Label-free optical vibrational spectroscopy to detect the metabolic state of M. tuberculosis cells at the site of disease

Tuberculosis relapse is a barrier to shorter treatment. It is thought that lipid rich cells, phenotypically resistant to antibiotics, may play a major role. Most studies investigating relapse use sputum samples although tissue bacteria may play an important role. We developed a non-destructive, label-free technique combining wavelength modulated Raman (WMR) spectroscopy and fluorescence detection (Nile Red staining) to interrogate Mycobacterium tuberculosis cell state. This approach could differentiate single “dormant” (lipid rich, LR) and “non-dormant” (lipid poor, LP) cells with high sensitivity and specificity. We applied this to experimentally infected guinea pig lung sections and were able to distinguish both cell types showing that the LR phenotype dominates in infected tissue. Both in-vitro and ex-vivo spectra correlated well, showing for the first time that Mycobacterium tuberculosis, likely to be phenotypically resistant to antibiotics, are present in large numbers in tissue. This is an important step in understanding the pathology of relapse supporting the idea that they may be caused by M. tuberculosis cells with lipid inclusions.

each phenotype subset. With only the first 3 PCs we can form a distinct cluster as seen in Supplementary Fig. 1 b,  and LP (green curves) cells are calculated from ~60 WMR spectra taken from single bacteria. The colour-shaded area represents the associated single standard deviation. The x-coordinate in (Fig. 1   a, c, e) corresponds to the Raman shift (in wavenumber, cm -1 ) and the y-coordinate corresponds to the differential Raman intensity in arbitrary units. In a WMR spectrum, a Raman peak is located at the zero crossing point (denoted by the dash-dotted line) while the Raman peak intensity is

Separation technique and control
An aliqot (2mL) of bacterial suspension were heat inactivated at 80°C for 20 minutes. The heat inactivation (80°C for 20 minutes) induces no significant modification in the Raman spectra 3 .
The bacterial suspension was separated using a density based separation technique (1.04 g.ml -1 D2O/H2O solution) similarly as previously described 1 . Both LR and LP fractions were stained using Nile red (Sigma-Aldrich) and viewed under a fluorescence microscope x100 (Leica DM5500) (excitation: 480/40 nm, 560/40 nm; emission: 527/30 nm, 645/75 nm) in order to understand the quality of the separation, the purity had to be greater than 90% over a 100 cell count to be validated.

LR and LP Sample preparation for Raman spectroscopy analysis
Separated cells fractions were re-suspended with 100 μl of PBS and 20 μl were placed on a thin quartz slide (SPI Supplies, 01015T-AB) left to air dry at 4 degrees and mounted on a thick quartz slide (SPI Supplies, 01016-AB) with a spacer filed with 15 μl of PBS. The bacteria were captured between the two slides in PBS. The mount was sealed with nail polish. This preparation was interrogated using Raman spectroscopy Section 2. Raman data analysis using different Raman shift window show equivalent output All data sets for M. tuberculosis in-vitro or in tissue are analysed with different Raman shift windows. The main differences between the two phenotypes are present in the two lipid bands A and B, which can be indicated by the increments in the variances from first 7 PCs in Supplementary Tab1 to 3. Therefore we still have good discrimination ability when only those two bands are chosen for analyses. The maximum standard deviations in sensitivity and specificity obtained using different Raman shift windows (in the region of 700 -1800 cm -1 ) is less than 0.044 in all our results for different samples as shown in Supplementary Tab. 1 to 3.

Section 3. Identification of M.tuberculosis LR and LP cells using a peak-topeak ratio (R Band A/Ref band)
A ratio R Band A/Ref band between the lipid Band A and the internal reference peak (1050-1070 cm -1 ) was also investigated. Both LR and LP (from in-vitro culture) groups can be clearly identified (Supplementary Fig. 2

Section 5. Immunostaining
Frozen section of mice lymph node infected with BCG were attached on a positively charged glass slide or a positively charged quartz slide and then interrogated by WMR spectroscopy.
Optimal Raman spectra can be achieved using quartz slides. A charged quartz slide was confirmed that it would not affect the Raman spectra (data not shown).
After WMR spectroscopy, the sample slide was used to perform immuno-staining, targeting T- Immuno-staining has been successfully performed on tissue sample slides that have been interrogated by WMR spectroscopy without any difficulties. In contrast, the CY5 and Hoechst staining were poor in quality making very difficult to identify the T-cells once the tissue has been previously stained using Nile red. Therefore, Nile red staining procedure makes the immuno process much less efficient or even impossible.

Supplementary Figure 3 Comparison of immuno-staining procedure on tissue that have been previously interrogated by WMR spectroscopy (a, c) or Nile red staining (b, d). The first row shows the overlay pictures of CY5 (showing T-cells, red color) and Hoechst counter staining (showing the nuclei, blue color). The second row shows the Hoechst staining in gray scale. Insets at the corner of (a, b, c, d) show the enlarged image in the corresponding smaller rectangle.
Based on the presented results we conclude that WMR spectroscopy and immuno-staining can be performed on the same sample and on the same slide though the slides need to be positively charged in advance. This charging process is fast, simple and costless by using Silane (Sigma) while it has no impact on the Raman spectra. Therefore, we confirm that using this label-free It is also clear that the immuno-staining procedure was strongly affected if the tissue sections were stained previously by Nile red. The CY5 emission was not specific showing lot of background and poor quality counter stain (Hoechst).

• Protocol of positively charging quartz slides
First the quartz slides were washed with distilled water and then with 70% ethanol. The slides were left to air-dry protected from dust for few minutes. In order to coat the slide, a 2% Silane solution was prepared using use 15 ml of acetone and 0.3 ml of 3-Aminopropyltriethoxysilane (Sigma A-3648). The quartz slides were dipped in Silane solution for 30 seconds. The slides were washed using distilled water for a few seconds and then left to air-dry at 37°C for several hours, protected from any potential dust.

• Frozen section of Mice lymph node tissue
Fresh BCG infected mice Lymph node tissue was stored at -80°C in the freezer. The preparing procedure was conducted on dry ice. First a sample mould was partially filled with OCT (Thermo Scientific) and the tissue was placed onto the mould before the OCT froze. Then another layer of OCT was added and left to freeze. A cryostat chuck was prepared by partially flooding the chuck with OCT and allowing it to freeze. The sample block was then removed from the mould and placed onto the chuck, over the frozen OCT layer. More OCT was added until the sample block was completely encased in frozen OCT. The chuck was placed on the cryostat (-20°C) and 10 micron thick sections were cut serially until the tissue was visible. The tissue sections were attached on positively charged slides (glass or quartz).

• Attach tissue section to positively charged slides
Three ways were investigated to attach BCG infected mice Lymph node frozen section on positively charged glass or quartz slides. The montage was then sealed by transparent nail polish at the corners.

• Nile red Staining of frozen BCG infected lymph node section
The 10-μm thick BCG infected mice lymph node frozen section were stained by Nile red (10μl of 25 μg.ml -1 solution). The stained tissue was then washed with MiliQ water and left to air-dry. The tissue sections were covered by a coverslip and observed in Leica fluorescent microscope.