DNA images to genomes.

Biological sciences/Biotechnology/Sequencing/DNA sequencing

Biological sciences/Biological techniques/Genetic techniques/Genetic mapping

The visualization of DNA sequences has been a powerful tool for biochemical and genomic analysis. Since the first DNA visualization by electron microscopy in 1948 ¹, numerous attempts have been made to directly observe the sequences on the genome, but direct observation remains elusive. In 1966, Inman developed DNA denaturation mapping to visualize high AT frequency regions under an electron microscope ². Since the development of fluorescence dyes in 1970s, DNA visualization moved primarily to fluorescence microscopy ³. Fluorescence in situ hybridization (FISH) has been used to visualize the position of specific sequences in the metaphase chromosome ⁴ and was further developed to fiber-FISH ⁵.

In 1993, Schwartz and colleagues invented the single-molecule DNA optical mapping system, in which elongated DNA molecules were sequence-specifically digested by restriction enzymes ⁶. This platform has been used for de novo sequence assembly and structural rearrangement analysis ⁷. In 2007, optical mapping system was advanced to a new platform, in which DNA was sequence-specifically labelled by nicking enzymes and polymerase with fluorochrome-labelled nucleotides ^{8, 9}. Alternatively, sequence-specific methyltransferases have also been employed for optical mapping ^{10, 11}. These sequence-specific labelling approaches allowed the analysis of DNA confined in nanochannels, which became the basis for the commercialized BioNano Genomics ¹². Currently, it is one of the essential long-read genome analysis tools to finish the first complete gapless telomere-to-telomere assembly of the human haploid genome ¹³, and is now approaching the completion of the human diploid genome ¹⁴. The most important advantage of optical mapping is to offer genomic information about incomparably long-read lengths. However, they utilize sequence-specific barcode-type labels, resulting in less sequence information than other sequencing platforms.

One approach to enrich more information from observed DNA image is the use of A/T-specific DNA staining, which provides more detailed and contextual sequence information than barcode-type mapping, as each base pair contributes to the intensity profile. In 2010, Reisner et al. reported DNA denaturation mapping that utilizes the partial melting of YOYO-1-stained DNA ¹⁵. An increase in temperature and formamide preferentially denatures DNA strands at AT-rich regions, releasing YOYO-1 from DNA, similar to Inman’s denaturation mapping under an electron microscope ². However, this method requires careful control of the temperature in the imaging platform, and formamide, used as a DNA denaturant, reduces the fluorescence of the dye. Alternatively, Nyberg et al. reported another approach using competitive binding of YOYO-1 and netropsin ¹⁶. Netropsin specifically binds to AT base pairs and blocks the intercalation of YOYO-1 into DNA, thus the fluorescence intensity profile of stained DNA indicates the ratio of GC base pairs. However, there is still a challenge in adjusting the ratio of netropsin and YOYO-1. Furthermore, YOYO-1-based approaches are plagued by unwanted photo-induced DNA cleavage caused by YOYO-1 ¹⁷. In 2018, we reported an approach that uses a fluorophore-linked octapyrrole as an AT-specific DNA staining agent ¹⁸. The pyrroles in polyamide form hydrogen bonds with the minor groove of AT base pairs. This dye has the advantage of using only one substance for DNA staining. By simply mixing the polyamide dye and DNA, the AT profile of the DNA can be generated. This profile mapping approach is advantageous because it can reveal the sequence information at optical resolution. However, the biggest challenge with this approach was aligning the AT-specifically stained DNA images with a large genome map, as many regions in the genome map appeared similar with only AT-specific staining patterns. If the highest and second-highest cross-correlation values did not differ significantly, it would be challenging to determine where to align with confidence. Therefore, the usage of profile mapping for the analysis of whole genome was limited due to the presence of many regions with similar cross-correlation scores.

To address this issue, we developed DOM, which combines barcode-like optical mapping and AT-specific staining ^{19, 20}. In addition, we present an alignment program, DOM.py. For barcode mapping, we used methyltransferase-based DNA labelling, a commercialized kit (DLE-1) from BioNano Genomics. The combination of the two optical mapping methods enhanced confidence in aligning DNA molecules to the reference genome. Our alignment program calculates cross-correlations using Python and compares these data with other optical mapping programs, such as Maligner ²¹ and OMBlast ²². For the E. coli K-12 MG1655 genome, we collected 182 DNA molecules larger than 100 kb, which were aligned to the reference genome with 5.9× coverage. Notably, we achieved a successful alignment of 98% of the DNA images (178 out of 182). As a result, DOM is a powerful genome analysis tool that harnesses the full potential of optical mapping, offering an unparalleled wealth of information.

Dual optical map

Figure 1a illustrates a representative DOM image of a DNA fragment, which is composed of two distinct microscopic images. The first, shown in green, is a sequence-specific labelled optical map. We utilized a commercialized kit replete with enzymes and reagents obtained from BioNano Genomics (Fig. 1b). The DLE-1 (direct labelling enzyme) is based on methyltransferase ^{10, 11}, which can prevent DNA breaks induced by processive polymerase reactions during nick-translation ⁸. Although barcode-like optical mapping has been widely used for genome analysis, it has an inherent limitation: the label-to-label distance is so large that a substantial sequence information between adjacent labels can be easily omitted. For example, the magenta arrow in Fig. 1a corresponds to 93.2 kb, containing 91 genes in the E. coli K-12 MG1655 genome (NC_000913.3) (Supplementary Fig. 1). The distance between closely labeled spots is 3.1 kb (the yellow arrow in Fig. 1a), which still includes five genes in the genomic region (Supplementary Fig. 1). Therefore, such information might be missing when using solely barcode-based optical mapping. In addition, experimental errors like false positives and negatives are also problematic, as this approach utilizes enzymes to cleave or label DNA, which can lead to the bypassing of enzymatic reactions targeting specific sequences. Moreover, technical constraints such as image noise or the bleaching of fluorescent dyes could affect the accurate interpretation of OM sequence data.

To address these limitations, DNA profile mapping technologies has been introduced ^{11, 15, 18, 23}. Profile mapping generates continuous fluorescent intensity signals, which can directly reflect DNA sequence. We have previously obtained sequence information from the large DNA of the E. coli genome employing TAMRA-conjugated pyrrole octamer ¹⁸. This approach demonstrates fluorescence intensity that correlates with the A/T base pair ratio. To leverage the improved characteristics of fluorophores, here we employed AP8, ATTO-647N conjugated pyrrole octamer (Fig. 1c) to generate an AT-profiled DNA image ^{24, 25}. The fluorescence-based AT profile offers a distinct advantage in providing more comprehensive sequence information compared to sparsely labeled sequence beacons.

Pyrrole polyamides in AP8 are known to bind weak bases (represented as W, denoting A or T) within the minor groove of the DNA double helix ²⁶. To characterize the binding of AP8 capability, we conducted a fluorescence resonance energy transfer (FRET) experiment using AP8 and fluorescein amidite (FAM)-conjugated DNA oligomers, varying the number of W bases as shown in Fig. 1d. Upon excitation at 498 nm, FAM (acting as the donor) transfers its resonance energy to ATTO-647N (functioning as the acceptor). The graph reveals that the FRET efficiency starts increasing with W4 and reaches a plateau when W is repeated six times (5’-ATATAT-3’). In comparison to the W6 FRET intensity, W4 demonstrates a 5.5-fold reduction in intensity (18.1%), while W5 shows a 2.3-fold decrease (43.9%). These results align with earlier studies: dimeric pyrrole-imidazole trimers select five base pairs ²⁷ and dimeric pyrrole-imidazole tetramers selectively bind to six base pairs ²⁸. Thus, AP8 is expected to bind six base pairs because AP8 comprises two pyrrole-tetramers with a beta-alanine linkage at the center. Subsequently, we assessed the influence of guanine (G) within the hexamer binding sites and observed a reduction in binding affinity in contrast to W6, FRET intensity for 5’-ATAGAT-3’ was reduced by 4.2-fold (24.1%). The introduction of an additional G in the hexamer, as seen in sequences like 5’-ATGTAG-3’ and 5’-ATAGAG-3’, resulted in decreased FRET intensities by 11.9- (8.4%) and 10.2-fold (9.8%), respectively. Note that the sequence 5’-ATGTGC-3’ exhibited mere binding affinity, showing a 48.9-fold (2.0%) difference in discrimination compared to W6. Taken together, these results suggest that W6 predominantly functions as the binding site for AP8, while displaying a lesser binding affinity for other sequences containing W.

AT Frequency Reference Map from the Genome Sequence

To accurately align DOM images with the genome, we first sought to develop an in silico referencing method. Given our findings for the binding characteristics of AP8, we simply calculated the frequency of W6 in the sequence by dividing it into 200 bp segments, considering the pixel resolution of our microscopic system. Figure 2a illustrates the in silico W6 frequency graph of the bacteriophage λ genome (48.5 kb) with an image-like DNA map for W6 and W1 profiles and displays the fluorescent microscope images of λ DNA molecules stained with AP8. Despite our efforts to achieve uniform full-length stretching using a recently described method ²⁵ that involves tethering, elongating, drying, and immobilizing the DNA molecules, there are still variations in fluorescent intensity profiles among the captured DNA images. For example, AP8-stained λ1 DNA exhibits a dim spot in the 6.3 kb region (indicated by the yellow arrow), which is absent in λ2. Additionally, λ5 displays a smoother fluorescent profile at the 30 kb region (marked by the cyan arrow) compared to other images. These results suggest that inherent characteristics of single-molecule experiments persist in optical mapping experiments.

Therefore, we assumed that simple binning of an in-silico map does not convey the exact binding modality of AP8 to large DNA molecules. To this end, we conducted computer simulations for AP8 binding, taking molecular characteristics into account. We hypothesized that when AP8 randomly binds to W6 DNA, it might disturb the adjacent binding due to steric hindrance. We also considered the potential binding occurrence of AP8 to other repetitive Ws, from W1 to W5, for instance, requiring consideration of numerous molecular binding capabilities. Given the assumption that AP8 putatively occupies around 10 base pairs with W6 at the center, we simulated AP8 binding events to the in silico λ DNA sequence using the data of Fig. 1d. We generated individual image-like DNA maps from the random binding events simulation (denoted as SIM01 ~ SIM10) and applied a 2D Gaussian point spread function to the map. As a result, we obtained each simulation outcome, which brings resemblance to one another yet shows subtle differences (Fig. 2b). Further, we averaged the results of ten individual simulated images to construct a reference consensus map for AP8-stained DNA molecules (SIM Consensus in Fig. 2b).

Next, we sought to investigate whether our simulation-based consensus map is more accurate than the sequence frequency-based in silico map. So, we computed Pearson cross-correlation (cc) values between the AP8 intensity profile and the AT profile map to score their resemblance. As shown in Fig. 2c, the cc values for 133 molecule images of λ DNA were calculated compared to W6 (black), W1 (red), individual simulation results (SIMs), and the SIM consensus map (green). The average cc value obtained using λ DNA image clips and W1 from the 200-bp map was 0.880 ± 0.00258, while with W6, it was 0.822 ± 0.00283. Notably, the cc value between DNA images and our SIM consensus map increased to 0.900 ± 0.00288 (Fig. 2c). When we compared individual experimental data, the cc values obtained using SIM consensus map were higher for 88% of molecules (117/133) compared to those using the W1 reference map, which was used in a previous study¹⁸. Moreover, there was an improvement for 97.7% of molecules (130/133) when compared to the values from the W6 reference map (Supplementary Fig. 2). These results suggest that we could obtain a more accurate AT frequency reference map through molecular binding simulations.

Encouraged by these results, we expanded our analysis target to unknown DNA fragments from E. coli genome. It is essential to possess an accurate sequence for the cell used in this experiment, so we conducted Illumina sequencing on our lab strain of E. coli via a sequencing company (DNA Link, Seoul, Korea). Compare to the standard sequence of MG1655 (GenBank U00096.3: 4,641,652 bp, https://www.ncbi.nlm.nih.gov/nuccore/545778205), the genome of our lab strain is 4,629,724 bp long, exhibiting a 99.4% alignment to the standard with some variations. In this study, all experimental data was aligned with the simulation map derived from our sequencing result. Complete sequence information for the lab strain can be found in the Supplementary Information (MG1655_Lab.fasta). Utilizing this sequence, we repeated the simulation procedure 10 times, averaging the results to produce an in silico consensus profile map. As illustrated in Fig. 2d, the cc value between the experimental data example and the SIM consensus map was found to be 0.741. In contrast, the cc value with W6 frequency map yielded only 0.633, and that with W1 frequency map produced 0.656. Note that these scores are considerably lower when compared to the improved SIM comparison approach utilized in this study.

Alignment DOM image to the SIM consensus map

We aimed to align dual-color DNA images with the SIM consensus map of the E. coli genome. To achieve this, we developed a Python program, DOM.py, depicted in Fig. 3. This program calculates the cc_rg value, derived from the product of two cc values: one from the red AT profile (cc_r) and the other from the green barcode-based alignment (cc_g). Figure 3a displays the aligned position on the genome with the reference map using green barcode markers, and Fig. 3b presents the red AT profiles. Because the E. coli genome is circular, the reference map includes a dark-colored region. This is done by appending the front region of the genome to the end of the reference map, preventing misalignment at the terminus (as shown by the dark color in Fig. 3a and 3b). DOM.py offers candidates based on the highest cross-correlation (cc_rg) values, and also calls Maligner ²¹ and OMBlast ²⁹, the data shown in Fig. 3a. Using green barcode labels, Maligner proposes ten candidates with corresponding scores, and OMBlast indicates a single location with the highest likelihood. In the example given, all three methods pinpointed the same genome location (4.257 Mbp with 194.8 kb).

Figure 3c presents the dual-color DNA image juxtaposed with the in silico dual-color image, and Fig. 3d delineates their associated intensity graphs. The program further provides comprehensive details at the bottom, listing information such as filename, DNA length, alignment location, cc_rg, cc_r, cc_g, stretching scale, ranking, and direction. Given that surface-immobilized DNA molecules manifest varying stretch extents, DOM.py accounts for potential under- or over-stretching, ranging between 88% and 123%. To optimize computational efficiency, especially considering the multiple calculations involved, we implemented multiprocessing for parallel computing. Additionally, we employed the binary compile option (numba.njit) when writing DOM.py. This approach substantially cut down processing time; the software can process 182 images of the E. coli genome in merely 18 minutes, a stark contrast to the over 20 hours needed without parallel computing. In summary, Fig. 3 demonstrates a 194.8 kb DNA molecule alongside its aligned map, detailing a profile comparison where the cc_rg stands at 0.265 at a 99% stretch. The red profile depicted in Fig. 3d aligns closely with the violet in silico profile derived from the SIM consensus map. It's noteworthy that, for the green barcode labels, this DNA still presents one false positive and misses five labels from the anticipated twenty (+ 1/-5/20).

In Fig. 4, our goal was to evaluate the enhanced genome alignment accuracy of DOM.py by contrasting it with existing optical mapping analysis tools like Maligner and OMBlast. Often, barcode-like markers alone are inadequate for pinpointing the exact location. As an illustration, Maligner identified the top match at 0.65 Mbp, labeled as Maligner1 in Fig. 4a. However, this alignment fails to adequately align with the AT frequency, leading to a meager cc_r value of 0.006. The reason for such a low cc_r is depicted in Fig. 4b. Contrarily, DOM.py identifies 2.3 Mbp as the best match based on cc_rg, even though Maligner ranks it as the fifth-best match (Maligner5). Figure 4c displays the molecular images along with the related graphs, elucidating why Maligner placed this molecule fifth. Despite the presence of certain false positives and negatives in Fig. 4c's barcode map, DOM.py successfully maps the DNA segment to a specific genomic region, boasting high values for both cc_r (0.493) and cc_g (0.409). On the other hand, OMBlast couldn't determine the position for this particular molecule. From our set of 182 DNA images, OMBlast could only correctly identify the alignment position for 74 molecules, a success rate of 41%.

Complete DOM of E. coli genome

Finally, we aimed to produce a complete dual optical map of the 4.6 Mbp long E. coli MG1655 genome. Figure 5a depicts 178 out of the 182 DNA image clips we acquired, all successfully aligned to the SIM consensus map. Notably, a subset of just 34 DNA image clips provided full coverage of the optical map (Supplementary Fig. 4). Out of the 178 successful alignments, 177 were based on the cc_rg values; only one DNA image was exclusively aligned using Maligner, as highlighted in Fig. 5b. One primary factor for this misalignment via cc_rg might be the aggregation of the initial 20 kb of DNA, creating a pronounced red profile, thereby yielding a diminished cc_r value of 0.166. A contrasting scenario is presented in Fig. 5c, where the green labels appeared only at eight of the 16 predicted sites (indicated by the magenta box in both Fig. 5a and 5c, 8/16 = 50.0% efficiency). Such a pattern complicates alignment through Maligner or OMBlast. However, the pronounced red profile has a high matching cc_r value of 0.747, leading to a cc_rg value of 0.331 when paired with the SIM consensus map. These two distinct cases possibly elucidate the four DNA molecules that remain unaligned. In instances where DNA is inadequately labeled as well as unproperly stretched, it becomes highly challenging to map the DNA image against the SIM consensus map.

Upon scrutinizing the alignment of the 178 DNA molecules, Maligner aligned with 131 image clips, while OMBlast matched with 74. Of the 131 images identified by Maligner, the majority (115) received the top matching score, leaving 16 ranked second or in subsequent positions, as demonstrated in Fig. 4. Across our entire dataset, we identified 218 (6.9%) false positives and 1066 (33.9%) false negatives. Our observed mean DLE-1 labeling efficiency at predicted sites was 77 ± 1.8% (refer to Supplementary Fig. 3). This efficiency resonates with past research, where methyltransferase efficiencies of 78% and 75% were respectively reported ^{30, 31}.

Identifying genes in a single DNA molecule image merely by examination is profoundly enticing. This strategy could allow rapid DNA information acquisition, bypassing the complicated and time-consuming sequencing process. Unlike contemporary sequencing technologies that read DNA sequences by destroying DNA samples, DNA imaging preserve the molecular structure. This crucial difference allows for in-depth analysis and examination without the loss of the physical sample, ensuring the integrity and availability of the DNA for further study and analysis. Such innovation could signal considerable advancements in genetic research, disease investigation, and personalized medicine. Direct observation of a single molecule could facilitate more accurate and swift genomic analysis, leading to the development of enhanced diagnostic and treatment methods. This potential unfolds new scientific horizons. Despite its potential, the actualization of such technology has remained limited thus far. This is attributed to inherent flaws in the most advanced method to date, sequence-selective labeling at the single-molecule level. Employing single-molecule DNA in this method persistently yields errors stemming from random processes, thus muddling interpretation and eroding confidence in the outcomes.

To enhance confidence in single DNA molecule image identification, we developed DOM, combining barcode-like mapping and AT-specific profiling. DOM shows exceptional alignment efficiency with a 98% success rate in mapping the E. coli genome as a model. It stands out for its unparalleled accuracy in identifying unknown DNA fragment images, potentially outpacing most existing optical mapping applications. As a significant advancement in DNA optical mapping, DOM can effectively investigate large-scale genomic variations up to the mega-base pair scale, proving invaluable for de novo assembly with incomplete reference maps. When paired with modern sequencing technology, DOM marks a new era for high-precision DNA analysis, promising enhanced insight into complex genomes from single-molecule DNA images.

Synthesis of ATTO647N-β₂-Py₄-β-Py₄-Dp (AP8)

AP8 was synthesized as previously described ^{24, 25}. Briefly, a computer-assisted Fmoc solid-phase synthesis of H₂N-β₂-Py₄-β-Py₄-Dp (H₂N-P8) was conducted from Fmoc-Py-oxime resin. Cleavage from the resin was performed with 1 mL of 3-(dimethylamino) propylamine (Dp) for 3 h at 55°C. Subsequently, H₂N-P8 (1.3 mg, 1.0 µmol) and ATTO647N NHS ester (1.0 mg, 1.2 µmol) were dissolved in 200 µL of N,N-dimethylformamide (DMF) and 0.50 µL of N,N-Diisopropylethylamine (DIEA, 2.4 µmol), followed by mixing at room temperature with protection from light. After the reaction was finished, the product was purified by high performance liquid chromatography (HPLC), followed by lyophilization of the collected fractions to afford AP8 as a blue powder.

Preparation of positively charged glass surface

The glass coverslips were immersed in piranha etching solution (70:30 v/v H₂SO₄/H₂O₂) for 3 h. Subsequently, they were washed with deionized water and sonicated for 30 min. 300 µL of N-trimethoxysilylpropyl-N,N,N-trimethylammonium chloride (50% in methanol, Alfa Aesar) was added to 250 mL deionized water. The acid-cleaned coverslips were incubated in that solution at 65°C for 16 h, followed by washing with ethyl alcohol. The positively charged coverslips were stored in ethanol.

Preparation of E. coli genomic DNA

E. coli K-12 MG1655 cells were cultured in LB medium at 37°C with agitation at 200 rpm until they reached OD600 of 0.6. Subsequently, the cells were harvested by centrifugation at 10,000×g for 10 min, washed twice with 1×PBS, and resuspended in the same buffer to OD600 of 2. The bacterial suspension was mixed with molten 2% low melting point (LMP) agarose solution to achieve a final concentration of 0.7%. This mixture was dispensed as 20 µL droplets and solidified at 4°C. E. coli cells within a gel plug were lysed by 2 mg/mL proteinase K (Enzynomics) for 2 h at 42°C. The proteinase K reaction was repeated with freshly prepared enzyme. The gel plugs were washed several times with 1×TE. To remove RNA from the gel plugs, 200 U of RNase I (Enzynomics) was added and incubated for 1 h at 37°C. Two gel plugs (40 µL) were used for subsequent experiment.

E. coli reference genome sequence

The genome sequence of E. coli strain used in this study was corrected by Illumina sequencing data, resulting in a final genome length of 4,629,724 bp. When compared to the NCBI E. coli str. K-12 substr. MG1655 sequence (NC_000913.3, 4,641,652 bp), our lab strain exhibited a 99.4% alignment with some variations. The complete sequence file (MG1655_Lab.fasta) is provided in the Supplementary Information (SI) files.

Labelling and staining of E. coli genomic DNA

The labelling of E. coli genomic DNA with DLE-1 (Bionano Genomics) was conducted following the Bionano prep direct label and stain (DLS) protocol with some modifications. Initially, E. coli DNA within an agarose plug was labelled by DLE-1 and DL-green dye in DLE-1 buffer. After the labelling reaction, enzyme digestion by proteinase K was performed and then the gel plugs were thoroughly washed using 1×TE buffer to remove residual DL-green dyes. The agarose plugs containing DNA were melted at 65°C for 15 min. The molten agarose was cooled to 42°C and incubated with 1 U of β-agarase I (NEB) for 1 h. β-agarase I was heat inactivated at 65°C for 15 min. The E. coli DNA solution was mixed with 1.7 µM AP8 in 1:1 volume ratio at RT for 30 min. Before observation, this mixture was diluted tenfold using 1×TE.

Microscopy

For single molecule observations, an inverted microscope (Olympus IX70, Tokyo, Japan) equipped with 100×Olympus UPlanSApo oil immersion objectives was used. Illumination was provided by an LED light source (SOLA SM II light engine, Lumencor, Beaverton, OR, USA). We used corresponding filter sets from Semrock (Rochester, NY, USA) to set the excitation and emission wavelengths. Fluorescence images were captured using a scientific complementary metal-oxide-semiconductor (Prime sCMOS, Teledyne Photometrics, AZ, USA) and saved in a 16-bit TIFF format using the Micro-manager software. Image processing and analysis were performed using ImageJ with Java plug-ins and python programs developed in our lab.

Affinity measurement of AP8 to the AT repeated sequence using FRET

The binding affinities between AP8 and DNA oligonucleotides containing 6-FAM (fluorescein amidites) at the 5’ terminal were determined through fluorescence resonance energy transfer (FRET) analysis. Each oligonucleotide contained AT base pairs ranging from 0 to 10-mer, and their sequences were listed in SI Table S1. For this experiment, each oligonucleotide was diluted to 400 nM using 1×TE buffer and then mixed with AP8 at concentrations of 1.6 µM at a 1:1 volume ratio. After incubating the mixtures at room temperature for 30 min, the fluorescence intensities were measured through a Hitachi F-7000 fluorometer.

Simulation map based on AP8-DNA interaction

To create an in silico simulation map reflecting actual DNA-AP8 interaction data, we conducted a computational simulation. AP8 randomly bound to specific sites on a reference sequence, creating fluorescent signal patterns. Key variables in the simulation included binding length, AT threshold, block length, and the point spread function. The AT threshold and block length denote the necessary AT ratio and size occupied by bound AP8, respectively. Bayesian optimization helped find optimal values for these parameters, with scores calculated using the Pearson cross-correlation coefficient between the AP8-stained λ DNA image and the simulation map (Supplementary Fig. 5). After determining ranges for binding length, AT threshold, and block length, we further refined the point spread function using a Gaussian function with full width at half maximum as a variable. These optimized variables, alongside DNA binding affinity determined by FRET, allowed the generation of the E. coli genome's in silico map.

Alignment python program (DOM.py)

All python codes used in this study have been included in the SI python program section and are available on GitHub. Seq2RGmap.py converts a FASTA file into two-channel in silico map, representing AT-frequency and labelling site. Simulation.py converts a FASTA file into a more accurate simulation map that reflects the AP8-DNA interaction. Pcc_Folder.py searches for matching positions with reference map at various stretches. It utilizes the Pearson cross correlation coefficient for both AT profile and sequence-specific labels. Mal_omb_tifFolder.py performs alignment using MalignerDP ²¹ and OMBlast ²², which are software tools available on GitHub. DOM.py shows the alignment result. DOM_lib.py contains a library of functions.

CODE AVAILABILITY

Codes for DOM.py is available at https://github.com/

ACKNOWLEDGEMENT

Authors (^†) contributed equally to this work. We used ChatGPT to check grammar and expression.

FUNDING

This study was supported by the Samsung Research Funding Center (SRFC-MA2101-07).

CONFLICT OF INTEREST

The authors declare no competing financial interests.

Scott, J.F. Electron micrograph studies on sodium desoxyribose nucleate. Biochimica et Biophysica Acta 2, 1-6 (1948).
Inman, R.B. A denaturation map of the lambda phage DNA molecule determined by electron microscopy. J Mol Biol 18, 464-476 (1966).
Matsumoto, S., Morikawa, K. & Yanagida, M. Light microscopic structure of DNA in solution studied by the 4',6-diamidino-2-phenylindole staining method. Journal of molecular biology 152, 501-516 (1981).
Bauman, J.G., Wiegant, J., Borst, P. & van Duijn, P. A new method for fluorescence microscopical localization of specific DNA sequences by in situ hybridization of fluorochromelabelled RNA. Experimental cell research 128, 485-490 (1980).
Heng, H.H.Q., Squire, J. & Tsui, L.C. High-Resolution Mapping of Mammalian Genes by Insitu Hybridization to Free Chromatin. P Natl Acad Sci USA 89, 9509-9513 (1992).
Schwartz, D.C. et al. Ordered restriction maps of Saccharomyces cerevisiae chromosomes constructed by optical mapping. Science 262, 110-114 (1993).
Gupta, A. et al. Single-molecule analysis reveals widespread structural variation in multiple myeloma. Proc Natl Acad Sci U S A 112, 7689-7694 (2015).
Jo, K. et al. A single-molecule barcoding system using nanoslits for DNA analysis. Proc Natl Acad Sci U S A 104, 2673-2678 (2007).
Xiao, M. et al. Rapid DNA mapping by fluorescent single molecule detection. Nucleic Acids Res 35 (2007).
Neely, R.K. et al. DNA fluorocode: A single molecule, optical map of DNA with nanometre resolution. Chemical Science 1, 453-460 (2010).
Grunwald, A. et al. Bacteriophage strain typing by rapid single molecule analysis. Nucleic Acids Res 43 (2015).
Lam, E.T. et al. Genome mapping on nanochannel arrays for structural variation analysis and sequence assembly. Nature Biotechnology 30, 771-776 (2012).
Nurk, S. et al. The complete sequence of a human genome. Science 376, 44-53 (2022).
Jarvis, E.D. et al. Semi-automated assembly of high-quality diploid human reference genomes. Nature 611, 519-531 (2022).
Reisner, W. et al. Single-molecule denaturation mapping of DNA in nanofluidic channels. Proceedings of the National Academy of Sciences of the United States of America 107, 13294-13299 (2010).
Nyberg, L.K. et al. A single-step competitive binding assay for mapping of single DNA molecules. Biochem Biophys Res Commun 417, 404-408 (2012).
Lee, S. et al. DNA binding fluorescent proteins for the direct visualization of large DNA molecules. Nucleic Acids Res 44, e6 (2016).
Lee, S. et al. TAMRA-polypyrrole for A/T sequence visualization on DNA molecules. Nucleic Acids Res 46 (2018).
Torstensson, E., Goyal, G., Johnning, A., Westerlund, F. & Ambjornsson, T. Combining dense and sparse labeling in optical DNA mapping. PLoS One 16, e0260489 (2021).
Jin, Y. et al. Twelve Colors of Streptavidin-Fluorescent Proteins (SA-FPs): A Versatile Tool to Visualize Genetic Information in Single-Molecule DNA. Anal Chem 94, 16927-16935 (2022).
Mendelowitz, L.M., Schwartz, D.C. & Pop, M. Maligner: a fast ordered restriction map aligner. Bioinformatics 32, 1016-1022 (2016).
Leung, A.K.Y. et al. OMBlast: alignment tool for optical mapping using a seed-and-extend approach. Bioinformatics 33, 311-319 (2017).
Nilsson, A.N. et al. Competitive binding-based optical DNA mapping for fast identification of bacteria - multi-ligand transfer matrix theory and experimental applications on Escherichia coli. Nucleic Acids Res 42 (2014).
Jin, Y. et al. Twelve Colors of Streptavidin–Fluorescent Proteins (SA-FPs): A Versatile Tool to Visualize Genetic Information in Single-Molecule DNA. Anal Chem 94, 16927-16935 (2022).
Bong, S. et al. AT-specific DNA visualization revisits the directionality of bacteriophage λ DNA ejection. Nucleic Acids Res 51, 5634-5646 (2023).
Urbach, A.R. & Dervan, P.B. Toward rules for 1:1 polyamide:DNA recognition. Proceedings of the National Academy of Sciences 98, 4343-4348 (2001).
Geierstanger, B.H., Mrksich, M., Dervan, P.B. & Wemmer, D.E. DESIGN OF A G-CENTER-DOT-C-SPECIFIC DNA MINOR GROOVE-BINDING PEPTIDE. Science 266, 646-650 (1994).
Mrksich, M. & Dervan, P.B. RECOGNITION IN THE MINOR-GROOVE OF DNA AT 5'-(A,T)GCGC(A,T)-3' BY A 4-RING TRIPEPTIDE DIMER - REVERSAL OF THE SPECIFICITY OF THE NATURAL PRODUCT DISTAMYCIN. Journal of the American Chemical Society 117, 3325-3332 (1995).
Leung, A.K. et al. OMBlast: alignment tool for optical mapping using a seed-and-extend approach. Bioinformatics 33, 311-319 (2017).
Deen, J. et al. A general strategy for direct, enzyme-catalyzed conjugation of functional compounds to DNA. Nucleic Acids Res 46 (2018).
Bouwens, A. et al. Identifying microbial species by single-molecule DNA optical mapping and resampling statistics. NAR Genom Bioinform 2, lqz007 (2020).

There is NO Competing Interest.

MG1655Lab.fasta.zip
E coli MG1655 sequence file
SI.docx
Supplemntary Information

Download PDF

Version 1

posted

You are reading this latest preprint version

DOM: Dual Optical Mapping Combining Sequence-specific Markers and A/T Frequency-dependent Profiles

Status:

Version 1

Abstract

Figures

Main

RESULTS

Dual optical map

AT Frequency Reference Map from the Genome Sequence

Alignment DOM image to the SIM consensus map

Complete DOM of E. coli genome

DISCUSSION

METHODS

Synthesis of ATTO647N-β₂-Py₄-β-Py₄-Dp (AP8)

Preparation of positively charged glass surface

Microscopy

Affinity measurement of AP8 to the AT repeated sequence using FRET

Simulation map based on AP8-DNA interaction

Alignment python program (DOM.py)

DECLARATIONS

REFERENCES

Additional Declarations

Supplementary Files

Status:

Version 1

DOM: Dual Optical Mapping Combining Sequence-specific Markers and A/T Frequency-dependent Profiles

Status:

Version 1

Abstract

Figures

Main

RESULTS

Dual optical map

AT Frequency Reference Map from the Genome Sequence

Alignment DOM image to the SIM consensus map

Complete DOM of E. coli genome

DISCUSSION

METHODS

Synthesis of ATTO647N-β2-Py4-β-Py4-Dp (AP8)

Preparation of positively charged glass surface

Microscopy

Affinity measurement of AP8 to the AT repeated sequence using FRET

Simulation map based on AP8-DNA interaction

Alignment python program (DOM.py)

DECLARATIONS

REFERENCES

Additional Declarations

Supplementary Files

Status:

Version 1

Synthesis of ATTO647N-β₂-Py₄-β-Py₄-Dp (AP8)