Localisation and characterisation of incipient brown-rot decay within spruce wood cell walls using FT-IR imaging microscopy.

Spruce wood that had been degraded by brown-rot fungi (Gloeophyllum trabeum or Poria placenta) exhibiting mass losses up to 16% was investigated by transmission Fourier transform infrared (FT-IR) imaging microscopy. Here the first work on the application of FT-IR imaging microscopy and multivariate image analysis of fungal degraded wood is presented and the first report on the spatial distribution of polysaccharide degradation during incipient brown-rot of wood. Brown-rot starts to become significant in the outer cell wall regions (middle lamellae, primary cell walls, and the outer layer of the secondary cell wall S1). This pattern was detected even in a sample with non-detectable mass loss. Most significant during incipient decay was the cleavage of glycosidic bonds, i.e. depolymerisation of wood polysaccharides and the degradation of pectic substances. Accordingly, intramolecular hydrogen bonding within cellulose was reduced, while the presence of phenolic groups increased.

Introduction

Brown-rot fungi are aggressive colonisers of wood. They degrade and mineralise the wood polysaccharides leaving behind a strongly modified lignin, characterised by its lower methoxyl and higher hydroxyl group content, which also suffers some depolymerisation [1], [2]. The ability to break down the recalcitrant lignocellulose complex suggests employing these organisms for biomass pre-treatments to gain better access to the structural wood polymers and facilitate biorefinery fractionation processes [2], [3]. Therefore, it is of great importance to trace and characterise degradation processes within lignocellulosic substrates.

Chemical changes within wood cell walls in early stages of brown-rot – often referred to as incipient ones – are marginal and often beyond the limit of detection of standard methods of chemical analyses. Mechanical properties, however, suffer dramatically during incipient decay, often even before mass loss becomes apparent and the overall appearance of the wood has been altered [4]. Accessibility of the wood cell wall polymers to water becomes significant very early in the decay [4], [5]. Both features lead to the conclusion, that major changes during early brown-rot concern the structural chemistry and assembly of the wood cell wall components cellulose, hemicelluloses, pectins, and lignin, rather than the chemical composition in terms of cellulose, hemicelluloses, and lignin content.

FT-IR spectroscopy is a powerful tool to investigate the composition of biodegraded wood [6], [7], [8], the molecular interactions between wood polymers [9] as well as hydrogen bonding within and between cellulose chains [10], [11]. Classical FT-IR spectroscopy is not suited to investigate wood on the cellular level. FT-IR imaging microscopy is like FT-IR spectroscopy but allows for the localisation of wood degradation processes on the spatial level of single softwood tracheids [12], [13].

FT-IR imaging microscopy was here applied to brown-rot degraded wood in early degradation states to localise traces of incipient decay. FT-IR spectra were analysed by means of multivariate methods to obtain the descriptive information relating to fungal degradation processes. Based on multivariate models, several multivariate image analysis (MIA) techniques were applied to allocate the observed spectral changes to distinct cell wall regions. As a reference, mainly to support the spectral data, the composition of amorphous wood polysaccharides of all wood samples was analysed with the methanolysis method.

Methods

Fungal cultures

Poria placenta (syn. Oligoporus placenta) MAD 698 and Gloeophyllum trabeum CBS 900.73 that had been maintained on agar slants at 4 °C were pre-incubated on malt extract agar plates (Fluka) prior to their use.

Preparation of degraded spruce wood samples

Spruce (Picea abies (L.) Karst.) sapwood blocks (3.5 radial × 3 tangential × 4 longitudinal cm3) from a 120-year-old tree grown near St. Pölten, Lower Austria were dried to a constant weight at 50 °C. After weighing, the blocks were impregnated in a vacuum with water and γ-sterilised (25.5 kGy min). Sterilised blocks were transferred into glass jars (900 ml, approx. 12 cm Ø, 2 blocks for one jar), in which fungal cultures had been pre-cultivated on malt extract agar for 14 days. A sterile plastic grid served as spacer between the agar surface and the wood block. The samples were incubated for several weeks at 28 °C (Table 1) at a high relative humidity. After incubation, mycelia were carefully wiped off the wood surfaces and wood samples were dried again at 50 °C to a constant weight to determine the mass loss. Selected samples (Table 1) were impregnated with water and thin sections of one tracheid diameter thickness (approx. 30 μm) were carefully cut in the radial direction by means of a sliding microtome. Samples were cut from a region at least 1 mm from the surface of the wood block to avoid non-representative regions of stronger degradation. Dried thin sections were mounted on transparency frames and subjected to transmission FT-IR microscopy at ambient temperature at approximately 40% relative humidity.

Table 1

Investigated spruce wood samples: decay fungus, incubation time, mass loss, and number of recorded FT-IR images.

	Incubation time [days]	Mass loss [%]	No. of FT-IR images
Spruce	–	–	8
G. trabeum	21	4	3
G. trabeum	21	6	3
G. trabeum	28	12	7
G. trabeum	28	16	4
P. placenta	21	n.d.	2
P. placenta	56	8	4
P. placenta	56	16	5

n.d., not detectable.

Analysis of monomeric sugars

Monomeric sugars were quantified by methanolysis according to Sundberg et al. [14]. Contrary to acid hydrolysis, this method allows for the selective lysis of amorphous wood polysaccharides and for the quantification of uronic acids. Ten milligrams of thin sections adjacent to the ones investigated by FT-IR were analysed at least in duplicates from each wood block sample. Differences exceeding twice the standard deviation of repeated measurements were allocated as significant.

Fourier transform imaging infrared microscopy

The FT-IR images were recorded in transmission mode on a Spectrum Spotlight 400 FT-IR microscope connected to a Spectrum 100 FT-IR spectrometer (PerkinElmer Inc.). The area of interest was first displayed by a CCD camera and then it was irradiated with mid-IR light (Fig. 1A and B). The scanning was performed in imaging mode by a 16 point dual array liquid N2 cooled MCT detector. Samples were moved stepwise in the x- and y-direction resulting in the assembly of a chemical image of the pre-defined area. The measurement provided a pixel resolution of 6.25 × 6.25 μm2. The FT-IR spectra were recorded in 8 cm−1 spectral resolution, zerofilling factor 1, between 4000 and 720 cm−1. 16 scans per pixel were averaged, to increase signal to noise ratio.

Fig. 1

CCD camera image of a degraded spruce wood section from the sample P. placenta – 16% mass loss (A); pre-selected area M (B), and FT-IR pseudo-colour spectral absorbance image (C). Areas of tracheids (T01–T08) are marked with rectangles. Zones of high total absorbance (red) show a high contribution of middle lamellae (ML), primary cell walls (P), and outer layers of secondary cell walls (S1) those with low total absorbance (blue) can be assigned to regions with a high contribution of secondary cell walls (S2) and pits. (D) shows an example of a multivariate PLS-DA prediction image – degraded pixels are plotted yellow.

Image areas were chosen at random positions within the wood section (Table 1). However, special care was taken to avoid any contribution of ray cells or resin channels. Most FT-IR images were recorded in early-wood as thick late-wood cells absorbed too much IR radiation to obtain proper spectra.

FT-IR microscopic images result in a 3D data matrix (I × J × K), where I and J denominate the pixels along the radial and longitudinal directions and K represents the variables of the IR spectrum (Scheme 1). The first output of IR-imaging microscopy, however, is a pseudo-colour FT-IR absorbance image calculated from the average absorbance of the whole IR range after atmospheric compensation (Fig. 1C).

Scheme 1

Approaches to multivariate image analysis (MIA) for visualisation of brown-rot degradation (A) PCA based MIA; and (B) PLS-DA based MIA for classification of unknown spectra.

Data processing for multivariate data analysis (MVA) and multivariate image analysis (MIA)

Raw spectra of each pixel were subjected to atmospheric compensation by means of the PerkinElmer (PE) Spotlight software. Spectra were transferred into The Unscrambler® (Vsn. 9.8; www.camo.com) for all further manipulations. For that purpose, extracted single PE spectra were converted to OPUS spectra (OPUS software version 6.5; www.brukeroptics.com) and as such imported into The Unscrambler® resulting in an unfolded 2D data matrix (IJ × K; Scheme 1).

For normalisation, spectra were subjected to SNV (standard normal variate) transformation, a data pre-treatment that scales and centres each spectrum and removes scattering effects. Afterwards, spectra were differentiated to 2nd derivatives using the Savitzky and Golay algorithm (9 smoothing points, 2nd order polynomial) [15] to compensate adverse baseline effects, to resolve overlapping IR bands, and to facilitate the interpretation of the MVA results. 2nd derivatives of the spectra are characterised by amplitude maxima and minima. The latter are positioned at the original maxima of the raw absorbance spectra. Principal component analysis (PCA) and partial least squares (PLS) regression models were calculated after mean-centring of the data in the spectral range between 1840 and 1096 cm−1, and 1016 and 736 cm−1. The range between 1092 and 1020 cm−1 (C–O valence vibrations, ν C–O) was omitted due to the high absorbance of many of the samples in this region.

Spectra of tracheids were extracted from the FT-IR absorbance images by calculating the average spectrum of the tracheid area between two middle lamellae (Fig. 1C). Pre-treatment was performed in the same manner as for the pixel spectra (atmospheric compensation, SNV, 2nd derivative).

Crystallinity indices were calculated by means of the OPUS software. For that purpose, 2nd derivative spectra were calculated from raw tracheid spectra after atmospheric compensation (Savitzky–Golay algorithm, 9 smoothing points, 2nd order polynomial). Ratios between the amplitude minima near 1368 and 896 cm−1 (SD1368/896), between those near 1112 and 896 cm−1 (SD1112/896), and between 1508 and 1368 cm−1 (SD1508/1368) were calculated.

Absorbance spectra presented in Fig. 4 have been subjected to a rubber band baseline correction (concave base line correction, 64 points, 2 iterations) using OPUS and were vector normalised over the presented range.

Fig. 4

Vector normalised absorbance spectra of spruce tracheid spectra highlighted in Fig. 3(A). (a) P. placenta – 16% ML; (b) G. trabeum – 6% ML, (c) G. trabeum – 16% ML, (d) P. placenta – 8% ML, and (e) non-degraded spruce.

Partial least squares discriminant analysis (PLS-DA) was performed using the PLS2 algorithm of The Unscrambler® (cross validation; 30 data blocks). Dummy response variables were introduced to describe the categories: Y = 0 for spectra from non-degraded samples and Y = 1 for selected spectra from degraded samples for a first regression model and vice versa for a second.

Visualisation of the multivariate images

PC scores and residuals, and PLS-DA prediction and deviation values were imported via MS Excel (www.microsoft.com) into Solo + MIA (www.eigenvector.com), refolded and plotted as pseudo-colour images. Pixel rows with missing data were not considered in the plots (Fig. 8).

Fig. 8

Photographic, FT-IR spectral absorbance and PCA scores and residuals images (1840–1096 and 1016–736 cm−1) of spruce wood sections degraded by brown-rot fungi to different extents. (A) Non-degraded spruce; (B) P. placenta – 16% ML; (C) G. trabeum – 16% ML (barsize 50 μm). Spectral absorbance images were slightly scaled to obtain similar contrast.

Results and discussion

Fungal degradation – mass loss

In the laboratory, degradation is traditionally determined as the mass loss (ML) of a wood sample in a wood block decay test. It is well known, that degradation processes are not evenly distributed. Therefore, ML only serves as a crude measure. The intention was to investigate early stages of brown-rot rather than pronounced ones. Therefore samples exhibiting a ML up to 16% were chosen in this study (Table 1). The incubation time to achieve these 16% ML was 28 days for G. trabeum, and 56 days for P. placenta, indicating the different aggressiveness of these two strains under the test conditions.

Composition of monomeric sugars from amorphous polysaccharides

Brown-rot led to an overall decrease of sugar monomers from amorphous polysaccharides (Fig. 2), particularly of the non-glucosic ones. The losses were in accordance with the observed ML. The most characteristic change in low ML samples was the decrease of galactose (Gal) and galacturonic acid (GalA), monosaccharides that are enriched in the middle lamellae, primary cell walls (P), and the outer secondary cell wall layer (S1) [16]. At non-detectable ML (P. placenta) 25% Gal and 19% GalA were lost. In spruce wood, Gal is most abundant in arabinogalactans of the middle lamellae and P, only about 7% of Gal in spruce wood is found in galactoglucomannans. GalA, however, is located in the middle lamellae as a main constituent of galacturonan, a pectic substance [16]. Gal and GalA degradation became more pronounced as brown-rot proceeded. At 16% ML (G. trabeum and P. placenta) more than half of the Gal and a high proportion of GalA (44 and 24%, respectively) were lost.

Fig. 2

Contents of sugars in amorphous polysaccharides in non-degraded spruce wood and spruce wood degraded by G. trabeum (A) and by P. placenta (B). The starting material is represented by 0% ML, and so was the P. placenta sample with non-detectable ML (B). Error bars indicate standard deviations of replicate measurements.

Arabinose (Ara), a constituent of arabinogalactans and arabinans, is also more abundant in the middle lamellae and P than in the S cell walls, where about 6–10 mg/g Ara is found in arabino-4-O-methylglucuronoxylan, a hemicellulose that is slightly enriched in S1 and the outer part of S2 layers. Ara degradation became significant at more pronounced degradation states (G. trabeum 12 and 16% ML; P. placenta 8 and 16% ML). Mannose (Man) from galactoglucomannan, a hemicellulose abundant in the S cell walls of softwoods, was also degraded. The xylose (Xyl) content, however, remained quite stable during degradation up to 16% ML; its degradation therefore corresponded to the ML of wood. The glucuronic acid (GlcA) content, however, increased significantly in the degraded samples exhibiting 16% ML, indicating the recalcitrance of xylo-oligomers in vicinity of GlcA side-chains. This was also reported for the brown-rot fungus Coniophora cerebella (syn. C. puteana) that is lacking extracellular enzyme activity catalysing the hydrolysis of GlcA residues from xylan main-chains [17]. The glucose (Glu) in amorphous polysaccharides also increased while brown-rot proceeded and even doubled in the G. trabeum sample exhibiting 16% ML, indicating the conversion of crystalline cellulose domains to amorphous ones.

Characterising brown-rot decay processes on the cellular level – multivariate analysis of FT-IR spectra of tracheids

The aim of MVA is to reduce the number of variables of e.g. IR spectra to only a few latent orthogonal variables – the principal components (PCs). The contribution of each original spectral variable to each PC becomes visible in the loadings spectra. PCA can be used to identify similarities and dissimilarities of spectra.

Fingerprint region

After PCA of 197 tracheid spectra (1840–1096 and 1016–736 cm−1) the data variance related to brown-rot degradation appeared in the first two PCs which accounted for 61% of the total variance. IR spectra from most tracheids from samples subjected to brown-rot fungi were similar to those from non-degraded spruce wood tracheids; also from samples exhibiting significant ML up to 16%. This finding confirmed that brown-rot processes were not evenly distributed throughout a wood block. However, for some tracheids the spectra separated from the others either on PC1 or on PC2 and scored more positively on either one or both PCs. Tracheid spectra separating along the PC1 axis were from the more severely degraded samples (P. placenta 16% ML and G. trabeum 16% ML) (Fig. 3A). The first PC loading spectrum (Fig. 3B) was dominated by lignin bands (1588, 1508, 1456, 1268, and 852 cm−1; Table 2) which showed minima in the loadings plot, indicating a lignin content higher than the average of all investigated tracheids and higher than that of all non-degraded tracheids, whose score on PC1 is lower. A strong positive loading band at 1168 cm−1, the absorbance region of the C–O–C valence vibration of various glycosidic bonds of polysaccharides, particularly mannan [18], indicates that glycosidic bonds had been cleaved in these tracheids. Other positive loadings (1376, 1116, 896, 808 cm−1) can be assigned to cellulose and hemicelluloses and reveal the lower content of these polysaccharides in these degraded samples.

Fig. 3

PCA scores (A) and loadings plot (B) of spectra (2nd derivative, 1840–1096 and 1016–736 cm−1) of tracheids (Pp: P. placenta treated spruce, Gt: G. trabeum treated spruce; Spruce: non-degraded spruce wood). Spectra plotted in Fig. 4 are highlighted with large markers.

Table 2

Assignments of IR band maxima to various components of wood.a.

Wavenumber [cm−1]	Assignment	Source	Reference
>3450	Free O–H valence vibration	Carbohydrates lignin	[10]
3500	Dimeric formation of an intermolecular hydrogen bond between phenols	Lignin	[30]
3480	Free O–H valence vibration	Spruce wood	This work
3450	O–H valence vibration of weakly hydrogen bonded or free O–H mainly C(2)–O(2)–H secondary alcohol	Cellulose	[10]
3432	Free or weakly H-bonded valence vibration	Spruce wood	This work
3410	O–H valence vibration of C(6)H2O(6)–H primary alcohol (main conformation)	Cellulose	[10]
3375	Parallel oriented hydrogen bonded O–H valence vibration	Cellulose	[11]
3365	O–H of C(6)H2O(6)–H primary alcohol (secondary conformation)	Cellulose	[10]
3342	O(3)H···O(5) intramolecular in cellulose	Cellulose	[10]
3340	Mainly C(3)–O(3)–H secondary alcohol	Cellulose	[10]
3328	Hydrogen bonded O–H valence vibration	Spruce wood	This work
3310–3305	Intermolecular O(6)H···O(3)′ in cellulose	Cellulose	[11]
3280	O(6)H···O(3) intermolecular in cellulose (Iβ)	Cellulose	[34]
3275	O(2)H···O(6) intramolecular in cellulose	Cellulose	[10]
3219	Intermolecular hydrogen bond between biphenols	G-lignin	[30]
< 3200 (3191–3165)	Intermolecular hydrogen bond between biphenol and other phenolic groups	G-lignin, lignin	[30]
1730–1725	C <svg xmlns="http://www.w3.org/2000/svg" version="1.0" width="20.666667pt" height="16.000000pt" viewBox="0 0 20.666667 16.000000" preserveAspectRatio="xMidYMid meet"><metadata> Created by potrace 1.16, written by Peter Selinger 2001-2019 </metadata><g transform="translate(1.000000,15.000000) scale(0.019444,-0.019444)" fill="currentColor" stroke="none"><path d="M0 440 l0 -40 480 0 480 0 0 40 0 40 -480 0 -480 0 0 -40z M0 280 l0 -40 480 0 480 0 0 40 0 40 -480 0 -480 0 0 -40z"/></g></svg> O valence vibration of acetyl groups	Xylan	[39]
1738–1709	CO stretch in unconjugated ketones, carbonyls and ester groups		[40]
1675–1655 (1660)	CO stretch in conjugated p-substituted aryl ketones	Lignin	[40]
1588	Aromatic skeletal vibrations plus CO stretch	Lignin	[40]
1508	Aromatic skeletal vibrations	Lignin	[40]
1430	O–H in plane bending of alcohol groups	Carbohydrates	[10]
1424	Aromatic skeletal vibration with C–H in plane deformation	Lignin	[40]
1372	C–H deformation vibration	Cellulose	[39]
1315	O–H in plane bending of alcohol groups	Carbohydrates	[10]
1317–1315	CH2 rocking vibration	Cellulose	[39]
1268	G-ring plus CO stretch	G-lignin	[40]
1162–1139	C–O–C asymmetric valence vibration	Polysaccharides	[39]
1110–1107	Ring asymmetric valence vibration	Polysaccharides	[39]
1060	C(3)–O(3)H valence vibration mainly from C3–O3H	Polysaccharides	[10]
897	Anomere C-groups C(1)–H deformation, ring valence vibration	Polysaccharides	[39]
864	Glucomannan	Glucomannan	[41]
858–853	C–H out of plane deformation in position 2,5,6	G-lignin	[34]
810	Glucomannan	Glucomannan	[41]

As a result of spectral resolution, zerofilling, and differentiation to 2nd derivatives, bands of spectra in this work slightly differ from literature values.

Other tracheid spectra separate from the non-degraded state along the PC2 axis. These spectra belong to tracheids of the lower ML samples G. trabeum – 4%, G. trabeum – 6%, and P. placenta – 8%. The corresponding loading spectrum was dominated by a positive band at 1160 cm−1. Other polysaccharides derived loading bands that can be assigned to polysaccharides (glucomannan at 868 and 808 cm−1, xylan at 1736 and 896 cm−1) exhibited a loading in the opposite direction, indicating that the absorbance of the glycosidic bonds was negatively correlated to all other hemicelluloses bands. In other words: glycosidic bonds have already been cleaved but the products – saccharides of a lower degree of polymerisation – have not readily been metabolised by the fungus. Lack of utilisation of decomposed polysaccharides was already suggested by Winandy and Morrell [4] who found increased equilibrium moisture content (EMC) as a result of initial liberation of water bonding sites during carbohydrate decomposition. They suggested a turning point at about 10% ML, when EMC decreases in more severely degraded samples [19]. Many studies on the mechanisms of brown-rot support the theory that aggressive and diffusible low molecular mass agents – very likely hydroxyl radicals (OH•) – rapidly depolymerise wood polysaccharides. OH• are produced via a Fenton reaction during which H2O2 generated by fungal enzymes is reduced by ferric ions. These radicals lead to oxidative cleavage of wood polysaccharides. Hydroquinones excreted by the fungus or catecholic structures derived from modified lignin serve as reductants of ferrous to ferric ions [20], [21], [22]. This initial attack opens the wood structure and carbohydrate decomposition products become accessible to fungal enzymes. The brown-rot fungus G. trabeum excretes a number of polysaccharides hydrolysing enzymes such as endoglucanases, processive endoglucanase releasing cellobiose, β-glucosidase and xylanases [1], [23]. In synergism with the low molecular mass carbohydrate depolymerising system, these enzymes convert the wood polysaccharides to monomeric sugars, the energy source of the fungal organism. Furthermore, the secretome of P. placenta during growth on cellulose was consistent with the involvement of Fenton chemistry [24]. The decay pattern found here by means of FT-IR microscopy supports the hypothesis that an unspecific primary attack of polysaccharides leads to the accumulation of decomposition products that then become accessible to enzymatic hydrolysis.

Fig. 4 shows five tracheid absorbance spectra allocated as average and extreme in the PCA scores plot (Fig. 3A). Major differences concern mainly the band near 1160 cm−1, which is narrower in the degraded samples. Pronounced brown-rot leads to significantly increased lignin content, visible in IR spectra [8], [25], [26]. Therefore the lignin derived absorbance bands (1596–1588, 1508, ∼1264 cm−1) were higher in the spectra of the tracheids from the samples exhibiting 16% ML. Only minor differences have been found at the CO valence vibration bands 1740–1710 cm−1. An increasing number of CO bands were related to oxidative processes during brown-rot degradation by Körner et al. [27]. However, CO groups from O-acetylglucuronoxylan overlap with non-conjugated carbonyls from oxidised lignin and the products of oxidative depolymerisation of polysaccharides, e.g. lactones of galactonic acid, mannonic acid and galactaric acid [28]. Putative degradation of acetyl groups and oxidative processes during brown-rot therefore affect this band in a different way.

O–H valence vibrations

O–H valence vibrations of the alcoholic groups of polysaccharides and those of phenolic groups in lignin absorb in this spectral region. Depending on the formation of hydrogen bonds (H-bonds) and the strength of these H-bonds, the excitation energies vary. Generally, free O–H groups show higher resonance energies and IR bands at higher wavenumbers than weakly and strongly H-bonded OH groups. Wood is a composite of crystalline, para-crystalline, and amorphous polysaccharides and lignin, exhibiting a large spectrum of differently H-bonded O–H groups that are highly overlapping. Many of these bonds have been assigned to IR bands (Table 2)

After PCA of 197 2nd derivative spectra of tracheids in the O–H region (SNV, 3600–3100 cm−1) the data variance describing brown-rot degradation was found in the first two PCs which accounted for 76% of the total variance. Some P. placenta – 16% ML, G. trabeum – 16% ML and 12% ML, and also some G. trabeum – 6% ML tracheid spectra separate along the first PC from those of non-degraded samples (Fig. 5A). However, also the variation within the non-degraded spruce tracheids was high in respect of this PC. The loading spectrum of PC1 (Fig. 5B) shows a huge maximum at 3348 cm−1 indicating that the O(3)H···O(5) intramolecular hydrogen bond was lower in these degraded tracheids than in non-degraded ones. The lateral O(3)H···O(5) hydrogen bond in β-1-4 linked wood polysaccharides is in proximity to the glycosidic bond, which was shown to be reduced during brown-rot degradation. Negative loadings at 3460 and 3404 cm−1 indicate an increase of free or only weakly H-bonded O–H groups of C(2) and C(6) of cellulose. These evolving O–H bands suggest the depolymerisation of the polysaccharides rather than their mineralisation in this stage of brown-rot, and may serve as the water bonding sites proposed by Winandy and Morrell [4]. A broad negative loading band at around 3312 cm−1 suggests a relative increase of intermolecular hydrogen between O(6)H and O(3) of the vicinal cellulose chain in cellulose crystals. This feature shows the relative resistance of crystalline domains over amorphous ones against brown-rot attack.

Fig. 5

PCA scores (A) and loadings plot (B) of spectra (2nd derivative, 3600–3100 cm−1) of tracheids. (Pp: P. placenta treated spruce, Gt: G. trabeum treated spruce; Spruce: non-degraded spruce wood).

Although the variation of the non-degraded samples was also high in respect of PC2, degraded samples tended to score positively compared to spruce – particularly those exhibiting pronounced degradation (>12% ML). The positive loading bands at the position of strongly H-bonded O–H vibrations (3376, 3344, ∼3324 to 3320 cm−1) as well as weakly H-bonded or free ones at 3456, 3428 and 3408 cm−1 indicated that loosening of interactions between polysaccharides took place in pronounced brown-rot stages. The number of free O–H groups of alcoholic and/or phenolic origin was higher in these samples indicated by negative loading bands at wavenumbers >3500 cm−1. Furthermore, negative loadings at 3224 and 3192 cm−1 could be indicative of increased phenolic moieties generated during the oxidative cleavage and demethoxylation of lignin [4], [29] and their strong intra- and intermolecular H-bonding with other phenols and alcoholic groups [30]. The negative loading at 3268 cm−1 (intermolecular H-bonded OH of cellulose Iβ) was a further indication for the relative stability of cellulose crystallites during brown-rot degradation.

Univariate tools to characterize brown-rot processes – IR crystallinity indices

Brown-rot processes reduce cellulose crystallinity [19]. However, less than one-third of the wood polysaccharides are crystalline. The rest are hemicelluloses, pectic substances, or amorphous and para-crystalline regions of the cellulose fibrils. If those amorphous polysaccharides are degraded in preference to crystalline cellulose, the overall crystallinity is expected to increase. Howell et al. [31], who investigated brown-rot of pine wood by X-ray diffraction analysis, found such a slight increase of wood crystallinity in samples up to a ML of about 50%.

Several attempts have been made to quantify the degree of crystallinity of pure celluloses using the intensities of certain bands in the IR spectra [12], [32], [33]. Schwanninger et al. [34] found the best representation for wood crystallinity using the ratios of the areas from 1400 to 1289 cm−1 and 1143 to 1089 cm−1 to the area below the band at 898 cm−1. As lignin contributes to the region between 1400 and 1289 cm−1, the authors subtracted a pure milled wood lignin spectrum before calculation. This procedure has the potential for error, since lignin modification takes place during brown-rot [29], [35].

To overcome the problem of overlapping bands, 2nd derivatives were calculated to obtain well resolved amplitude minima at 1508, 1368, 1108, and 896 cm−1 to which neither the lignin band at 1424 cm−1 nor the glucomannan band at 868 cm−1 contribute. Second derivative (SD) amplitude ratios were calculated from these minima.

Regarding the crystallinity indices SD1368/896 and SD1108/896 which were plotted vs. SD1508/1368, a measure for the relative lignin content [25], only a few spectra showed significantly lower or higher crystallinity than the average (Fig. 6A and B). Furthermore, tracheids with lower crystallinity according to the ratio SD1368/896 were not the same as those detected by the ratio SD1108/896. If all ML (at 16%) were caused by degradation of amorphous substances and no conversion of crystalline to amorphous cellulose took place, the overall cellulose crystallinity based on the wood substance would increase by approx. 20% rel. This difference would be too small to be detected amongst the highly variable non-degraded tracheids.

Fig. 6

2nd derivative (SD) amplitude ratios of tracheid spectra SD1368/896, SD1108/896, plotted versus SD1508/1368. Grey zones highlight the interval of the average ratio of non-degraded spruce tracheids ± standard deviations times 1 and times 2.

Characterising brown-rot processes on the cellular level – multivariate image analysis (MIA)

The aim of MIA was to detect characteristic changes during brown-rot discussed in the previous chapters within the FT-IR images of spruce wood sections. For that purpose, spectra from each pixel were analysed. The size of an FT-IR image pixel was 6.25 × 6.25 μm2. Therefore, typically 5–6 spectra were recorded over the average diameter of an early-wood tracheid (Fig. 1C). However, the true physical resolution is in the range of the wavelength of IR radiation (8.6 μm at 1160 cm−1). Consequently, each pixel spectrum has some contribution from adjacent pixels. These restrictions concerning spatial resolution lead to some contribution of the inner cell wall layers (S2, tertiary wall) to spectra on the tracheid boundaries, where the middle lamellae, P, and S1 cell wall layers are located. Furthermore, the IR radiation in transmission microscopy of radial wood sections also passes the middle lamellae, P, and S1 cell wall layers in the centres of tracheids, although the S2 layers make up approx. 80–90% of the wood cell wall material [16]. This setup of FT-IR microscopy thus allowed distinguishing spectra with a high contribution of S2 and spectra with a high contribution of S1, P and middle lamella (Fig. 1C).

For the first approach to MIA, spectral data of selected FT-IR images with decayed tracheids were unfolded and combined into one matrix of single pixel spectra (17,290) × IR fingerprint variables (1840–1096 and 1016–736 cm−1). A first PCA was calculated from this matrix. Spectra with high spectral residual or high leverage – spectra with very low or high total absorbance – were omitted for the calculation of a second PCA (14,944 pixel spectra in the same spectral range). The scores were then refolded to pseudo-colour score images (Scheme 1) to localise spectra of degraded spots.

A second approach to MIA classified pixel-spectra of all images into degraded and non-degraded to trace characteristic degradation markers of early brown-rot. PLS-DA models were built using the spectra from three native spruce images and early brown-rot spectra which clearly separated from the former in the PCA. The PLS-DA model was used to predict and assign the degradation state (0 – non-degraded to 1 – degraded and vice versa) to spectra of images not included in the regression model (Scheme 1).

PCA based image analysis

Only the scores plot of the PCA (Fig. 7) is presented here, as the loading spectra were very similar to those presented in Fig. 3B. Second derivative spectra from pixels with high lignin and low polysaccharide content (pronounced brown-rot) are separated from those with low lignin content and higher polysaccharides content along PC1 (41% of the data variance). Spectra from early degraded pixels separate on PC2 (17% of the data variance) according to the relative intensity of the band assigned to the glycosidic bond

Fig. 7

PCA scores plot of single pixel spectra (2nd derivative, 1840–1096 and 1016–736 cm−1) (P. placenta treated spruce, G. trabeum treated spruce; Spruce: non-degraded spruce wood).

The score values of each spectrum on PC1 and PC2 were plotted as score images to localise the features of degradation within selected details of thin sections. Fig. 8 shows CCD camera images, FT-IR absorbance images, the score images of PC1 and PC2, and images of the spectral residuals after two PCs. Yellow and red pixels were allocated to brown-rot degradation (score values exceeding + 0.05 on PC1 and PC2, respectively). In the residual images, pixels with spectral residuals exceeding the average residual by the factor three (>3 × 10−5) were highlighted in yellow to red to identify outliers.

The PC1 score images clearly demonstrated some degraded pixels in the topmost tracheids of a P. placenta (16% ML) thin section (Fig. 8B). The detail of a more severely degraded G. trabeum (16% ML) section (Fig. 8C) showed pronounced degradation. The lowermost row of pixels within the detail of the non-degraded section (Fig. 8A) also appeared degraded. However, the residuals image identifies these pixels as outliers and false-positives caused by the high IR absorbance of the latewood boundary.

The PC2 score image of the P. placenta detail visualised the early degradation pattern of the same topmost tracheids of the image (Fig. 8B). Some pixels of the G. trabeum image also showed some characteristics of early degradation, although most pixels were identified as strongly degraded by their PC1 scores.

Partial least squares discriminant analysis (PLS-DA)

The PCA of the single pixel spectra resulted in three groups of spectra: non-degraded, early brown-rot, and pronounced brown-rot (Fig. 7). The goal was to build a PLS-DA model capable detecting early brown-rot in spruce wood. Spectra of non-degraded spruce and spectra separating on PC2 (PC2 score > 0.05 and PC1 score < 0.05) were considered. Those 3514 spectra and 4882 non-degraded spectra were extracted from the PCA data matrix, before two PLS-DA models were built.

PLS-DA allowed for a sharp separation of the calibration samples on the first factor (81% explained Y-variance, Fig. 9A). The PLS factor plot explained which variables contained the class-discriminating information (Fig. 9B). Although most of the data variance was described by the first factor, a five factor model with a coefficient of determination was chosen. The two resulting PLS-R models were complementary and their regression coefficients differed only in their algebraic signs. Therefore, only one model is discussed in detail here. The most important variable was 1160 cm−1, which dominated the first PLS factor loading and the regression coefficients (Fig. 9B and C), indicating the degradation of glycosidic bonds as the spectral marker of early brown-rot.

Fig. 9

PLS scores plot (A), first PLS loading (B) and regression coefficients (Y = 1 setting for degraded pixels) (C) of the PLS models calculated for PLS-DA based MIA (2nd derivative, 1840–1096 and 1016–736 cm−1).

Predicted values of spruce spectra ranged from −0.38 to +0.35, those for brown-rot spectra from 0.5 to 1.54. The prediction error (RMSEP) was 0.123, suggesting that predicted values >0.60 (i.e. 0.35 + 2 × RMSEP) assign pixels affected by early brown-rot, and values <0.26 (i.e. 0.5 − 2 × RMSEP) assign non-degraded pixels. Intermediate values (0.26–0.60) cannot be assigned to either group.

Each FT-IR image was unfolded for the PLS-DA based MIA and subjected to the same data pre-processing (SNV, 2nd derivative). Then the same spectral range of the IR fingerprint was extracted. The prediction value and deviation matrices were then refolded and plotted as pseudo-colour images. All spectra of non-degraded spruce images showed some variation in their prediction values, which in most cases did not exceed 0.30. Deviation values generally were lower than 2 × RMSEP (0.246). The deviation in PLS prediction is a measure for the similarity of an unknown sample to the calibration samples (www.camo.com). This value can be used to identify false-positive results, if they occur. However, also spectra from samples with strong degradation pattern showed very high deviation (Fig. 10G2) indicating that this PLS-DA prediction model was limited to samples in early degradation stages.

Fig. 10

FT-IR spectral absorbance images (top), PLS-DA prediction (centre) and deviation (bottom) images of spruce wood thin sections. S1, S2: non-degraded; P1: P. placenta – 0% ML with pattern of incipient decay in cell wall regions close to middle lamellae; P2: P. placenta – 8% ML – pronounced decay; P3: P. placenta – 16% ML, uneven decay; G1: G. trabeum – 12% ML, uneven decay; G2: G. trabeum – 16% ML: pronounced decay; G3: G. trabeum – 16% ML – incipient decay. The colourmap indicates the assignment of degraded and non-degraded pixels in the prediction images.

In Fig. 10, FT-IR spectral absorbance images, PLS-DA prediction and deviation images of selected details of non-degraded spruce wood sections (S1, S2) and sections exhibiting different degrees of degradation are presented. No false-positives were observed in the non-degraded spruce images. The P. placenta section P1 from a sample without a detectable ML showed some degraded spots. A high number of other pixels in proximity to the former could not be identified as “non-degraded”. Most deviation values of these pixels were low. The affected pixels were located at the middle lamellae, P, and S1 cell wall layers, suggesting that brown-rot commences in those areas. Sections G3 (G. trabeum – 16% ML) and the one presented in Fig. 1D (P. placenta – 16% ML) showed very similar patterns.

Highly branched and less-ordered amorphous pectic substances [36] are the main constituents of the outer cell wall regions. These substances are most susceptible to biodegradation processes and are the first targets of brown-rot fungi. The proximity of these polysaccharides to highly concentrated lignin in these cell wall regions could be a further explanation that brown-rot first became evident in cell wall regions most distant from where fungal hyphae grow. Lignin has been attributed an important role for mediating Fenton reactions during brown-rot [22]. Thus, the findings here support the involvement of fungal degradation systems employing diffusible agents. Degradation processes in amorphous parts of S cell walls may also take place but become significant later. Blanchette and Abad [37] investigated birch wood degraded to 53% ML by the brown-rot Fomitopsis pinicola using scanning transmission electron microscopy and found that the highest fraction of hemicelluloses had been degraded in cell corners and middle lamellae. FT-IR microscopy provides direct evidence to support their findings in a sample exhibiting non-detectable ML as well as a pattern of incipient decay also unevenly degraded samples with higher ML. Pronounced degradation is characterised by compositional changes within the wood cell walls, indicating that hydrolysing enzymes gained access to their substrates, or that the saccharides of lower molecular mass gained access to degrading enzymes. Only small oligosaccharides are able to diffuse in a non-degraded wood cell wall [38].

Most of the degraded spots in the degraded section P2 (Fig. 10; P. placenta – 8% ML) could be allocated to middle lamellae/primary cell walls/S1 cell walls, the least affected being the S2 cell walls. P3 (P. placenta – 16% ML) and G1 (Fig. 10, G. trabeum – 12% ML) were typical examples for very uneven decay within the same section. Most spectra of sections exhibiting a pronounced degradation pattern (G2 – G. trabeum – 16% ML) were identified as degraded.

Conclusions

FT-IR microscopy provided direct evidence for incipient decay processes in wood cell walls of intact early wood tracheids. Ray cells did not contribute to the FT-IR spectra, or was there any disturbing contamination from fungal biomass detected. This allowed for a selective and detailed analysis of the IR fingerprint and of the O–H region of the FT-IR spectra allowing for structural characterisation of brown-rot decay processes on the cellular and on the sub-cellular level, although with some limitations concerning the spatial resolution. The characteristic disappearance of the IR band at 1160 cm−1 in relation to other polysaccharide-derived IR bands could serve as a reliable marker for incipient brown-rot decay of softwood in general, because this feature is supposed to be related to the average degree of polymerisation of the polysaccharides and therefore shows less variation in spruce wood than does its composition.

Multivariate image analysis can be used to assign fungal degradation processes to distinct cell wall positions. Spatially resolved spectra of unevenly degraded wood samples allow for the detection of extreme situations within the sample. Therefore FT-IR imaging may also serve as a powerful technique for wood decay assessment. To obtain universal and more robust classification models, the method would have to be extended to samples from spruce wood of different sources. Uneven degradation, however, may impede the detection of ongoing decay processes if the number of investigated FT-IR images is too low.