Introduction

Kinkarakawa-gami represents an important example of Japanese art and its mixture within the European culture between the XIX and XX centuries. This manufacture derived from the importation in Japan of the so-called Kinkarakawa (i.e., foreign gilt leather) by the Dutch East India Company, which was one of the few foreign companies allowed to commercially trade with this eastern country up to 18541. Subsequently, the opening of Japan to the western world encouraged the exportation of authentic original Japanese goods, among which Kinkarakawa-gami (i.e., gilt leather-like wallpapers) gained international relevance and recognition, mostly with British, Dutch and German companies1,2.

Several European sources of the XIX century documented the traditional production stream of Kinkarakawa-gami, identifying four steps: (1) hand-making of the leather-like paper, (2) its embossing with a chosen pattern, (3) gilding of the paper with metal powders or foils, and (4) its final decoration3,4,5,6. Historically, three periods of Kinkarakawa-gami’s manufacture are distinguished: (1) the Japonist Enlightenment (1873–1884), during which the wallpapers were entirely hand-made, properly depicting Japanese patterns and featuring an “oily smell”7, which did not encounter the European consumer taste; (2) the period of industrial production (1882–1889), when several British factories were established on Japanese soil, leading to faster yet traditional manufacture, as well as a change in the wallpaper design; and (3) the mass-produced commodity period (1890–1905), where the patterns were directly chosen by British designers and cheap industrial processes were applied to answer to the large demand of European customers1, however resulting in the end—or almost all—of the Kinkarakawa-gami art2.

The uniqueness of these Japanese wallpapers relies on their limited presence in Europe, therefore physical–chemical studies aimed to unveil the features of these artefacts are still missing, being of paramount importance to develop successful and sound conservation-restoration strategies. This aspect significantly gains importance since, to the best of the authors’ knowledge, Kinkarakawa-gami belonging to the Vincenzo Ragusa and O’Tama Kiyohara Collection of Palermo (patrimony of the Liceo Artistico founded in the XIX century by the sculptor Vincenzo Ragusa) represents the first finding of this form of Japanese fine art in Italy. Thus, four wallpapers of this collection were studied through a multidisciplinary approach aimed to overview their state of conservation for their future restoration, therefore enhancing the artistic value of these artefacts. The wallpaper manufacture was evaluated through a standard operating procedure involving non-invasive and non-destructive physical–chemical techniques—that allow to perform an extensive sampling of the artefacts without altering them or their conservation state8—and microscopy observations, focusing on the (1) paper composition, (2) metallic layer on the recto of these artefacts, identification of (3) inorganic pigments, (4) organic pigments, and (5) substances (e.g., lacquers) exploited as binders for the pigments9. Physical–chemical results were validated through in blind statistical analyses, which, by identifying homogeneous clusters of curves with respect to their shape, further unveiled important differences and similarities among the four artefacts. Finally, the cultivable microbiota populating these wallpapers and their potential interaction with both inorganic and organic substrates were critically discussed to better understand how to properly restore and preserve these Kinkarakawa-gami through time.

Results and discussion

Paper composition

The support material used by Japanese artisans to produce the Kinkarakawa-gami wallpapers (henceforth named as INV_11, INV_13, INV_15, and INV_20; Fig. 1, Table 1) derived from either eastern plants or wood pulp, as indicated by the ubiquitous detection, through micro Attenuated Total Reflectance-Fourier Transform Infrared (µATR-FTIR) spectroscopy, of vibrational modes typical of lignin and cellulose (Fig. 2, Table 2, Supplementary Tables 1–4). The chemical similarity with wood pulp was also suggested by a statistical analysis performed on the µATR-FTIR spectra, which was carried out focusing on the 3770–2750 cm−1 and 1840–720 cm−1 intervals (Supplementary Fig. 1), where the shape of the curves gave information regarding the essential modes of variability among the spectra and their clustering. The cluster analysis showed the hierarchical relationship between the spectra (Figs. 3a, 4a), based on which the optimal number of clusters and the spectra allocation were determined (Figs. 3b, c, 4b, c). Principal modes of variations were derived to identify cluster effects on the shape of the curves, showing that the clusters were correctly visualized in an R2 space, as the first two functional Principal Components (PCs) described the 90.8% and 82.6% of the total variability for the first and second interval, respectively (Figs. 3d, 4d). Specifically, cluster 2 (first interval) and 1 (second interval) grouped the spectra collected for wallpapers’ verso, whose high values of silhouette (average si = 0.73 and 0.76) and distribution in the R2 space through PCA assessed their well-identification within the cluster (Fig. 3c–e, 4c–e). The large PC2 value of the first IR interval of INV_15_2 (Fig. 3d) indicated its greater variability than the other verso spectra, likely due to the presence of multiple –OH stretching modes characteristic of cellulose (Fig. 2c; Supplementary Table 3). The Modified Band Depth (MBD) function was used to identify a representative (i.e., deepest) curve for the clusters (Figs. 3f, 4f) describing the shape of the spectra and arrange them by resemblance from the centre (highest depth) outwards (lowest depth) (Supplementary Fig. 2, 3). INV_13_2 was the deepest curve of cluster 2 (first interval) and the other verso spectra featured high depth values, constituting the core (70%) of the cluster (Supplementary Fig. 2b). An analogous situation was observed for the second interval, although the most representative curve was INV_13_9.1 (Fig. 4f), confirming the high degree of similarity of the wallpapers’ verso. Moreover, morphological and structural features of the wallpapers’ verso were studied through Confocal Laser Scanning Microscopy (CLSM) (Fig. 5), revealing a strong fluorescence signal in the visible-light spectrum, and a tangled matrix of organic fibres with a lateral width between 10 and 20 µm for all the samples (Fig. 5a, Supplementary Video 1), which, along with the cellulose and lignin contributions detected within µATR-FTIR spectra (Supplementary Table 1–4), suggested the vegetable origin of the paper. However, it cannot be unequivocally indicated whether the paper was hand-made or made out of mechanical wood pulp1, although the detection of silicon (Si) traces by X-Ray Fluorescence (XRF) spectroscopy in the verso sampling points (Fig. 6a, b) could indicate the use of sandstone grinders for the mechanical generation of the pulp10. Further, the detection of Si, calcium (Ca) (Fig. 6a, b), and IR absorption bands typical of silicates or carbonates (Supplementary Table 1–4) might infer the use of quartz, calcite, or chalk as fillers11.

Figure 1
figure 1

Kinkarakawa-gami wallpapers. In (a) is depicted the verso of one of the wallpapers and the points 1 (written portion) and 2 (non-written portion) that were chosen—for all the artefacts—for the analyses, while the wallpapers’ recto and the investigated sampling points (numbers) are represented in (b) INV_11 (3–10), (c) INV_13 (3–9), (d) INV_15 (3–4), and (e) INV_20 (3–5).

Full size image
Table 1 Description of the sampling points of the four analysed wallpapers.
Full size table
Figure 2
figure 2

µATR-FTIR spectra performed on the wallpapers. The spectra were collected on the selected sampling points (as indicated in the legend) of (a) INV_11, (b) INV_13, (c) INV_15, and (d) INV_20. For clarity, the µATR-FTIR spectra were offset of 0.10 arbitrary units (a.u.).

Full size image
Table 2 Organic and inorganic substances detected through μATR-FTIR performed on INV_11, INV_13, INV_15, and INV_20.
Full size table
Figure 3
figure 3

Statistical analysis performed on µATR-FTIR spectra for the 3770–2750 cm−1 interval. The hierarchical relationship between the spectra is represented by the dendrogram depicted in (a), while (b, c) show the estimation of the number of clusters using the average silhouette width and the obtained silhouette of clusters, respectively. PCA results, µATR-FTIR spectra, and the most representative (i.e., deepest) curve for each identified cluster (highlighted with different colours) are illustrated in (d–f). For clarity, the most external spectra of the clusters are underlined in (d) with the grey colour.

Full size image
Figure 4
figure 4

Statistical analysis performed on µATR-FTIR spectra for the 1840–719 cm−1 interval. The hierarchical relationship between the spectra is represented by the dendrogram depicted in (a), while (b, c) show the estimation of the number of clusters using the average silhouette width and the obtained silhouette of clusters, respectively. PCA results, µATR-FTIR spectra, and the most representative (i.e., deepest) curve for each identified cluster (highlighted with different colours) are illustrated in (d–f). For clarity, the most external spectra of the clusters are underlined in (d) with the grey colour.

Full size image
Figure 5
figure 5

Fluorescence microscopy of Kinkarakawa-gami wallpapers. CLSM imaging of representative portions of INV_15 on both verso and recto are illustrated in (A, B) respectively, while (C) depicts a representation of the 3D reconstruction performed on INV_15 recto, whose multilayer structure is quantitatively reported in depth-colour code (analysis depth 75 µm) in (D). Scale bars are 100 µm.

Full size image
Figure 6
figure 6

Elemental composition and distribution of Kinkarakawa-gami wallpapers obtained through XRF spectroscopy. XRF results are depicted by comparing artefacts’ sampling points collected for (a) verso (INV_11_2, INV_13_2, INV_15_2, INV_20_2), (b) written portions of verso (INV_11_1, INV_13_1, INV_15_1, INV_20_1) and (c) recto’s background (INV_11_5, INV_11_7, INV_13_3, INV_15_3, INV_20_5), as well as (d) red (INV_11_6, INV_15_4), (e) green (INV_11_3, INV_11_8, INV_11_9, INV_13_4, INV_13_8, INV_20_3), (f) brown (INV_11_4, INV_11_5, INV_13_6, INV_13_7, INV_13_9), and (g) black (INV_11_10) drawings on the wallpapers’ recto. Elemental compositions of wallpaper sampling points are reported in terms of Net Area (× 104 a.u.), which represents the integral X-ray emission intensity characteristic of each recognized element.

Full size image

Traces of Sulphur (S) and manganese (Mn) were observed on the wallpapers’ verso, although S was solely recognized in the not-written areas, except for INV_15, while a higher distribution of Ca and iron (Fe) was seen in this side as compared to the recto (Fig. 6a, b, Supplementary Fig. 4a). Since these two elements, along with lead (Pb), were preponderant in the written portions of INV_11 and INV_13 (Fig. 6b, Supplementary Fig. 4a), they may be components of the ink (e.g., iron gall ink) used for the selvages12, which are a trademark of Kinkarakawa-gami1. This conclusion cannot be drawn for INV_15_2 and INV_20_2, due to the strong presence of arsenic (As) and mercury (Hg) (Supplementary Fig. 4a), deriving from the coloured areas of the recto (Fig. 6c, Supplementary Fig. 5b) and/or a metal diffusion from one overlapping wallpaper to another, which may have hidden Ca and Fe signals.

Gilding of the paper with metals

Tin (Sn) was ubiquitously present in the wallpapers, revealing a greater signal distribution on their recto than their verso (Fig. 6), which is coherent with the gilding technique used by Japanese artisans, who covered the paper surface with a lacquer mordant and a metal foil (i.e., silver, gold or tin) and adhered to the embossed pattern7. In this regard, CLSM imaging of the wallpapers’ recto showed an amorphous and heterogeneous surface attributable to the artefact’s pattern, the inorganic pigments, and/or organic pigments (Fig. 5b). To better validate whether the wallpapers featured a multilayer structure as described in the literature1,2, a 3D reconstruction of a portion of INV_15 recto only partially covered by the motif pigmentation was performed (Supplementary Video 2). As a result, two distinct layers were detected: the top layer displayed the amorphous structure characterising the pigmentation, while the bottom layer featured an underlying matrix of fibres resembling that of the verso (Fig. 5c, d, Supplementary Video 2). Since the gilding practice was first carried out by Rottmann, Strome & Co factory in Yokohama (1882) yet abandoned from 1890 onwards2, the presence of multilayers containing Sn within the wallpapers suggested their belonging to the second period of Kinkarakawa-gami production.

Kinkarakawa-gami’s decoration

Despite the published literature on Kinkarakawa-gami art, indications regarding inorganic and organic pigments, as well as binders used for artefacts’ decoration is not reported yet, hence constituting a gap to be filled to enable their conservation-restoration.

Inorganic pigments

XRF spectroscopy revealed the recurrent presence in the artefacts of Si, S, Fe, Mn, Zn, As, whereas Hg, copper (Cu), chromium (Cr), and Pb were detected in specific coloured portions (Fig. 6), indicating the use of several inorganic pigments for the decoration. Although most of these pigments have strong IR contributions below 600 cm−113, some absorption bands attributable to the yellow pigment orpiment (As2S3, Sekio; 1035, 913, and 820 cm−1) and its degradation (-AsOFe stretching mode 830–840 cm−1) were frequently found within the µATR-FTIR spectra (Table 2, Supplementary Table 1–4), agreeing with the detection of As and S through XRF spectroscopy (Fig. 6). Since up to 1885 the wallpaper production did not involve As-compounds2,6, these results suggested artefacts’ manufacture between 1886 and 1889, which is a rather tight timeline unlikely to be achieved through other techniques (e.g., Accelerator Mass Spectroscopy (AMS)-radiocarbon dating featuring ± 20 years of experimental error)14.

Fe ubiquity on the recto’s background (Supplementary Fig. 4b) may instead indicate the application of natural iron oxide pigments, such as yellow—Odo—, brown—Taisha—and red—Kincha—ochres15, to obtain gold-bronze shades, which is another distinctive feature of this Japanese art7. Moreover, these pigments contain silicates and manganese dioxide (MnO2) as either components or impurities16, supporting the frequent detection of Si and Mn in traces (Fig. 6). An additional trace element observed in INV_11_7 and INV_13_3 was S (Supplementary Fig. 4b), whose contribution might rely on the use of tin sulphide (SnS2), which was known in Asia since 300 BC17, as a bronze-gold pigment for these backgrounds. The variability of the elemental composition in recto’s backgrounds was inferred by the detection of Zn (INV_11_5, INV_11_7, and INV_15_3), As (INV_15_3 and INV_20_5), Pb (INV_11_5), or Hg (INV_15_3; Fig. 6c). In this regard, the simultaneous presence of S and Hg (INV_15_3), or S and As (INV_20_5; Supplementary Fig. 4b) suggested that red mercury- (HgS; cinnabar or Shu) and yellow/red arsenic-sulphides (Sekio or red realgar—Yuo, AsS or As4S4) were the main pigments of INV_15 and INV_20 recto’s backgrounds15. As and Zn traces in INV_15_3 could be also linked to Shu, which generally contains these elements as impurities17. Furthermore, Zn-based paint driers may have been applied to favour the pigments’ deposition and solidification on the tin foil18, while Chinese white (zinc oxide; ZnO) and zinc sulphide (ZnS) could be used as white pigments19. To improve the brightness of the latter, artisans even mixed them with lead white [basic lead carbonate; 2PbCO3·Pb(OH)2]20, justifying the Pb contribution observed in INV_11_5 (Supplementary Fig. 4b).

The painted areas of the wallpapers’ recto were deeply analysed focusing on the most representative colorations (i.e., red, green, brown, and black). Similar chemical compositions were observed for the red colour in INV_11_6 and INV_15_4, where the featuring elements were Hg and S, indicating the utilization of Shu15,17, while traces of Si were detected only in INV_11_6 (Fig. 6d, Supplementary Fig. 5a). Moreover, the large signal of Hg observed in both INV_11_6 and INV_15_4 might be responsible for the Sn underestimation, as the high atomic number of Hg can mask Sn contribution. The green-coloured areas of INV_11, INV_13, and INV_20 displayed instead a high variability in elemental composition (Fig. 6e), inferring that these portions could contain multiple pigments. Particularly, Pb was present only in INV_11, whose great variability in composition was further indicated by the detection of Cr and Cu (INV_11_3; Supplementary Fig. 5b), typical components of green pigments, as chromium oxide (CrO2), malachite green [Cu2(CO3)(OH)2], Verdigris [Cu2(CH3COO)2Cu(OH)2)], Scheele green (a mixture of CuHAsO3 and CuAs2O3), or emerald green [Cu(CH3COO)23Cu(AsO2)2]17. The presence of Pb may be instead ascribed to (1) lead white for the outline of the leaves15, (2) lead–tin yellow21, or (3) yellow lead oxide (PbO)22 to achieve diverse green shades. Regardless, all the green areas exhibited S and As, whose distribution varied depending on the different colour shades, as noted in INV_11_3, INV_11_9, INV_13_4, and INV_13_8 (Supplementary Fig. 5b). Thus, Sekio15 seemed to be constantly present in the green areas23, suggesting (1) a combined use of Verdigris, Scheele green, or emerald green with sulphide compounds in INV_11_3 to improve both brightness and durability of the pigments24, as well as (2) the addition of Sekio to an organic dark (blue/purple) pigment to obtain the green colour15,25,26. Except for INV_20, the intense Fe signal characterising the recto’s green-painted sections (Fig. 6e) could imply the combination of green/yellow inorganic pigments with natural ochres15, as also supported by the detection of Si, Mn, and Zn in traces (Fig. 6e, Supplementary Fig. 5b). Nevertheless, µATR-FTIR spectra of INV_20_3 exhibited –SO4 asymmetric (1120–1040 cm−1 and 827 cm−1) and symmetric (983 cm−1) stretching, –FeOSOFe vibration (955 cm−1), as well as –FeO stretching (830 cm−1) characteristic of the Fe-based mordant green vitriol (FeSO4·xH2O) (Table 2, Supplementary Table 4), which was likely used in this wallpaper. Several brown drawings were part of INV_11 and INV_13 (Table 1), where Fe detection (Fig. 6f, Supplementary Fig. 5c) might infer to the utilization of Taisha7, also supported by the low content of Si, Mn, and Zn, which confer diverse yellow–brown and red-brown shades to the pigment16. Additionally, lead white and/or lead oxide15,22 could be present in INV_11_4 and INV_13_5, as indicated by Pb traces, while INV_13_6 and INV_13_9 showed a relatively high distribution of both As and S (Fig. 6f) attributable to the addition of Sekio or Yuo to get unique brown shades23,26. Finally, the black colour was only present in INV_11_10, whose elemental analysis revealed traces of S and the three ubiquitous elements Ca, Fe, and Sn (Fig. 6g), implying the use of an organic black pigment for this colour15.

Organic pigments

Several IR absorption bands typical of the natural purple-pigment Indigo were detected in the green areas of INV_11_3 and INV_13_8 (Fig. 2a, b, Table 2, Supplementary Table 1–2), whose similarity was confirmed by the partitioning of their spectra in cluster 4 (average si = 0.56; first interval) or 2 (average si = 0.70; second interval), as well as their PC2 values (Figs. 3c–e, 4c–e). This observation agrees with the practice of Japanese artists, from the late Edo period (1840–1860) onwards, to mix Indigo with lime, ashes, and Sekio as an alternative to malachite green16,25,26. On the other hand, the INV_20_3 spectrum, belonging to cluster 1 for both the first (average si = 0.83) and second (average si = 0.76) intervals (Figs. 3c–e, 4c–e), suggested that diverse techniques could have been used to obtain the green colour in INV_20 as compared to INV_11 and INV_13. Although INV_20_3 featured some Indigo’s IR contributions, the detection of vibrational modes typical of -SO4 and -FeO groups (Supplementary Table 4) indicated that iron sulphate mordants, such as green vitriol, were supplied to Indigo and Sekio obtaining the green colour. Indeed, a similar procedure was developed by English artisans, who added pearl ashes, lime, Sekio, and green vitriol to Indigo syrup, favouring both the precipitation of Fe protoxide by the lime, as well as the deoxygenation of Indigo, which acquired a dark green shade27,28. Besides, only a few of Indigo’s IR contributions were identified in red and brown drawings (Table 1, Supplementary Tables 1–2, 4), reasonably due to (1) Indigo diffusion from adjacent areas containing the organic pigment or (2) its utilization to obtain diverse colour shades. The µATR-FTIR spectra collected for brown and red drawings of INV_11 and INV_13 (Table 1) showed a high degree of similarity, as they all clustered with INV_11_3 and INV_13_8 spectra for both the 3770–2750 cm−1 and 1840–719 cm−1 intervals (Figs. 3c–e, 4c–e). INV_13_9 constituted the only exception, as it was grouped in cluster 1 (Figs. 3c, 4c), underlining that diverse organic substances, alongside the inorganic Sekio (Supplementary Fig. 5c), may be used for this brown sampling point. Since Indigo features low IR contributions in the 3770–2750 cm−1 interval29, indications regarding its presence within red and brown drawings can be better obtained by focusing on the MBD-based analysis of the second IR interval, where INV_13_5.1 was the deepest curve (Fig. 4f), while INV_11_6, INV_15_4, and INV_20_4 were the most external and variable of the cluster (Fig. 4d, Supplementary Fig. 3b). Hence, it is reasonable to suggest that Indigo diffusion was responsible for IR contributions in INV_11_6 and INV_20_4, while it was likely used for INV_11_4, INV_13_3, and INV_13_5, being part of the core cluster (Supplementary Fig. 3b). Moreover, INV_15_4 partitioning within cluster 2 for the 1840–719 cm−1 interval (Fig. 4c–e) may be due to the presence of other organic substances having absorption bands in common with Indigo. Indeed, any of typical IR contributions of this organic pigment were observed in INV_15_4 μATR-FTIR spectrum (Supplementary Table 3), which also featured a low depth value, as well as a high PC2 score, with respect to the other spectra clustering together (Fig. 4d, Supplementary Fig. 5b).

Organic binders

The identification of inorganic pigments highlighted the need for Japanese artisans to use binders (e.g., proteinaceous binders, vegetable oils, waxes, and lacquers) for their painting15. Although the type of these substances is generally determined through FTIR, the wallpapers’ complexity impaired their univocal identification, since they had similar IR absorption bands and their specific vibrational modes often overlapped with each other (Supplementary Tables 1–4)30,31. In this regard, the clustering analysis in the 1840–719 cm−1 interval revealed that spectra featuring Indigo’s IR contributions grouped in cluster 2, while cluster 1 featured the verso spectra and those where the organic pigment was absent or less present, in which overlapping vibrational modes of cellulose, lignin, silicates, carbonates, and organic binders were ubiquitous (Fig. 4c, Table 2). Similarly, although four clusters were identified for the µATR-FTIR spectra in the 3770–2750 cm−1 interval (Fig. 3c–e), several absorption bands attributable to these substances were detected (Supplementary Tables 1–4). Nevertheless, the type of organic substances was presumed, when possible, by comparing their IR absorption bands with those characterising the organic binders (Supplementary Tables 1–4). Thus, vibrational modes typical of proteinaceous binders (INV_15) or waxes (INV_13_9 and INV_20) were detected (Supplementary Tables 3–4), which is in agreement with the Japanese traditional use of animal glue (nikawa), gelatine cubes, and beeswax as binders15,25. Since IR absorption bands distinctive of proteinaceous binders or waxes are more present in the 3770–2750 cm−1 interval30,31, the partitioning of µATR-FTIR spectra of INV_15 recto in cluster 3 (average si = 0.66), INV_13_9 and INV_20_3 in cluster 1 (Fig. 3c, d) statistically supported these observations. Particularly, the MBD-based analysis of cluster 1 revealed a high degree of similarity between these µATR-FTIR spectra, representing both the deepest curve (INV_13_9) and the core (70%) of the cluster itself, while INV_11_4.1 was the most external spectrum (Fig. 3f, Supplementary Fig. 2a), where IR absorption bands typical of waxes were not detected (Supplementary Table 1). Moreover, although vibrational modes characterising waxes are less present in the 1840–719 cm−1 interval30, the comparable PC2 values of INV_13_9, INV_13_9.1, INV_13_9.2, and INV_20_3 spectra highlighted their higher variability, yet intrinsic similarity, than other curves belonging to cluster 1(Fig. 4d).

Several µATR-FTIR spectra displayed also vibrational modes distinctive of organic vegetable lacquers (Table 2, Supplementary Tables 1–4), which grouped in cluster 3 in the 3770–2750 cm−1 interval (Fig. 3c–e). In this regard, the so-called lacquering procedure for the wallpapers’ production involved the mixing of plant-extracted lacquers—likely the Urushi one, based on the Japanese tradition32—with Odo, Taisha, and Kincha to confer the yellow-gold background characteristic of the Kinkarakawa-gami’s recto7. Indeed, the Urushi lacquer features a strong IR contribution around 3400 cm−1 attributable to –OH stretching32, which, alongside the vibrational modes of proteinaceous binders, was a distinctive trait of the µATR-FTIR spectra belonging to cluster 3 (first interval; Fig. 3c–e). In line with this, INV_15_4 resulted the deepest curve identified for this cluster (Fig. 3f), while INV_13_5, where the –OH stretching of the Urushi lacquer was absent (Supplementary Table 2), was the most external one (Fig. 3d, Supplementary Fig. 2a).

Isolation and identification of microorganisms populating the wallpapers

A microbiological evaluation of specific areas of these artefacts was carried out to assess the role of microorganisms in either the degradation or damage prevention of the wallpapers, as well as the hindrance in identifying IR contribution of organic substances due to the interference derived from macromolecules’ vibrational modes33,34. A wide array of microorganisms populates different works of art, as their organic and inorganic substances constitute carbon and essential element sources for bacteria and fungi34. The microbial presence on artistic items is also favoured by environmental factors (e.g., low ventilation, high humidity) and the poor state of conservation of the artefacts, which allows the microbial growth under oligotrophic conditions34. Particularly, microorganisms tend to irreversibly attach to the artefact surface, forming communities defined as a biofilm, where its complex hydrogel matrix—made of proteins, lipids, and polysaccharides—confers to bacteria and fungi protection from external factors35. Additionally, less than 10% of microorganisms populating different niches can be cultured through standard procedures34, making it almost impossible to entirely identify the microbial community associated with artistic items. Several bacteria and fungi (Table 3) were isolated from sampling points of INV_11, INV_13, and INV_15, while any cultivable microorganism was not retrieved from INV_20. Bacterial isolates were identified as part of the Firmicutes (i.e., Bacillus coreaensis, B. pocheonensis, B. onubensis, Staphylococcus capitis, and S. epidermidis) and Actinobacteria [i.e., Kocuria rizophila (former Micrococcus luteus) and Micromonospora chokoriensis] phyla, while all the detected fungi belonged to the Ascomycetes phylum, although four genera were observed (Table 3). These results are in line with the studies reporting the isolation of cultivable microorganisms from wall or easel paintings34,36.

Table 3 Identification of bacterial and fungal strains isolated from the four Kinkarakawa-gami analysed.
Full size table

The microbial isolates C4(11)3, C4(13)1, C7(11)1, and C4(11)2 closely related to B. coreaensis, B. pocheonensis, M. chokoriensis, and Cladosporium cf. ramotenellum respectively (Table 3), and autochthonous of Kinkarakawa-gami, are commonly found in several environmental matrices of Korea, China, Thailand, and Japan37,38,39,40, confirming the eastern provenance of these wallpapers. Besides, these microorganisms hold enzymatic assets (i.e., xylanases and laccases) that enable them to degrade and depolymerize (1) lignin and cellulose compounds37,40,41,42,43,44, (2) lacquers45,46,47, and (3) organic pigments46,48. Indeed, Bacillus, Staphylococcus, and Micrococcus spp. produce extracellular xylanases42,43,49, which efficiently hydrolyse lignocellulosic material and the Indigo pigment42,43, a biotic aspect that supports their presence on INV_11_1, INV_11_4, INV_13_5, and INV_13_4 (Table 3). Laccases are instead Cu-polyphenol oxidases firstly identified in the exudates of the Japanese Rhus verniczfera plant50 from which the Urushi lacquer is extracted32, yet also recognized as enzymatic catalyst of most fungal strains47. Since these enzymes are responsible for the degradation of phenol substrates45,47, the partial identification of vibrational modes typical of urushiol polymerization or Indigo (Supplementary Tables 1–4) could be ascribed to the biotic deterioration of both Urushi lacquer and organic pigment, as Japanese lacquer IR contributions were found within sampling points where fungal strains were isolated (Tables 2, 3). A similar conclusion can be made for C4(11)3 and C4(11)2 isolates phylogenetically related to B. coreaensis and C. ramotenellum respectively, whose production and secretion of xylanases and laccases is one of their distinctive metabolic traits37,40,44. Further, Bacillus [C4(13)2], Staphylococci [C5(13)2 and C4(15)1)], and Micrococci [C2(11)1 and C5(13)1] species are among the most persistent strains capable of metabolizing honey and beeswax51, while M. chokoriensis [C7(11)1] and C. ramotenellum [C4(11)2] are proficient in degrading gelatine and starch, which were two of the most used binders in Japanese tradition15,27.

The presence of microorganisms can also be linked to their tolerance and/or resistance towards a broad spectrum of metal and metalloid compounds52, which are the main components of inorganic pigments. For instance, bacteria belonging to the Firmicutes and Actinobacteria phyla can detoxify their surrounding environment from toxic metals or metalloids, even using them as a terminal electron acceptor to produce energy52,53. Bacillus, Staphylococcus, and Micrococcus spp. are overall present in paintings where carbonates (chalk—CaCO3) and silicates (quartz—SiO2) are abundant54,55, due to their ability to overcome the challenge deriving from these minerals. Bacillus spp. are also highly tolerant towards Pb-56 and Fe-containing compounds57, being able to transform lead acetate [Pb(CH3CO2)2]56, hematite (Fe2O3) and its hydrated forms58, as well as to oxidize Fe(II) to Fe(III), as in the case of B. pocheonensis59. Analogously, S. epidermidis, M. chokoriensis, and K. rizophila strains were studied for their resistance against Pb-containing compounds60,61,62, while the fungal strain Hirsutella spp. showed tolerance to Ca- and Fe-sulphates63. Thus, the elemental composition of INV_11 and INV_13 (Fig. 6) justified the presence of microorganisms (Table 3) that hold metabolic traits allowing them to survive on metal-rich wallpapers. Although INV_15_4 displayed a high amount of Shu (HgS) (Supplementary Fig. 5a), one bacterial [C4(15)1] and two fungal [C4(15)2 and C4(15)3] strains related to S. epidermidis, Penicillium georgiense, and Aspergillus costaricaensis were isolated (Table 3), highlighting their great tolerance towards Hg-containing compounds64,65. Indeed, these fungal strains were responsible, among others, for the darkening of cinnabar (Shu), as reported elsewhere66. The isolation of S. epidermidis and Hirsutella strains could also derive from anthropogenic/environmental contamination, as they are human and nematode pathogens respectively63,64, reflecting the artefacts’ poor state of conservation.

Besides, the local heterotrophic microflora (i.e., Bacillus, Micromonospora, and Micrococcus species) can produce secondary metabolites with antimicrobial properties as a defence mechanism under stress conditions34,67,68,69. Thus, Firmicutes and Actinobacteria could act as biocontrol agents for the preservation of cultural heritage34. These observations, along with the advanced state of deterioration that was macroscopically visible for INV_20, may indicate the key role played by bacteria belonging to Bacillus, Micromonospora, and Micrococcus genera in controlling and preventing the further degradation of the colonized wallpapers; indeed, INV_20 was the only artefact from which any cultivable microorganism was not isolated.

Conclusion

This multidisciplinary study allowed to unveil physical–chemical features regarding the composition, manufacture, and dating of the collectively important cultural heritage represented by Kinkarakawa-gami works of art. The experimental evidence gathered was corroborated through the uncommon yet resourceful and innovative in blind functional data statistical analysis. Indeed, given the complexity of the studied wallpapers in terms of IR vibrational modes ascribed to (1) substances used for their fabrication, (2 physical–chemical degradation products, and (3) the presence of a microbiota, this statistical approach has proved greatly helpful to identify and confirm trends, differences, and similarities observed among the four Kinkarakawa-gami artefacts. Microbiological investigations supported the Eastern provenance of these wallpapers, however, whether the cultivable microbes act as deteriogen or biocontrol agents is yet to be defined; hence, DNA sequencing-based technology (e.g., study of the microbiome) represents the new frontier to unveil the identity of uncultivable microorganisms and, alongside physical–chemical characterisation, will improve the development of innovative, promising, and eco-friendly restoration strategies for the conservation of cultural heritage.

Experimental section

Materials

The four wallpapers here studied belong to the V. Ragusa-O’T. Kiyohara collection of Palermo (Italy). Given the complexity and richness of these artefacts in terms of details and depicted colours, an extensive sampling of the wallpapers (Table 1) was performed to thoroughly analyse them.

Tryptic soy, malt extract, and agar technical were purchased from Sigma-Aldrich® (Milan, Italy), while QIAquick PCR purification kit was obtained from QIAGEN (Milan, Italy).

X-Ray fluorescence (XRF) spectroscopy

A Tracer III sd Bruker AXS (Bruker, UK) equipped with Rhode anode and working at 40 kV and 11 µA was exploited for XRF analyses, whose acquisition time was 30 s. Element identification and XRF spectra analysis was performed by using ARTAX® software, which was provided with the instrument, while R 3.6.1 and OriginPro® 2016 software were used for the representation. Elemental compositions of wallpaper sampling points are reported in terms of Net Area (× 104 a.u.), which represents the integral intensity of X-ray emission characteristic of each element obtained after performing the Bayes deconvolution and deduction of the background intensity70. These results are to be considered as a qualitative estimation regarding the presence and abundance of diverse elements within the chosen sampling points, as any appropriate standard was not used to determine the concentration of each element.

Attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy

ATR-FTIR spectra were recorder by using a µFTIR Lumos (Bruker, UK) equipped with a Platinum ATR and an IR microscope featuring 0.1 µm as lateral resolution. The spectra were collected in the 4000–600 cm−1 range, with a resolution of 2 cm−1 and 60 scans per each sampling point, and they were subsequently analysed through the software OPUS(7.5)®, which was provided with the instrument, as well as OriginPro® 2016.

Statistical analyses of µATR-FTIR spectra

Based on the structure of µATR-FTIR spectra, they can be assimilated to complex sets of data (i.e., functions, where each spectrum corresponds to a distinct function) varying over a continuum (i.e., wavenumber range) that, taken together, can be considered as single curves. Hence, in blind functional data analysis (FDA) was applied as the statistical methodology on these spectra71 that were split, based on the wavenumber range acquisition, in 3770–2750 cm−1 and 1840–719 cm−1 intervals, which were singularly analysed by performing a hierarchical clustering of µATR-FTIR curves with the respect to their shape. The quantification of each curve’s cohesion to its own cluster as compared to the separation from the other clusters was obtained through a silhouette (si) measurement, while the optimal number of clusters was determined by maximizing the average si to the configuration obtained in the hierarchical clustering. Functional Principal Component Analysis (FPCA) was then performed to derive the PCs inside the final clusters71 and find an optimal orthogonal linear projection of the curves on a d-dimensional subspace (Rd when d = 2 or d = 3) to minimize the expected value of the squared error due to the projection. Lastly, a data depth algorithm based on the Modified Band Depth (MBD) was constructed focusing on centrality and separation of the spectra72, allowing to identify both the most representative and the most external curves of each cluster. All the statistical analyses were performed by using the R 3.6.1 software; a more extensive review of the performed statistical analyses and the R-packages used is reported in the Supplementary Methods section.

Fluorescence microscopy

Fragments of the wallpapers were imaged both on recto and verso sides by a Leica TCS SP5 fluorescence confocal laser scanning microscope (CLSM), using a 40X-1.25 NA objective (Leica Microsystems, Germany). Images were acquired under two-photon excitation at 830 nm in the 450–700 nm emission range. The samples were soaked in glycerol during the imaging procedure. The same setup was applied for a 3D reconstruction of the wallpaper multilayers. The data were analysed by ImageJ software.

Microbiological analyses

The sample areas of interest of Kinkarakawa-gami were gently swiped with sterile cotton swabs, which were then suspended in the physiological solution (sodium chloride 0.9% w/v) for 30 min. Afterward, the suspensions were serially diluted, being aliquots (100 μL) spread onto both tryptic soy and malt extract agar plates to isolate either bacterial or fungal strains respectively, whose biomass growth was carried out at 30 °C for 5 days under static conditions.

Polymerase chain reaction (PCR) was performed (Supplementary Methods), following thermocycler conditions described elsewhere73, on the extracted and purified genomic DNA to obtain the 16S rRNA gene product and the internal transcribed spacer (ITS) region in the ribosomal RNA operon. To identify—at genus level—the isolates retrieved from Kinkarakawa-gami’s art, PCR products were purified through QIAquick PCR purification kit, according to the manufacture’s protocol, sequenced (BMR Genomics, Padova, Italy), and searched for nucleotide homology with other microorganisms (Supplementary Methods).