ABSTRACT
There is increasing interest in understanding how the three-dimensional organization of the genome is regulated. Different strategies have been employed to identify chromatin interactions genome wide. However, due to the current limitations in resolving genomic contacts, visualization and validation of these genomic loci with sub-kilobase resolution remain the bottleneck for many years. Here, we describe Tn5 transposase-based Fluorescence in situ Hybridization (Tn5-FISH), a Polymerase Chain Reaction (PCR)-based, cost-effective imaging method, which achieved the co-localization of genomic loci with sub-kilobase resolution, to fine dissect genome architecture at sub-kilobase resolution and to verify chromatin interactions detected by Chromatin Configuration Capture (3C)-derivative methods. Especially, Tn5-FISH is very useful to verify short-range chromatin interactions inside of contact domain and Topologically Associated Domain (TAD). It also offers one powerful molecular diagnosis tool for clinical detection of cytogenetic changes in cancers.
Introduction
Chromatin interactions play essential roles in three-dimensional organization of the eukaryotic genome, a process critical to vital cellular functions such as transcriptional regulation1,2. Several strategies were employed to study genome 3D structure3–5: One is genome mapping techniques, including ligation-based and ligation-free methods; another is in situ imaging of DNA, RNA, and protein in the nucleus; and the last one is computational methods to predict either chromatin interactions or 3D chromatin organization6. Among these strategies, 3C-derivative methods rely on digestion and proximity ligation to capture pair-wise5,7 or multiple chromatin contacts8, while ligation-free methods, such as Genome Architecture Mapping (GAM)9, Split-Pool Recognition of Interactions by Tag Extension (SPRITE)10, and ChIA-Drop11 can detect multiple chromatin interactions genome-wide, by demonstrating that multiple enhancers and highly transcribed regions are associated simultaneously.
As Complementary approaches to these molecular mapping methods, fluorescence in situ hybridization (FISH) utilizes plasmids or PCR fragments that contain certain region of genome, to amplify and label fluorescence-tagged probes for hybridization. Multiplexed super-resolution FISH12,13 and different orthogonal CRISPR–dCas9 systems14,15 can be employed to label DNA at large scale and to identify cooperative higher-order chromatin interactions. With super-resolution16–21 and even electron microscopy with higher resolution22, more and more detailed genomic features about regulatory architecture such as chromatin loops, TADs and contact domains could be investigated.
On the other hand, chromatin interactions, either identified by 3C-based technologies or predicted by new designed algorithms, require rapid experimental verification by either 3C or imaging methods23. However, due to the current limitations of traditional FISH in resolving genomic contacts, validation of these genomic contacts remains the bottleneck for investigating the regulation of genome architecture. For example, many chromatin loops are inside of a TAD or contact domain2 (0.2-1 Mb), the basic unit of genome 3D structure and gene regulation, but typical BAC clones expand 100-300 Kb24, which is even comparable with the length of a TAD. Therefore, the genomic resolution of traditional FISH based on BAC clones is too low to precisely label and to image the interactions within the genomic distance less than 500 Kb, such as inside of a TAD or contact domain, whereas most of the chromatin interactions identified by 3C-based techniques fall into this range25–27.
There are advanced FISH techniques with higher genomic resolution, such as oligopaint28, HD-FISH29, CasFISH30 and MB-FISH31, etc. But they are either very expensive28,31 or require complex preparation29,30, which is difficult for laboratories that do not routinely use various FISH techniques to adopt and to study genomic loci. Molecular beacon-based FISH31 can offer a resolution of 2.5 kb, but is also technically complex and lacks cost-effectiveness.
Here we report a novel and cost-effective method named Tn5-FISH (Tn5 transposase based fluorescence in situ hybridization) that offers more than one order of magnitude higher resolution than traditional FISH. Tn5-FISH utilized the hyperactive Tn5 transposase for probe library construction, and PCR for probe library amplification and labeling. Foreign DNA sequences can be inserted into host genome with the help of Tn5 transposase through “cut and paste” mechanism32, which granted Tn5 transposase the ability to induce the DNA double breaks at insertion site. Efficient segmentation of targeted DNA sequence is favored in probe library construction33,34. With the aid of bioinformatic tools, we can obtain the repeat-free DNA sequence suitable for probing via genomic PCR.
Then Tn5-FISH was employed to verify the chromatin interactions, for example, the ones inside of type II KRT locus in chromosome 12 of K562 cells measured by Hi-C contact map from published data. Triple-color Tn5-FISH can verify these interactions well, with triple-color traditional BAC FISH as positive control, indicating that Tn5-FISH is suitable to visualize specific genomic loci and chromatin interactions inside a typical TAD, in either normal cell types, cancer cell line, or even tissues. Furthermore, Tn5-FISH potentially becomes one of the best clinical tools for the rapid molecular diagnosis, to detect or to confirm classic cytogenetic changes in cancers.
Results
The Scheme of Tn5-FISH to label target genomic loci
Foreign DNA sequences can be inserted into host genome with the help of Tn5 transposase through “cut and paste” mechanism32, offering Tn5 transposase the ability to induce DNA double breaks at insertion site. Therefore, Tn5-FISH utilizes the hyperactive Tn5 Transposase for probe library construction, and PCR for probe library amplification and labeling (see the schematic diagram in Fig. 1. For more details, see Supplementary Note 1). Efficient segmentation of targeted DNA sequence is favored in probe library construction33,34. With the aid of UCSC Genome Browser, one could easily obtain the repeat-free DNA target sequence suitable for probing via genomic PCR.
We next tested the specificity of Tn5-FISH by using two probes targeting two adjacent loci GM19705 and Platr22 (only 6.5 kb away from each other, Supplementary Fig. S1a), in wild-type mouse Embryonic Stem Cells (WT mESCs) and Platr22 knock-out mESCs (Platr22-KO mESCs, Supplementary Fig. S1b), respectively. As proposed, Platr22 loci were only visible in WT mESCs, but not in Platr22-KO mESCs, while GM19705 loci can be visualized in either WT or Platr22-KO mESCs (Fig. 2a-d). These results were simultaneously verified by traditional FISH with BAC clone (BMQ-36B16) which covered both GM19705 and Platr22 loci, indicating that Tn5-FISH is a versatile technique for labeling unique genomic loci with high specificity. A full width at half maximum (FWHM) of each fluorescent signal density, which approximately follow a Gaussian distribution, is 230 ± 20 nm and 300 ± 40 nm (mean ± SD) for the TAMRA and Alexa Fluor 488, respectively (Fig. 2e), suggesting a significant improvement of spatial resolution from ~300 nm for BAC FISH to ~230 nm for Tn5-FISH (Fig. 2f), benefited from the smaller size of probes (~4Kb).
Resolution of Tn5-FISH
Next we tried to test the genomic resolution limit of Tn5-FISH. With traditional FISH as reference, we gradually shortened the target genome locus fragment sizes from ~8kb downwards (7.99kb, 4.14 kb, 2.58Kb, 1.17Kb, 0.8Kb, 0.5Kb, and 0.3Kb, respectively, as shown in Fig.3 a-g). These seven different fragments were imaged in K562 cells with Structured Illumination Microscopy (SIM) (Fig.3 a-g).
Tn5-FISH signals were then quantified and compared with traditional BAC FISH respectively (Fig. 3h). Even with a target as short as around one kilobase, good fluorescent signal can still be observed (with 86% hybridization efficiency compared with traditional BAC FISH, Supplementary Table S1). Therefore, Tn5-FISH can offer a genomic resolution of sub kilobase, which is more than one order of magnitude higher than traditional FISH methods.
Furthermore, we compared the cost-effectiveness of some different FISH methods published so far28–31,35. With the cost of only ~50$ and only 1 day per replicate in one experiment (Table 1), Tn5-FISH is the most cost-effective one among various FISH methods, according to our knowledge.
Imaging the chromatin interactions with Tn5-FISH in a TAD of cancer cells
Sub-kilobase-resolution Tn5-FISH offers great potential to be one powerful tool in clinical molecular diagnosis. Therefore, we asked whether Tn5-FISH can be used for the fine dissect genome architecture in cancer cells. Different combination of keratin proteins encoded by KRT locus (keratin-encoding gene locus in one TAD), can be assembled into keratin intermediate filaments as structure scaffolds36. Mutations in KRT genes may cause human disease such as epidermolysis bullosa simplex and pachyonychia congenita. Chromatin interactions inside of type II KRT locus in chromosome 12 of K562 cells (Fig. 4a) were measured by Hi-C contact map (Fig. 4b) from the published data37 which indicates that there are multiple chromatin loops within the TAD. The chromatin loops were verified by triple-color Tn5-FISH (Fig. 4c and Supplementary Figure S2), with triple-color traditional BAC FISH as a positive control (Supplementary Figure S3).
The chromatin loops were verified by triple-color Tn5-FISH (Fig. 4c), with triple-color traditional BAC FISH as a positive control (data not shown). Three interacting sites were closely adjacent to each other spatially (no matter labeled by BAC FISH or by Tn5-FISH), while the negative control was ~500 nm away from the interacting sites (Fig. 4d and e). The independent verification by both different FISH methods further indicates that Tn5-FISH is a robust method to investigate chromatin interactions inside of a given TAD in 3D genome structure.
Discussion
As very different strategies, both genome mapping and imaging provide very different, yet complementary in many ways, information about the chromatin looping and genome architecture. Many chromatin loops fall into the range of one BAC, thus can’t be confirmed by traditional BAC FISH. Tn5-FISH, a cost- and time-effective imaging method, expanded access to validate or reconcile the contact frequency of genomic loci in Hi-C contact map, in either normal cell types or cancer cell lines, thus is a good choice for the imaging and verification.
Improving the specificity and reducing the uncertainty
Tn5 Transposase could generate short segments around 40nt, reaching a theoretical maximum labeling density of 50 probes per kb, therefore helps the signal to be distinguished from background. There are three reasons to use Tn5 for probes preparation in mammalian FISH. First, the hyperactive Tn5 provides the flexibility of DNA probe template (BAC or PCR fragments), in small quantities (1-50ng, can be varied according to commercially available Tn5 library construction kits), which is 20 times lesser than the amount BAC clone required. Second, the full length of probes in constructed library is normally within 100-200 bp, which is appropriate for FISH. Third, the DNA sequence from fully functional Tn5-DNA complex is orthogonal to mammalian genome38, therefore the specificity of probes was guaranteed.
The uncertainties of FISH experiment might arise from probe size, chromatin displacement during denaturation and hybridization, background noise and ambiguity caused by homologous chromosomes, etc26. Tn5-FISH method, which uses the fragments of several kilobases to generate probes for labeling, will greatly reduce the uncertainty caused by probe size. However, the uncertainty caused by the hybridization process still exist due to the denatured hybridization conditions of Tn5-FISH similar to that of traditional FISH. On the other hand, for one specific site, when probe size decreases, the mis-target rate increases (Table 1). In this case, the quality of Tn5-FISH images can be complemented with BAC FISH as positive control or backup.
What could be the highest genomic resolution for Tn5-FISH theoretically, and how to image it?
With the following assumptions: (1) Nucleotide sequences of a human genome are randomly distributed; (2) The designed oligonucleotide was hybridized very efficiently to the target sequence without or with little hybridization to other similar sequences presented in the genome; (3) The single fluorescent molecule labelled on this designed oligonucleotide can be imaged very efficiently and successfully; we may conclude that theoretically, the highest genomic resolution of FISH can reach 17 nucleotides, as an ideal 17-base-long oligonucleotide should find its specific and the only position on the genome, and 417 nucleotides can already cover the whole 3-billion-base-pair human genome.
However, the super-resolution imaging of the single-fluorophore labelled on the 17-base oligonucleotide can be very challenging, if not impossible. The two-dimensional localization precision of an individual fluorescent molecule is defined by the sum of photon noise, pixelation noise and background noise, as shown by39,40
Here S is the standard deviation of the PSF of a given microscope system, N is the number of photons collected, a is pixel size, and b is background noise.
In order to obtain high-quality imaging, one needs to optimize the microscopic system to increased s, employ the best-performing organic fluorophore for high photon number N, and adopt appropriate strategy to deliver fluorophores into cell nuclei efficiently, to reduce the background noise b. Furthermore, one needs to consider how to tradeoff these optimized parameters. For example, during sample preparation, what labeling density and washing conditions should be employed; during imaging, what wavelength and detector parameters should be considered; whether TIRF illumination should be considered to reduce the background noise, etc. Photobleaching curve41 could be employed, to decide whether to increase the photon number N, or to reduce the background noise b in formular (1). According to our experience, for the single-color FISH with one labelled typical fluorophore molecule, the localization precision of less than 10nm can be obtained, with the photon number N larger than 104.
In summary, Tn5-FISH can be a powerful tool for broad image-based chromatin structure studies with joint consideration of contact frequency and high-throughput bar-coding strategy. Furthermore, with much higher resolution, Tn5-FISH will potentially become one of the best clinical methods for the rapid (prenatal) diagnosis and prognosis42,43, to detect or confirm classic cytogenetic changes, such as chromosomal rearrangement, aberrations/abnormalities, tiny chromosomal changes in patients with Edward syndrome, Williams syndrome, hematologic malignancies, or certain solid tumors.
MATERIALS AND METHODS
Cell culture
K562 and GM12878 cells were purchased from China Infrastructure of Cell Line Resources (Peking Union Medical College, Beijing, China) and maintained in RPMI1640 medium (Gibco, U.S.A.) supplemented with 10% fetal bovine serum (FBS, Gibco, U.S.A.), 50 units/ml penicillin and streptomycin (Gibco, U.S.A.), and Non-Essential Amino Acids (NEAA, Gibco, U.S.A.) as instructed.
Tn5 FISH probe preparation
Tn5-FISH probes were constructed as follows. First, sequences of interacting sites were downloaded from UCSC genome browser with repeats masked as N. DNA fragments used for probe library generation were amplified by PCR and recovered by DNA Cleanup kit (D4014, Zymo research, U.S.A.). The amplified probe library could serve as templates for Tn5-FISH probes. The Tn5-FISH probes were obtained by a second PCR amplification with fluorescence-tagged primers.
Traditional FISH probes generated by Nick Translation of BAC clone
Traditional FISH probes were prepared as previously described35. Briefly, BAC clones were ordered from Thermo Fisher, and obtained by BACMAX kit (Epicenter, U.S.A.). For 1 μg BAC clone, 1 μL DNA Pol I and 20 U DNase I (NEB, U.K.), 1 mM each of dATP, dCTP, dGTP and fluorescence-tagged dUTP/dTTP were added together and incubated at 15°C for 2 h, and finally re-dissolved in DNA FISH buffer.
Multi-color Tn5-FISH preparation
The procedure of Tn5-FISH was the same as traditional FISH as previously described35. Briefly, cells were fixed by 4% paraformaldehyde and permeated by 0.1% Triton-X 100, then incubated at 20% glycerol for 30 min. As for two-color Tn5-FISH, each color of 10 ng Tn5-FISH probe was mixed with DNA FISH buffer and applied to cells. The FISH program was set to 75°C for 5 min, then 37°C overnight. The slides were then imaged either by Carl Zuess LSM780 confocal microscope or Nikon A1 SIM microscope.
Image processing and Quantification of FISH signals
The obtained Tn5-FISH images were processed by FIJI software (version 1.25h, from NIH) and the videos were processed and all images were quantified by Imaris 9.2.0 (Bitplane, Switzerland).
Statistical analysis
The statistical analyses are carried out using build-in two-tailed t-test models provided by Graphpad Prism7 software (Graphpad Software, San Diego, U.S.A.). The significance indicators are labeled according to p-values analyzed from the data in this manuscript.
The statement for the conflict of interest
The authors certify that they have NO affiliations with or involvement in any organization or entity with any financial interest, or non-financial interest in the subject matter or materials discussed in this manuscript.
Acknowledgement
Due to space limitations, we apologize that only a very selected list of FISH methods could be cited here. We thank Prof. Minping Qian and members from Prof. Michael Q Zhang’s lab for useful discussions. This work was supported in part by the National Natural Science Foundation of China (31671383 and 81890990), the State Key Research Development Program of China (2017YFA0505503 and 2016YFC1200300). This work was supported by funds from Beijing Advanced Innovation Center for Structural Biology, Tsinghua University.