Complex biomolecules absorb in the mid-infrared (λ = 2–20 μm), giving vibrational spectra associated with structure and function. We used Fourier transform infrared (FTIR) microspectroscopy to “fingerprint” locations along the length of human small and large intestinal crypts. Paraffin-embedded slices of normal human gut were sectioned (10 μm thick) and mounted to facilitate infrared (IR) spectral analyses. IR spectra were collected using globar (15 μm × 15 μm aperture) FTIR microspectroscopy in reflection mode, synchrotron (≤10 μm × 10 μm aperture) FTIR microspectroscopy in transmission mode or near-field photothermal microspectroscopy. Dependent on the location of crypt interrogation, clear differences in spectral characteristics were noted. Epithelial-cell IR spectra were subjected to principal component analysis to determine whether wavenumber-absorbance relationships expressed as single points in “hyperspace” might on the basis of multivariate distance reveal biophysical differences along the length of gut crypts. Following spectroscopic analysis, plotted clusters and their loadings plots pointed toward symmetric (νs)PO2− (1,080 cm−1) vibrations as a discriminating factor for the putative stem cell region; this proved to be a more robust marker than other phenotypic markers, such as β-catenin or CD133. This pattern was subsequently confirmed by image mapping and points to a novel approach of nondestructively identifying a tissue's stem cell location. νsPO2−, probably associated with DNA conformational alterations, might facilitate a means of identifying stem cells, which may have utility in other tissues where the location of stem cells is unclear.
Disclosure of potential conflicts of interest is found at the end of this article.
Infrared (IR) spectroscopy is based on radiation absorption between 2-μm and 20-μm wavelengths (400 to 4,000 cm−1) giving rise to molecular vibrations . Biomolecules absorb the mid-IR (λ = 2–20 μm), and this allows the acquisition of vibrational spectra that contain absorbance intensity peaks corresponding to different chemical bonds (including C–H, O–H, CN, C=O, C=N, N–H, C=C, C–O, C–N, and PO2−) . Various IR microspectroscopic techniques, in particular those using Fourier transform infrared (FTIR) spectroscopy, have potential applications in diagnostics  or fingerprinting cells following toxic insult  and can provide unique information about structural and conformational changes of the molecules inside the cell. Coupled with computational tools such as principal component analyses (PCA), data reduction of the large numbers of derived variables facilitates the identification of (bio)chemical entities responsible for underlying variance.
It has always been something of a conundrum that we seem to know a great deal about intestinal stem cells and yet reliable markers for these cells have proved to be elusive. It is generally agreed that in the small intestinal crypt, the stem cells are located in a ring, at approximately cell positions 4–5 (counting from the crypt base), whereas in the large bowel crypts, in the absence of Paneth cells, the stem cells are at the very base [5, 6]. The mesenchymal intestinal subepithelial myofibroblasts and their secreted growth factors (in particular Wnts) are believed to form and maintain the stem cell niche and thereby regulate epithelial cell function . The stem cells give rise to transit-amplifying cells, which proliferate and differentiate as they migrate upward in the crypt toward the intestinal lumen.
As in other tissues, the recognition of stem cells in the intestine has often been inferred from high expression of the likes of adhesion molecules and drug efflux pumps and/or the constitutive expression of signaling molecules of the Wnt and Notch/Delta pathways. Thus, Musashi-1 (Msi-1) mRNA and protein expression has been confirmed in the murine intestinal stem cell zone , whereas nuclear-localized β-catenin has been reported in the murine colonic crypt base epithelial cells . Proteins encoded by β-catenin target genes (e.g., CD44 and c-Myc) also appear to be expressed in putative stem cells, as well as in some other proliferative cells. Likewise, receptor tyrosine kinases of the Eph subfamily, particularly EphB2 and EphB3, are targets of β-catenin/Tcf-4 signaling and are thus expressed in the stem cell zone, but they also appear in other proliferative cells in superior cell positions [10, –12].
Another potential indicator of intestinal stem cells has been a lack of cytoplasmic E-cadherin expression; E-cadherin is reportedly absent from the cells in the base of human small intestinal crypts, approximately at cell positions 5–7 . In common with many other tissues, the location of the murine intestinal stem cell zone has also been inferred by the presence of label-retaining cells . Also, in the mouse small intestine, a CD45-negative side population fraction that was enriched for Msi-1 and the β1-integrin, implying a stem cell phenotype without actually demonstrating the key properties of self-renewal, clonogenicity, and multipotentiality in this fraction, has been isolated . Interestingly, also found to be highly enriched among murine intestinal basal cells that had been Paneth cell-depleted was a high expression of an RNA-binding protein, nuclear fragile X-interacting protein .
Stem cell populations are believed to be slow-cycling, possessing an infinite proliferative capacity and giving rise to daughter cells from which transit-amplifying populations may emerge; these latter cell populations then differentiate with functional characteristics. Because of their different in situ roles, one might reasonably expect that these three cell populations would have differing biochemical profiles that would result in alterations in IR spectral signatures [17, 18]. One of the most studied stem cell systems is that which resides within intestinal crypts , but specific markers remain elusive. To this end, we set out to determine whether IR spectral analyses (900 to 1,800 cm−1) of paraffin-embedded archived human intestinal tissue using FTIR microspectroscopy would allow us to spatially discriminate the putative stem cell location from that of transit-amplifying and differentiated locations. In a nondestructive manner, IR spectra were collected using globar (15 μm × 15 μm aperture) FTIR microspectroscopy in reflection mode, synchrotron (≤10 μm × 10 μm aperture) FTIR microspectroscopy in transmission mode, or near-field photothermal microspectroscopy (PTMS). These different forms of IR microspectroscopy may differ in their respective means of IR spectral acquisition (i.e., reflection, transmission, or photothermal), along with their particular IR source (i.e., globar or synchrotron radiation); the application of these very different analytical approaches would allow the comprehensive evaluation of the use of IR microspectroscopy to segregate and characterize cell types within intestinal crypts. The aim of this study was to determine whether IR spectral markers of a putative stem cell location within a well-characterized system could be derived; such IR spectral markers may then be compared with other hypothesized markers or applied to tissues where the stem cell location is unknown. Our results using three different forms of IR microspectroscopy agree on a consistent pattern of a biomolecular signature that conforms to the perceived stem cell and transit-amplifying compartments in the human small and large intestinal crypts, suggesting that this technology is a powerful adjunct to our understanding of stem cell biology.
Materials and Methods
Human Intestinal Tissue Preparation for IR Microspectroscopy
Informed consent to obtain normal human intestinal tissue for research was obtained (Local Research Ethics Committee no. 06/Q0604/40). Paraffin-embedded tissue blocks were obtained, and sections 10 μm thick were floated onto either 1 cm × 1 cm Low-E reflective glass microscope slides (Kevley Technologies, Chesterland, OH, http://www.kevley.com) for PTMS or 0.5-mm-thick BaF2 windows (Photox Optical Systems, Sheffield, U.K., http://www.photox.co.uk) for reflection-mode globar or transmission-mode synchrotron FTIR microspectroscopy. The sections were then dewaxed by immersion in xylene (5 minutes) and then washed in absolute alcohol (74OP) (5 minutes). Serial sections (4 μm thick) were stained with H&E to be checked retrospectively to confirm correct orientation.
Globar FTIR Microspectroscopy
IR spectra were acquired with a PerkinElmer Spectrum One FTIR spectrometer and a Spectrum Spotlight microscope (PerkinElmer Life and Analytical Sciences, Boston, http://www.perkinelmer.com). Three large and three small intestinal crypts were interrogated on a cell-by-cell basis using a 15 μm × 15 μm aperture; 10 spectra were acquired per crypt, with the first spectrum obtained at the base of each and sequential analyses made along the length toward the crypt orifice. The tissue sections mounted on IR-transparent BaF2 windows were placed on a Low-E reflective glass microscope slide placed in the microscope platform. Spectral collection was made in reflection mode (2 cm−1 resolution, 256 co-additions), and spectra were converted to absorbance using Spectrum software (PerkinElmer). Raw spectra were processed by 13-point smoothing, baseline-corrected, and normalized to the amide I (∼1,650 cm−1) absorbance peak using Bruker OPUS software (Bruker Optics Ltd., Coventry, U.K., http://www.bruker.co.uk). An image map of large intestinal crypts was also obtained using the Nexus FTIR globar source (50 μm × 50 μm aperture with a step size of 50 μm, 512 coadditions).
Despite its wide application, optical detection FTIR microspectroscopy may not facilitate the acquisition of true IR spectra from opaque materials (e.g., condensed chromatin). Also, despite recent advances in FTIR microspectroscopy, there remains a need for even higher spatial resolution; near-field IR PTMS achieves this. PTMS is implemented by integrating an optical interface (Specac, Orpington, U.K., http://www.specac.com), a scanning-probe microscope (Explorer; Veeco Instruments, Santa Barbara, CA, http://www.veeco.com) equipped with a Wollaston wire thermal probe (Veeco Instruments), and an FTIR spectrometer (Vector 22; Bruker Optics Ltd.). A photothermal probe is positioned at a location selected from the acquired atomic force microscopy images, the sample is exposed to IR, and the heat generated as a result of the sample material absorbing radiation is directly measured. The crucial difference between PTMS and conventional microspectroscopic techniques based on optical detection is that this near-field sensor circumvents the diffraction limit and can be used for both imaging and thermal detection. Under ideal conditions, spatial resolution depends only on the dimensions of these ultrasharp thermal sensors (Anasys Instruments, Norwich, U.K., http://www.anasysinstruments.com), which are tens of nanometers wide at the tip. It is possible to achieve discrimination between phases better than 200 nm (in this study, we used a resolution of ∼10 μm), so that under certain conditions, a spectrum may be obtained from an individual cell (or specific organelle), and the chance of sampling across cell population boundaries is greatly reduced. In addition, PTMS measures IR absorption more directly than optical-detection FTIR spectroscopy, where absorption is calculated from transmission measurements.
True absorbance spectra were acquired using PTMS for 10 large intestinal crypts. The tissue section, mounted on a Low-E reflective glass slide, was placed in the spectroscopic chamber, with the Wollaston wire thermal probe being brought into contact. Along the length of randomly selected large intestinal crypts (supplemental online data), seven spectra were acquired using the thermal probe (8 cm−1 resolution, 1,024 co-additions) and were converted into absorbance spectra using Brucker OPUS software (Bruker Optics). The first spectrum was taken at the base of the crypt, and the probe was moved up the crypt with IR spectra acquired at an additional six sequential locations. Background spectra were obtained before spectral analysis of each crypt to compensate for atmospheric changes and ensure that the probe was free of cellular material. Spectra were baseline-corrected and normalized to the amide I (∼1,650 cm−1) absorbance peak using Bruker OPUS software (Bruker Optics).
Synchrotron FTIR Microspectroscopy
Unless otherwise stated, FTIR data were primarily obtained on beamline 11.1 on the synchrotron radiation source at Daresbury Laboratories (Warrington, U.K., http://www.srs.ac.uk); confirmatory experiments were subsequently carried out on the FTIR end-station of the ID 21 beamline at the European Synchrotron Radiation Facility (ESRF, http://www.esrf.eu), Grenoble, France (as indicated). On both synchrotrons, a Nexus-FTIR spectrophotometer (Thermo Scientific Inc., Waltham, MA, http://www.thermo.com) coupled to a Nicolet Continuum microscope and mercury cadmium telluride detector cooled with liquid nitrogen, with a measuring range of 650 to 4,000 cm−1, was used. Fifteen large and 15 small intestinal crypts were interrogated, and 10 spectra per crypt were obtained. Spectral collection was made in transmission mode (4 cm−1 resolution, co-added for 256 scans), and spectra were converted to absorbance using Thermo Omnic 7.1 software (Thermo Scientific Inc.). At Daresbury Laboratories, the high brilliance of IR from the synchrotron radiation source allowed an aperture of 10 μm × 10 μm, while still achieving a high signal-to-noise ratio . Raw spectra were processed by 13-point smoothing, baseline-corrected, and normalized to the amide I (∼1,650 cm−1) absorbance peak using Bruker OPUS software (Bruker Optics).
Synchrotron FTIR spectral image maps of intestinal crypts, in which image contrast is determined by the absorbance intensity at a chosen wavenumber, were obtained in transmission mode. At Daresbury Laboratories, an aperture of 10 μm × 10 μm was used with a step size of 10 μm, allowing maps composed of pixels (128 co-additions) at a 10 μm × 10 μm resolution to be generated within an acquisition time of ∼6 hours. Spectral maps were baseline-corrected, and two-dimensional (2D) maps were processed with either linear or spline smoothing using Thermo Omnic 7.1 software. During synchrotron FTIR microspectroscopy analyses, a new background was taken every 2 hours to correct for atmospheric alterations or changes in beam current. At the ESRF, an aperture of 8 μm × 8 μm was used with a step size of 8 μm, allowing maps composed of pixels (100 co-additions) at an 8 μm × 8 μm resolution to be generated within an acquisition time of ∼6 h; for these image maps, a background reading was taken every 30 minutes.
Ten spectra per crypt were obtained using the globar or synchrotron radiation source for both large and small intestinal samples. Individual crypts were interrogated on a location-by-location basis (Fig. 1) from evidence derived from the literature . To this end, we examined whether IR microspectroscopy might be used to segregate two regions, one at the base of large intestinal crypts and the other at cell positions 4–6 in small intestine crypts. Large intestinal crypts were interrogated by assigning the following spatial format: putative stem cell region (locations 1–4), transit-amplifying (locations 5–8), and differentiated (locations 9 and 10). Small intestinal crypts were interrogated by assigning the following spatial format: transit-amplifying (locations 1–3, although these might also be considered differentiating/differentiated due to the presence of Paneth cells), putative stem cell region (locations 4–6), and differentiated (locations 7–10). Seven spectra per large intestinal crypt were obtained using PTMS, and these were interrogated by assigning the following spatial format: stem cell (locations 1 and 2), transit-amplifying (locations 3–5), and differentiated (locations 6 and 7) regions. The spectral regions processed included the biochemical-cell fingerprint region (900 to 1,800 cm−1), protein conformation region (1,480 to 1,800 cm−1) and DNA/RNA region (900 to 1,425 cm−1). Initial transformation used second-derivative processing (as indicated), baseline correction, and normalization to the amide I (∼1,650 cm−1) absorbance peak.
FTIR microspectroscopy generates massive amounts of complex data, with each individual spectrum containing up to 900 wavenumbers (cm−1). With each experiment requiring the acquisition of tens of spectra, it is often difficult to identify important, often subtle, differences from the information that is collected. However, recent developments in the use of multivariate analysis and in particular PCA now allow more efficient analysis of spectra. Moreover in the biomedical field, IR spectroscopic studies can involve the processing of data derived from many samples, clustered into classes such as category of sample  or patient identity . This analytical approach requires reliable methods for acquiring two types of information: cluster (scores) plots to separate the classes, and loadings plots to identify the chief contributory variables that identify the class-specific information on which of the wavenumbers, representing various molecular groups, are responsible for observed class clustering. PCA is built on the assumption that variation implies information. It replaces the original several hundred wavenumber variables by linear combinations thereof, termed principal components (PCs), which seek to capture as much variability as possible. The PCs are automatically listed in order, each one accounting for a portion of the original data variance: often, typically no more than the first 10 or so need to be taken into account. For each spectrum obtained, the whole set of many hundreds of readings (one for each wavenumber) is replaced by a small number of “scores,” one for each significant PC. Thus, in a scores plot, each set of measurements (making up an IR spectrum) appears as a single point, whose coordinates are its scores on the one, two, or more PCs chosen as axes for the plot. Increasing spatial separation between scores in a scores plot is proportional to the level of dissimilarity in absorbance spectra. Spectra, transformed as described above, were processed by PCA performed using the Pirouette software package (Infometrix Inc., Woodinville, WA, http://www.infometrix.com) to allow the identification of significant spectral alterations.
PCA was used for preliminary data reduction, and the output was processed using linear discriminant analysis (LDA) . In LDA, a choice of predetermined classes (i.e., category of cell type based on locations [Fig. 1]) was taken into account during the derivation of clusters and loading plots. This allows new variables to be found such that the ratio of the between-cluster variance to the within-cluster variance is maximized, giving optimal separation. LDA has the additional advantage in that the best view for class discrimination is mechanically generated. PCA-LDA with second-derivative processing was performed on the biochemical-cell fingerprint region (900 to 1,800 cm−1), protein conformation region (1,480 to 1,800 cm−1), and DNA/RNA region (900 to 1,425 cm−1). The paraffin band region (1,420 to 1,480 cm−1) was excluded from these analyses.
Immunohistochemical Analyses for Other Phenotypical Markers
Serial tissue sections (4 μm) were placed onto frosted glass microscope slides and dried overnight at 37°C. Tissue sections were dewaxed by a 5-minute immersion in sulfur-free xylene (BDH Chemicals, VWR International Ltd., Lutterworth, U.K., http://www.vwrsp.com) and rehydrated by 3-minute incubations in decreasing concentrations of ethanol (100%, 95%, 80%, 70%, and 50%) in distilled water. Sections were then treated with 30% hydrogen peroxide (BDH Chemicals) in methanol to block endogenous peroxidases, rinsed in tap water, and washed in phosphate-buffered saline (PBS) with 1% Tween 20 (PBST; Sigma-Aldrich, St. Louis, http://www.sigmaaldrich.com). Antigen retrieval was performed by microwave treatment in preheated sodium citrate buffer (pH 6) for 10 minutes at 700 W. Sections were then rinsed in tap water, washed in PBST, and incubated in a serum-free protein block (Dako, Glostrup, Denmark, http://www.dako.com). Sections were then incubated in primary antibody (rabbit polyclonal CD133 [1:500 overnight; Abcam, Cambridge, MA, http://www.abcam.com] or mouse monoclonal β-catenin [1:500 overnight; BD Biosciences, Franklin Lakes, NJ, http://www.bdbiosciences.com]). Sections incubated with CD133 and β-catenin were then incubated with a biotinylated secondary antibody (Dako REAL detection system) for 30 minutes, and a tertiary layer of streptavidin-horseradish peroxidase (Strep-HRP; Dako REAL detection system) was finally applied for 30 minutes. Tissue sections were washed three times for 5 minutes each time in PBST between each antibody layer and following the tertiary layer. Strep-HRP was detected by 3,3'-diaminobenzidine (Dako REAL detection system) for 2–10 minutes, washed in PBS for 2 × 5 minutes, rinsed in tap water, and counterstained lightly with hematoxylin.
Results and Discussion
Globar FTIR Microspectroscopy
Used to interrogate small and large intestinal crypts on a cell-by-cell basis, a globar source FTIR has a relatively weak IR beam. However, this is a new approach to deriving a possible stem cell phenotype coupled with a completely objective computational analysis. Acquired absorbance spectra of the “biochemical-cell fingerprint” region (900 to 1,800 cm−1) of small (Fig. 1A) and large (Fig. 1B) intestinal crypts were obtained following sequential cell-by-cell analyses. Figure 1A shows PCA-LDA performed on the DNA/RNA spectral region (900 to 1,425 cm−1), with classes assigned as transit-amplifying (locations 1–3), putative stem cell region (locations 4–6), and differentiated (locations 7–10). Assigning LDA classes ensures that maximum separation is achieved reducing the influence of intraclass variation and maximizing interclass variation. PCA-LDA scores plots of the small intestinal crypts show clear clustering of the different cell types, with only slight overlap between the stem and differentiated region (Fig. 1A). Based on their spatial location along the length of individual crypts and according to their assigned function, a remarkable separation of scores relating to different putative cell types (i.e., stem cell, transit-amplifying cell, and differentiated cell) was obtained (Fig. 1A).
PCA-LDA was performed on the DNA/RNA spectral region (900 to 1,425 cm−1) of large intestinal crypts, with locations putatively assigned as stem cell, transit-amplifying, or differentiated regions (Fig. 1B). Again, clear discrimination was observed between cells based on spatial location along the length of individual crypts. An obvious segregation of scores from location 1 (base of the crypt) to location 10 (toward the intestinal lumen) of the large intestinal crypts was observed. As with small intestinal crypts, it is clear that cells in specific locations seem to share similar biochemical properties (Fig. 1). Segregation of cell spectra based on location following PCA-LDA of the DNA/RNA IR spectral region suggests that biochemical markers may be identified using this approach. Modifications in this spectral region may be associated with DNA-conformational alterations  and, therefore, chromatin remodeling. These findings lay the basis for a novel nondestructive methodology that allows tissue sample or cell interrogation using mid-IR technologies prior to subsequent staining with other markers or isolation of defined populations. Our results clearly indicate that the assigned transit-amplifying and differentiated regions are biochemically different based on IR spectral characteristics to the putative stem cell region. The power of this technology is that combining IR spectral acquisition with computational analysis allows one to characterize the cells in a spatial fashion, thus providing a novel approach to segregating different cell populations in situ within tissues.
PTMS facilitates the derivation of true absorbance spectra based on direct measurement of heat generated independent of IR opacity . Because thermal waves are detected instead of IR, resultant spectra are derived at a resolution that is not diffraction-limited (∼10 μm). Mitotic chromosomes (condensed chromatin) may be IR opaque, and this paradoxically can lead to decreases in intensity in the DNA/RNA spectral region (900 to 1,425 cm−1), even though there may actually be an increase in DNA levels . Testing the notion that DNA folding is altered in stem cells  requires a technique that generates a true measure of DNA state. In each of 10 randomly chosen large intestinal crypts, seven IR spectra were derived in sequential fashion from the crypt base to the intestinal lumen (supplemental online data). PCA-LDA scores plots of the DNA/RNA spectral region with second-derivative processing gave the best segregation between the assigned cell classes (Fig. 2). Although some overlap between different cell classes was noted (Fig. 2A), good segregation of clusters derived from the assigned stem cell (red symbols), transit-amplifying (black symbols), and differentiated (blue symbols) IR spectral locations were noted. The resultant loadings plot pointed to 1,155 cm−1 (carbohydrates) and 1,080 cm−1 (symmetric [νs]PO2−) as being responsible for the segregation of the average IR spectrum of the putative stem cell class from the average of all the spectra derived from all three assigned classes (Fig. 2B). Interestingly, for the different classes (i.e., assigned locations), the same chemical entities as identified in the loadings plot, segregated the putative differentiated cell locations in this objective statistical analysis.
Synchrotron FTIR Microspectroscopy
A synchrotron radiation source generates high-brilliance IR and allows spectral interrogation approaching the diffraction limit . Small (n = 15) and large (n = 15) intestinal crypts were interrogated, with 10 IR spectra sequentially acquired per crypt. Following interrogation of small and large intestinal crypts, marked differences (including peak location and intensity) were observed throughout the IR spectral “fingerprint” region (900 to 1,800 cm−1) (Fig. 3A, 3C). Using the 10 IR spectra acquired per individual crypt (Fig. 3 shows a representative example of the analysis for a small and large intestinal crypt; other analyses are found in the supplemental online data), PCA was performed on this entire biochemical-cell fingerprint region to determine whether segregation between the three assigned regions could be achieved (Fig. 3B, 3D). For individual crypts, clear segregation of the assigned regions was observed.
To determine whether cluster segregation of assigned cell locations (i.e., classes) was due to a particular IR spectral region, PCA-LDA was performed on all 15 crypts of either small or large intestinal crypts (Fig. 4). To this end, from each assigned cell class, a representative IR spectrum was selected per crypt. PCA-LDA with second-derivative processing was performed on the biochemical-cell region (900 to 1,800 cm−1), protein conformation region (1,480 to 1,800 cm−1), and DNA/RNA region (900 to 1,425 cm−1). Despite some overlap, in both instances the designated DNA/RNA region gave rise to the best clustering and segregation of assigned cell classes (Fig. 4A, 4B). LDA loadings plots were derived for the biochemical fingerprint region of small and large intestinal crypts (supplemental online data). IR spectral regions of common importance were those designated for protein conformation (1,480 to 1,800 cm−1), asymmetric PO2− (1,225 cm−1), and νsPO2− (1,080 cm−1).
To determine whether a common biomolecular pattern underlies the putative stem cell, transit-amplifying, or differentiated locations, both the small and large intestinal crypts were compared (Fig. 5). Average IR spectra derived from the 10 corresponding locations sequentially analyzed along the large (n = 15) and small (n = 15) intestinal crypt lengths were obtained (Fig. 5A, 5B). Again, large intestinal crypts were divided into putative stem cell (locations 1–4), transit-amplifying (locations 5–8), and differentiated (locations 9 and 10) populations, whereas small intestinal crypts were separated into transit-amplifying (locations 1–3), putative stem cell (locations 4–6), and differentiated (locations 7–10) populations. PCA-LDA with second-derivative processing was performed, with classes assigned as putative stem cell, transit-amplifying, or differentiated regions using the entire biochemical-cell region (900 to 1,800 cm−1). Excellent segregation was achieved for the assigned locations (Fig. 5C, 5D).
Spectral Image Maps
IR spectral image maps allow one to track the spatial distribution of chemical entities based on levels of relative absorbance intensity at a chosen wavenumber in a pixel-by-pixel fashion. This results in the acquisition of an image map in which the absorbance intensity is proportional to thermal color changes: blue (lowest intensity) < green < yellow < red (highest intensity). PCA-LDA consistently highlighted νsPO2− (1,080 cm−1) as being a major contributor to IR spectral segregation of the different assigned classes, pointing to an important role of alterations in the secondary structure of DNA in chromatin. Symmetrical stretching relates to bonds that stretch in-phase. Symmetric modes (νs) of PO2− (the in-phase dioxy O=P=O stretch) have been attributed to DNA backbone structure , and these signals may be absent in differentiated or metabolically inactive cells . DNA can adopt a number of helical conformations, broadly classified as A, B, and Z. νsPO2− stretching vibrations of the phosphate linkage are observed with dehydration as DNA undergoes a structural transition from B-form (hydrated) to A-form (dehydrated) ; the B-form is associated with histone-complexed DNA, whereas the A-form may be nonhistone protein DNA [26, 27]. Under physiological conditions, B-form DNA is generally most prevalent in a cell's genome, and the exact role of A-form DNA (which may occur in patches) remains unknown. A decreased intensity of νsPO2− stretching vibrations may be due to a decrease in the dielectric constant in the vicinity of the phosphate group . The presence of Mn2+ also decreases the IR intensity of νsPO2− , associated with resulting DNA-conformational alterations .
Although B-form DNA is most common, a number of microenvironmental influences, such as water activity and counter-ion identity and/or concentration, influence the relative stabilities of these different forms [27, –29]. The role of an A-form DNA patch (which νsPO2− may represent) within the matrix of B-form DNA remains unclear, but it may contribute significantly to DNA bending and provide a focal point for the binding of architectural proteins (e.g., nonhistone chromatin protein HMGB1 and histone H1) into the minor groove [27, –29]. Also, A-form patches may become a focal point for protein binding because B-form to A-form transition increases the hydrophobicity of the DNA surface in the minor groove. The A-form with exaggerated base-pair inclinations is also induced by TATA-binding protein, a transcriptional promoter in eukaryotes. As a point of note, the junction between B-form and A-form DNA is a known target for the anticancer drug cisplatin. However, the exact physiological function of A-form DNA remains unknown and is an area of intense research investigation. The biological significance of these findings and whether this may be a stem cell characteristic remain to be ascertained, and future experiments could involve DNA isolation from tissue regions of interest prior to subsequent spectral interrogation; used on archived or freshly obtained tissue, such IR analyses could yield insights into the biology underlying the role of DNA conformation in stem cell maintenance.
Figure 6A shows H&E staining of a serial tissue section of three small intestinal crypts and the resultant IR image maps obtained at a resolution of 10 × 10 μm using synchrotron FTIR microspectroscopy at Daresbury Laboratories. The generated 2D map (Fig. 6C) demonstrated increased spectral intensity associated with νsPO2− (1,080 cm−1) at locations that would correspond approximately to cell positions 4–5 (counting from the crypt base) thought to house the stem cells (Fig. 6B) . Our findings are promising because these intensity fluctuations were not unique to individual crypts.
IR image maps were also obtained from large intestinal crypts using both globar and synchrotron FTIR microspectroscopy at Daresbury Laboratories. Compared with a serial H&E-stained section, four adjacent large intestinal crypts were examined using globar FTIR microspectroscopy (50 μm × 50 μm pixel resolution) (Fig. 6D–6F). Despite this relatively low resolution mapping, three of the four crypts clearly demonstrate an increased intensity at 1,080 cm−1 at the base of the crypts, in the putative stem-cell region [5, 7]. Interestingly, the base of a fifth crypt in the top right corner is also identifiable (Fig. 6F); so intense is this signal that we believe it dampens the spectral intensity of the νsPO2− derived from the main crypts of interest. Using synchrotron FTIR microspectroscopy, two large intestinal crypts were mapped (Fig. 6G–6I). In comparison with a serial H&E-stained section (Fig. 6G), an image map at 1,080 cm−1 clearly discriminated the putative stem cell region of large intestinal crypts (Fig. 6I). νsPO2− appeared in all cases to be the most robust marker of the putative stem cell region in both small and large human intestinal crypts.
Comparison with Other Phenotypical Markers
Because of its high-brilliance IR, a synchrotron radiation source allows for its application to the imaging and/or localized spectroscopy of received samples. In doing this, it allows both for optical visualization and imparting chemical information. To confirm our initial findings at Daresbury Laboratories, a subsequent set of image maps obtained following IR spectral interrogation of a new set of tissue sections from the same samples was generated at the ESRF. These new samples were interrogated at a higher resolution (8 μm × 8 μm), totally independently from our previous analyses (Fig. 7A, 7D). In this case, we were able to superimpose the IR spectral image map exactly on the interrogated region. The findings of these subsequent experiments highlighted νsPO2− with greatest intensity located in a ring in the small intestinal crypt, at approximately cell positions 4–5 (counting from the crypt base) (Fig. 7A), whereas in the large bowel crypts, this occurred primarily at cell positions 1–4, at the base (Fig. 7D). In comparison with some proposed immunophenotypic markers of intestinal stem cells, the present results provide extremely consistent spatial biochemical differences between the three compartments, but, for example, nuclear β-catenin immunoreactivity was observed only in Paneth cells of the small intestine (Fig. 7B), whereas only weak immunoreactivity was seen in one or two basal epithelial cell nuclei in the occasional large intestinal crypt (Fig. 7E). CD133 expression has been used to isolate tumor-initiating cells from human colon cancer [30, 31], but the expression pattern in the normal human crypts was weak and nonspecific (Fig. 7C, 7F). We observed no Msi-1 expression in these normal crypts (data not shown).
Although a wide variety of so-called markers have been proposed for intestinal stem cells, none are unique and certainly many are not confined to the stem cell region. For example, Msi-1 positively regulates the transcriptional repressor molecule Hes-1  through its binding to m-Numb, thereby perpetuating Notch signaling. In the mouse small intestine, Msi-1 and Hes-1 are coexpressed in the stem cell zone, although Hes-1 expression also extends to proliferative cells higher up the crypt . Likewise in the mouse, after crypt damage, Msi-1 is expressed by many proliferative cells above the stem cell zone, possibly as a consequence of the long half-life of the protein in conjunction with rapidly cycling regenerating cells. In human colonic crypts, expression of Msi-1 in cells scattered between cell positions 1 and 10 supports the notion that Msi-1 is not a specific marker of stem cells . Nuclear β-catenin has been observed in basal epithelial nuclei in the mouse large intestine , but in the murine small bowel, it is only expressed in Paneth cell nuclei , an observation that we also made in our human tissue. The use of FTIR microspectroscopy is a novel approach to discriminating the stem cell compartment [17, 18]. It has potential for high-throughput and live-cell monitoring [3, 36, 37], thus laying the basis for novel cell-sorting technologies based on an integrated IR spectral phenotype. Specific structural DNA phenotypes may be identified within tissues [38, 39] and may, in cases where the source of stem cells is alternative to that of a specific tissue , even allow the identification of their origin. This study is highly promising, as it demonstrates that the biomolecular characteristics of locations associated with the small and large intestinal crypts, namely stem cell, transit-amplifying, and differentiated regions, can be spatially segregated based on their IR spectral characteristics. Although assigned different locations along the crypt length depending on whether they were derived from small or large intestine, spectra strongly cosegregated based on class, demonstrating that the biomolecular signatures of these different putative cell types coincide, regardless of their origin (Fig. 5). The most promising marker appears to be νsPO2− (1,080 cm−1), which appears important for such IR spectral segregation.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.
This work was funded by the Biotechnology and Biological Sciences Research Council (Grant BB/D010055/1; to A.H., F.L.M., and N.J.F.), an Engineering and Physical Sciences Research Council studentship (to O.G.), and the Rosemere Cancer Foundation (to M.J.W., P.L.M.-H., and F.L.M.). We also thank Science and Technologies Facilities Council and the European Synchrotron Radiation Facility (Award MD-271) for grant support during access to synchrotron facilities.