Infrared nanospectroscopic mapping of a single metaphase chromosome

Abstract The integrity of the chromatin structure is essential to every process occurring within eukaryotic nuclei. However, there are no reliable tools to decipher the molecular composition of metaphase chromosomes. Here, we have applied infrared nanospectroscopy (AFM-IR) to demonstrate molecular difference between eu- and heterochromatin and generate infrared maps of single metaphase chromosomes revealing detailed information on their molecular composition, with nanometric lateral spatial resolution. AFM-IR coupled with principal component analysis has confirmed that chromosome areas containing euchromatin and heterochromatin are distinguishable based on differences in the degree of methylation. AFM-IR distribution of eu- and heterochromatin was compared to standard fluorescent staining. We demonstrate the ability of our methodology to locate spatially the presence of anticancer drug sites in metaphase chromosomes and cellular nuclei. We show that the anticancer 'rule breaker' platinum compound [Pt[N(p-HC6F4)CH2]2py2] preferentially binds to heterochromatin, forming localized discrete foci due to condensation of DNA interacting with the drug. Given the importance of DNA methylation in the development of nearly all types of cancer, there is potential for infrared nanospectroscopy to be used to detect gene expression/suppression sites in the whole genome and to become an early screening tool for malignancy.


Metaphase chromosomes isolation
Metaphase chromosomes were obtained from HeLa cells according to the following procedure.
To enrich the number of cellular nuclei containing chromosomes in the metaphase stage, the cells were incubated with 5% colcemid for 4 hours. Xs human chromosomes derived from lymphocytes were obtained according to the same protocol, however colcemid was added to the medium 2 h before preparation. After drug treatment the cells were incubated with 0.75 mM KCl for 7 min and centrifuged (1). Then the cells were fixed in a mixture of acetic acid and methanol (3:1) and deposited on a cooled substrate -ZnSe crystal for AFM-IR measurements using nanoIR system and CaF2 windows (IR grade, Crystran) for measurements using nanoIR2 system. Chromosome from healthy female donors were investigated in accordance with the Human Bioethical Committee of the Regional Medical Board in Kraków (No.

Technical details of applied instrumentation and data acquisition
Both used lasers covered broad spectral ranges: OPO coupled to nanoIR2 uses two stages: 3600 cm -1 -2234 cm -1 and 2000 cm -1 -900 cm -1 and OPO coupled to nanoIR uses also two stages 3600 cm -1 -1610 cm -1 , 1610 cm -1 -1000 cm -1 . The essential difference between those two systems lies in the geometrical arrangement of the IR illumination. In nanoIR system the sample is placed on a ZnSe prism and it is illuminated from the bottom, whereas the laser beam is reflected (total internal reflection) from the upper surface of the prism just below the sample, where an evanescent wave penetrates the sample-prism interface. Site illumination is applied in nanoIR2 system. Therefore this setup does not require usage of prisms and the chromosomes were fixed onto flat CaF2 windows for measurements with nanoIR2.
The scan regions were related to chromosome size typically several microns, the scan rates were set between 0.02 Hz -0.01 Hz.

Data post-processing
Single point spectra were smoothed using Savitzky-Golay algorithm (2 nd order of polynomial, 3 smoothing points). PCA (cross-validation) was applied to not normalized spectra using Unscrambler 9.2 software (CAMO, Norway) on the smoothed data. AFM and infrared maps were analysed and processed using the SPIP software (Image Metrology, Denmark). The images were normalized to the applied laser power and the position of the images were correlated based on the topographies collected simultaneously with each absorption map. The z-offset for each map was changed in order to set minimum value to 1. Then the ratio (a map of IR absorption at 2952 cm -1 to a map of IR absorption at 1240 cm -1 ) was calculated and nonlinear median filter was applied to 0.5 % of extreme values (the highest and lowest). The window of median filter was rectangular of size 15 pixels x 15 pixels, borders included.
PCA confirmed the chemical distinction in the banding pattern. The total scale CH3/OPO ratio is in the range from 0 a.u. to 6 a.u. for thick (190-210 nm) chromosome 2, presented in Fig. 3.
For thin chromosomes such as Xi, Xa and chromosome 3 (60 -80 nm) the total scale CH3/OPO ratio is in lower range of 0 a.u. -2 a.u. The colour/intensity threshold was optimized for each chromosome separately because the IR signal is proportional to the thickness of the chromosome at each particular wavenumber but with a different angular coefficient. Thus, the absolute value of the ratio image is different depending on the chromosome thickness (Z direction in the AFM map). The thickness of each presented chromosome is given in the figure captions. Finally, the threshold was verified/improved based on results of multivariate data analysis described in the next part of the article. PCA classified eu-and heterochromatin spectra and the threshold on each map were chosen in order to demonstrate the areas of heterochromatin spectra location (in yellow colour) and the areas of euchromatin spectra location (in blue).

S2 OPO laser power
Averaged laser power from 24 measurements and its standard deviation are presented in the supplementary Fig. S1. Each AFM-IR spectrum was corrected to account for the varying laser power, but, it should be noted that during the measurement of such a small object like a chromosome (thickness less than 100 nm) only a very small amount of light is absorbed especially in the spectral regions where the laser power was relatively low. A drop in AFM-IR laser power in the 1230 cm -1 -1190 cm -1 results in a significant difference between the spectra of chromosomes, DNA and cellular nuclei, which is addressed in Fig. S1. The laser power was measured once per day, according to supplier's instructions, and the average laser power from 24 measurements and the standard deviation are also presented in the supplementary Fig. S1. Each AFM-IR spectrum was corrected to account for the varying power but it should be noted that during the measurement on such small objects like chromosomes (thickness usually less than 100 nm) only a very small amount of light is absorbed in the spectral regions where the laser power is relatively low. Interestingly, in each of three investigated independent chromosomes two types of spectra were collected (blue and red, Fig. 1).
For the spectral acquisition the laser power value was optimized for the entire fingerprint region 1800 cm -1 -1000 cm -1 . Prior to AFM-IR mapping, the laser power value was optimized in order to increase both the signal-to-noise ratio and contrast between signal from the chromosome and background, for each independent wavenumber. Then, each acquired map was normalized to the laser power value. Therefore, the low laser power in the spectral range 1230 -1190 cm -1 influences the spectral shape (peak ratio) but it does not affect the overall mapping of the chromosome. Additionally, the signal-to-noise ratio at 1240 cm -1 was much higher in comparison to the spectral range of O-P-O symmetric stretching region 1100-1080 cm -1 , as shown in Fig. 2.

Supplementary Figure S1
Averaged relative laser power used during the collection of AFM-IR spectra.

The spatial resolution on chromosomes
The AFM-IR technique is based on the photothermal induced resonance effect (PTIR) (3). If a pulse of IR light at a fixed wavenumber is absorbed by a sample, the local rise in temperature leads to a local photo-thermal expansion. The AFM tip enables the detection of this temporary dilatation of the scanned region with a lateral resolution defined in principle by the dimensions of the AFM tip at the nanometer scale and it allows reconstructing, together with the acquisition of conventional morphology imaging, the IR-absorption map of the sample. It has been demonstrated in the case of several biological samples that the resolution of the instrument is routinely in the order of 10-20 nm, which is indeed the typical radius of a conventional AFM tip (3)(4)(5).
In order to calculate how precisely we were able to detect the edges between hetero-and euchromatin areas, we applied a knife-edge method to our measurements. This methodology was applied to the image ratio at the wavenumbers corresponding to the IR absorption of CH3 , where zoomed sections are shown, the spatial resolution is in the order of 12-25 nm. In order to estimate quantitatively and more accurately the spatial resolution, we performed a first derivative of each section, we fitted it by a Gaussian function and we measured its full width at half maximum (FWHM). This quantity is intimately related to the sharpness of the transition and to our spatial resolution, which was in the best case 12 ± 7 nm. The experimental error in determining the spatial resolution has been determined by considering the sum of the statistical error in determining the FWHM (≈1 nm), plus the sensitivity error due to the image pixelization (≈6 nm). In particular, in our specific case, it is not the tip size, but rather the pixel size in the maps that determines the ultimate spatial resolution measureable. Indeed, as stated by the Shannon-Nyquist theorem, the highest resolution that can be obtained in a raster map is twice the pixel size and it is at the best ≈12 nm in our maps (6). Finally, for the sake of rigor, we should specify that the knife-edge method is not applied to the raw data (AFM-IR signal), but to a divided/normalised image. For this reason, though we are still measuring accurately the spatial resolution, a further error in the order of 10 nm has to be considered in the experiment of the absolute position measurement in the map of the borders between methylated and un-methylated types of chromatin. Thus, the above discussion leads to an average ultimate spatial resolution in in our measurements of 15 ± 17 nm.
The intensity CH3/OPO ratio in the substrate area was approximately 5 times lower than the signal collected from the chromosome (Fig. S2). This signal is a convolution of instrumental noise and low absorption of IR light in-cellular debris, which cannot be avoided during the sample preparation procedure.

The spatial resolution on fibrillar samples
In order to illustrate the ultimate spatial resolution of AFM-IR, we report here what we demonstrated in our previous studies on the chemical characterisation of individual amyloid fibrils (7), where an AFM-IR system with top illumination and an AFM tip with a nominal diameter of 30 nm was used (PR-EX-nIR2, Anasys, USA). Spatial resolution of an optical image can be also defined as the closest distance at which two different objects can be still distinguished. In the Fig. S5 a- Table S1 AFM-IR band assignments for spectra of single chromosomes, cellular nuclei, nucleic acids and their components. To confirm achieved results by AFM-IR technique, chromosomes isolated from an anonymous patient were directly immunofluorescently stained using the Imprint® Monoclonal Anti-5methylcytosine (33D3) antibody produced in a mouse (Sigma-Aldrich, SAB4800001) and counterstained using propidium iodide (Sigma-Aldrich) to better visualize bands and distribution of methylation of DNA ( Supplementary Fig. S7b). The first step of staining was incubation in saline-sodium citrate (SSC) for 10 min in 37°C. The sample was then incubated in pepsin solution (20 μg/ml in 0.1N HCl, Sigma-Aldrich) also in 37°C. After incubation for 10 min, sample was washed using phosphate buffered saline (PBS) (Sigma-Aldrich) and dehydrated using following concentrations of ethanol: 50%, 70%, 100%. Subsequently, the sample was rehydrated in PBS and incubated with anti-5mC antibody (92 µg/ml) and propidium iodide (0.25 μg/ml) for 3h in 37°C. Photos were taken using the Metafer system (MetaSystems) with a fluorescent microscopic module (63x) and manual MMC module using ISIS software. Based on the fluorescent photos (63x) obtained in different channels for fluorescein (FITC) and propidium iodide stain, a karyogram of chromosomes was set using the same method as before ( Supplementary Fig. 7 d). Two X chromosomes were also typed based on a karyogram (Fig. 7 c,f) and the distribution of methylation based on fluorescence staining was compared with the distribution based on the AFM-IR spectroscopic data (Fig. 7b-c, e-f ). Based on the PCA analysis of the chromosome spectra in the fingerprint region, it was possible to observe chemical modifications in histone structure in eu-and heterochromatin. Intensity changes of C=O at 1723 cm -1 also influenced the clustering of the spectra. This spectral change could be related with acetylation of histones in euchromatin (24)(25)(26). PC-1 (61 % of total variance) and PC-3 (6 % of total variance) indicate that bending from α-methylene group in cytosine at 1407 cm -1 (15,20,27) is also responsible for the clustering in the scores plot (Fig. S8) indicting different sugar-base conformations (20) related to the degree of DNA packing in euand heterochromatin. PCA performed on DNA spectra in the fingerprint spectral range indicated a different sugar-base conformation associated with the spectral change observed at 1408 cm -1 (20), which difference is a significant factor affecting the clustering of spectra collected from methylated and un-methylated DNA. This provides a unique insight into chromatin structure at the molecular level that is only achievable using AFM-IR spectroscopic methods. In the fingerprint region each PC indicates that the base stacking mode at 1712 cm -1 is responsible for the clustering observed in the scores plot ( Supplementary Fig. S9a). Intensity changes in this mode are considered to be related to different levels of DNA packing (28). Therefore, this result confirms the hypothesis that AFM-IR spectroscopy can detect lightly packed DNA (typical for euchromatin) from tightly packed DNA (heterochromatin). In addition, all PCs indicate intensity changes in the methyl deformation band of cytosine at 1407 cm -1 and 1346 cm -1 (15,16,27,29), which are also related to the DNA structure and packing, and intensity changes of the band assigned to right handed helices at 1440 cm -1 (27,29).

S9 Metaphase chromosomes Fluorescence in situ hybridization (FISH) and Imaging
FISH was performed using the commercially available Red Alu probe (Chrombios GmbH). The repartition of the Alu sequences along the chromosomes corresponds to GC rich areas, and provides a R-band profile linked to euchromatin repartition (30). Briefly, 4 μL of Alu Red probe was added to a slide containing human chromosomes and a 16 mm coverslip was placed on the sample and sealed with nail polish. Denaturation of the DNA was performed at 72°C for 5 minutes. Then, the slide was incubated at 37°C overnight to promote the hybridization process.
The coverslip was carefully removed from the slide and the slide was placed into 2X SSC buffer bath for 5 min at RT. The slide was transferred into a bath of 0.4X SSC /0.1% Tween prewarmed solution at 70°C for 1 min to remove un-hybridized probes. The slide was rinsed in a bath with 2X SSC for 5 min at RT before counterstaining with Dapi (Sigma-Aldrich) at 0.5 μg/mL for 15 min. Finally, the slide was rinsed with 2X SSC and mounted using Mowiol solution.
The application of Alu-DNA as the probe, produces a pattern of hybridization signals similar to negative Giemsa banding (31,32). The Alu probe (red) was used in FISH staining, to highlight the euchromatin areas of chromosomes and fluorescent dye. DAPI (blue) was used as a counterstain but this approach did not achieve the same spatial resolution as AFM-IR, since lateral resolution of fluorescent staining is diffraction limited. Moreover, the FISH technique is reliant on the detection of cytosine-guanine pairs as opposed to AFM-IR that relies on the detection of the degree of methylation. The latter is more important in discerning the active (euchromatin) from non-active (heterochomatin) DNA. Euchromatin is characterized by trimethylation at H3K4, H3K36 and H3K79 and heterochromatin is enriched in trimethylation at H3K9, K3K27, and H4K20. (33) Imaging was performed using an inverted Zeiss Axiovert Z1 system with a CoolSnap camera and a 63x, oil immersion objective (N.A. 1.4).

S10 Eu and heterochromatin content in single metaphase chromosome
In order to estimate content of heterochromatin in measured chromosomes, using AFM-IR spectroscopy, AFM-IR maps were extracted and analyzed using ImageJ software (NIH, open source, Fig. 11 a-e). First, background of maps was cut, image color type was changed for 8- was obtained (Table S2, Fig. S11 a-e). The obtained results were compared with values given by International Human Genome Sequencing Consortium of heterochromatin and euchromatin content in human chromosomes (presented graphically in Fig. S11 f and in Table S2) (34)(35)(36).
As it could be seen in the  (Fig. S11 f).
This phenomenon can be also observed when AFM-IR nanospectroscopy is applied to identify heterochromatin distribution on human chromosomes. It manifests itself in a different scale of the images when 2952/1240 ratio is calculated (Fig. 3, 4, 5).

S11 Theoretical procedures
In the present work, calculations in liquid phase were carried out using the M062X method (37) with cc-pVDZ basis set (38), for both geometry optimizations and frequency calculations. For the Pt atom the LanL2DZ effective core potential was used. All these quantum chemical calculations have been conducted using the GAUSSIAN09 software package (39).