Archaeological excavations of the six new sacrificial pits K3 to K8 were conducted in October 2020. The sacrificial pits K3, K4, and K8 were selected as the subject of this study. The shape, size, and some physical–chemical properties of the three sacrificial pits were introduced in the previous reports (Sun et al., 2024). A brief introduction to soil moisture content and pH value of buried environment was as follows. The soil moisture content in K3 was 22.64%, and the pH was 7.16. The soil moisture content of K4 was 22.08%, and the pH was 7.18. The soil moisture content of K8 was 20.20%, and the pH was 7.19. During the excavations, five soil samples (numbered “1–5,” test sample) were collected from the lower of artifact layer in K3, K4, and K8, respectively, for a total of 15 samples. In addition, the raw soil samples (numbered “0,” control group A) at the bottom of each pits in K3, K4, and K8 and one raw sample (numbered “S_0,” control group B) outside the sacrificial pits were also analyzed as control samples to comparative analysis. Soil samples were gathered using sterile scalpel and collected into Corning centrifuge tubes with −30°C storage temperature. As shown in Tables 1, 19 samples were categorized according to the excavated sacrificial pits and the type of artifacts near its location. All samples were sequenced three replicates to minimize errors and finally a total of 57 samples.

Total genomic DNA was extracted from 0.5 g soil samples using the FastDNA® Spin Kit for Soil (MP Biomedicals, LLC, USA) as recommended by the manufacturer’s instructions. The concentration and purification of DNA samples were detected with NanoDrop 2000 (Thermo Fisher Scientific Inc., USA). The integrity of DNA was examined using 1% agarose gel electrophoresis with the voltage of 5 V/cm for 20 min by the electrophoresis system JY-600C (Beijing Junyi Electrophoresis Co., Ltd., China).

Subsequently, PCR amplification was conducted using ABI GeneAmp® 9700 (Thermo Fisher Scientific Inc., USA). The primers of bacteria were designed based on the V4 hypervariable regions of 16S rRNA with 515F 5′-GTGYCAGCMGCCGCGGTAA-3′ and 806R 5′-GGACTACNVGGGTWTCTAAT-3′(Peiffer et al., 2013). The primers of fungi were amplified with the rRNA ITS1 hypervariable regions of ITS1F 5′-CTTGGTCATTTAGAGGAAGTAA-3′ and ITS2R 5′-GCTGCGTTCTTCATCGATGC-3′(He et al., 2021). After quality check of amplicons employing 2% agarose gel, the purification was conducted with the AxyPrep DNA Gel Extraction Kit (Axygen Scientific Inc., USA) according to the manufacturer’s instructions. The amplicons were quantified using Quantus™ Fluorometer (Promega Co., USA) for subsequent sequencing.

The DNA libraries were established utilizing the NEXTFLEX Rapid DNA-Seq Kit (BIOO Scientific Co., USA) in accordance with the manufacturer’s instructions. The purified amplicons were pooling in equimolar amounts, and then, the DNA libraries were validated by the Agilent Fragment Analyzer 5400 (Agilent Technologies, Inc., USA) with the approximate size 450 bp and 480 bp for bacteria and fungi, respectively. After passing the test, sequencing was performed on an Illumina MiSeq PE300 platform (Illumina, Inc., USA) at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China).

The raw sequencing files were quality-filtered and merged by fastp (version 0.19.6) (Chen et al., 2018) and FLASH (version 1.2.11) (Magoč and Salzberg, 2011), respectively, to select high quality data and remove the adapters. Reads were excluded with the score < 20. The optimized sequences were clustered into operational taxonomic units (OTUs) employing UPARSE (version 7.1) (Edgar, 2013) at the 97% sequence similarity (Li et al., 2017).

The most abundant sequence of each OTU was selected as a representative sequence, which was annotated the taxonomy via RDP Classifier (version 2.13). The bacterial sequences were submitted to the SILVA 138/16 s rDNA database¹ with the confidence threshold of 70%. Similarly, the fungal sequences were analyzed at the same confidence threshold against the UNITE 8.0/its_fungi database.²

The bioinformatic analysis of all samples was carried out via the Majorbio Cloud platform.³ Alpha diversity (α-diversity, diversity within-sample) of microbiome was assessed by Mothur (version 1.30.2) with the indices of the observed richness (Sobs), the Chao1 estimator (Chao 1), the Shannon diversity index (Shannon), the Simpson diversity index (Simpson), and Pielou evenness index (Pielou_e). The Kruskal–Wallis H-test was performed to examine the significant differences between the different samples (*p < 0.05; **p < 0.01; ***p < 0.001). The beta diversity (β-diversity, similarity and difference between groups of samples) was visualized with principal co-ordinate analysis (PCoA) based on the Bray–Curtis distance via the vegan package in R3.3.1. The differences in microbial taxa of groups were performed by stats package in R3.3.1 and scipy package in python v1.0.0.

In total, 3,922,031 bacterial sequences and 4,655,448 fungal sequences were obtained from high-throughput sequencing for analysis after verifying, with the average length 256 bp of former and 240 bp of latter. Subsequently, the low-abundance OTUs with <10 copies in all samples were filtered to improve the reliability of microbial composition estimates (Bokulich et al., 2013; Nikodemova et al., 2023). Meanwhile, the number of effective sequences from all samples was rarefied to the number of the sample with minimum sequences, to minimize the effect of sequencing depth on the subsequent bioinformatics analysis.

To ensure the sequencing quantity and depth were sufficient and representative, the rarefaction curves of bacteria and fungi are given in Supplementary Figure S1, which tended to flatten out, suggesting that the sequencing data were able to comprehensively and truthfully reflect the microbial composition and community. Meanwhile, the Good’ s coverage (Supplementary Figures S1E,F) over 99.3% also indicated the sequencing data of all samples were enough for the follow-up analysis.

The α-diversity of microbiome was characterized with Sobs, Chao 1, Shannon, Simpson, and Pielou_e indices, to assess the community richness, diversity, and evenness. The larger the values of Sobs and Chao 1, the higher the community richness. The larger the Shannon value and the smaller the Simpson value, the higher the community diversity. Similarly, the closer the Pielou_e value is to 1, the higher the species evenness. As shown in Table 2, the Sobs and Shannon indices of control group S_0 were largest for both bacteria and fungi, which meant the outside raw soil samples had the highest community richness and diversity, while the Pielou_e index on fungi of control group S_0 was smallest, meaning the lowest community evenness. As for the test groups, group K4 indicated the best biodiversity. The order of bacterial richness, diversity, and evenness in the lower of artifact layer was K4 > K8 > K3, while fungi was different with the order of K4 > K3 > K8.

In terms of bacteria (Figure 3), almost all α-diversity indices displayed a highly significant difference (p < 0.001) between group S_0 and groups K3, K4, and K8. Group K3_0 and group K4_0 also indicated a difference (p < 0.05) or a highly significant difference (p < 0.001) in Sobs, Shannon, and Pielous_e indices from the group K3 and group K4, respectively. Furthermore, group K3 and group K4 also shown the statistical difference (p < 0.05) on Sobs, Shannon, and Pielous_e indices. The results demonstrated that the raw soil samples outside sacrificial pits (control group B), the raw soil samples inside sacrificial pits (control group A), and the test groups exhibited statistical difference in bacterial richness, diversity, and evenness, following the order: control group B > control group A > test group. Referring to fungi, the Shannon and Pielou_e indices did not perform the statistical difference, which meant the fungal diversity and evenness of all samples were closed, but the Sobs index of control group S_0 also shown the highly significant differences (p < 0.001) from other groups.

It was clearly observed that the statistical difference in α-diversity of bacteria was much higher than that of fungi. In addition, the statistical difference between different samples from the same sacrificial pit in bacteria and fungi was still remained (Supplementary Figures S2, S3), especially for control groups K3_0, K4_0, and K8_0 and the test samples from corresponding sacrificial pits. The results probably were caused by the different sampling locations and the nearby artifacts in spite of the same sacrificial pit.

Through the trimmer and filtering methods, a total of 3,174 bacterial OTUs and 733 fungal OTUs were detected with high reliability and accuracy. According to Venn plot (Figure 4), there were 2,212 bacterial OTUs and 402 fungal OTUs in the control group S_0, which were much higher than other groups. In test groups, the order of the bacterial OTUs was following K4 > K8 > K3, and the order of the fungal OTUs was K3 > K8 > K4.

More importantly, Figure 4 also displays the number of unique (in side) or shared (in core) OTUs in different groups, which revealed the similarity and overlap across sacrificial pits. Only 203 bacterial OTUs and 8 fungal OTUs were common to all groups, while 580 bacterial OTUs and 35 fungal OTUs were shared in groups K3, K4, and K8, which indicated the test groups K3, K4, and K8 exhibited higher overlap with each other rather than with control groups. Simultaneously, there were also numerous OTUs shared between two of the sacrificial pits or only appeared in one pits, suggesting the heterogeneity of the test groups from different sacrificial pits on OTU level. The similar overlap and differences were observed as well in OTU results of individual samples (Supplementary Figure S4). Even though samples collected from the same sacrificial pit, the OTUs of different samples still showed differences. The results above probably indicated the impact of excavation location and buried artifacts on soil microbiome.

To analyze the community composition and structure, the microbial OTUs were further classified. The bacterial OTUs were divided into 42 phyla, 405 families, and 667 genera. As shown in Figure 5, there were 18 dominant bacterial (relative abundance >1%) phyla in groups. The top bacterial (relative abundance >2%) phyla were Proteobacteria, Actinobacteriota, GAL15, Chloroflexi, Acidobacteriota, Methylomirabilota, unclassified_k__norank_d__Bacteria, Thermoplasmatota, Crenarchaeota, Gemmatimonadota, Firmicutes, and Nitrospirota, which were detected in all groups, and the combined abundance of these bacteria exceeded 80%. The phyla above were common to the samples of human archaeological remains. The results indicated that the test and control groups were similar at the phylum level in bacteria but differed in relative abundance.

The differences in bacterial community and composition among different pits were further noticed at the genus level. There were 65 dominant bacterial (relative abundance >1%) genera in samples. It should be noted that some norank and unclassified genera, such as genera norank_f__norank_o__norank_c__norank_p__GAL15, unclassified_d__Bacteria, and norank_f__norank_o__Gaiellales, were identified in all groups but lacking the accurate identification information. Thus, the bacteria only identified to order level were filtered out. The bacterial composition showed the evident differences between different groups. In group S_0, the top bacterial genera (relative abundance >2%) were genera Arthrobacter (4.65%), Streptomyces (3.07%), as well as genera norank_f__Nitrososphaeraceae (5.75%), norank_f__Vicinamibacteraceae (3.48%), norank_f__Roseiflexaceae (3.66%), and norank_f__Gemmatimonadaceae (2.04%). In group K3_0, the genera Sphingomonas (3.20%), Nitrospira (2.41%), and Candidatus_Udaeobacter (2.02%) were found. Meanwhile, there were six top bacterial genera including genera Cupriavidus (7.72%), Sphingopyxis (5.19%), Variovorax (4.91%), Limnobacter (3.28%), Candidatus_Methylomirabilis (3.26%), and Caulobacter (2.56%) in group K3. Similarly, eight top bacterial genera including the genera Limnobacter (9.20%), Cupriavidus (5.86%), Pseudomonas (4.20%), Bacillus (3.55%), Methylotenera (2.63%), Variovorax (2.57%), Sphingopyxis (2.49%), and Streptomyces (2.53%) were detected in group K4. In group K4_0, the top bacterial genera were genera Sphingopyxis (4.42%), norank_f__Gemmatimonadaceae (2.93%), Pseudomonas (2.65%), Variovorax (2.63%), Pedobacter (2.37%), Luteimonas (2.32%), unclassified_f__Comamonadaceae (2.30%), and Nitrospira (2.13%). Obviously, group K3 and group K4 had more precise dominant bacterial genera with higher abundance than group K3_0 and group K4_0, respectively. As for group K8, the top bacterial genera were genera Candidatus_Methylomirabilis (2.95%), Bacillus (2.45%), Arthrobacter (2.29%), Variovorax (2.57%), and norank_f__Gemmatimonadaceae (3.60%), while group K8_0 included more bacterial genera, such as genera Hansschlegelia, Nocardioides, Pseudonocardia, Luteimonas, Variovorax, Rhodococcus, and Olivibacter, accounting for 28.01, 11.97, 10.78, 7.12, 6.88, 5.42, and 2.13%, respectively.

The fungal OTUs belonged to 12 phyla, 182 families, and 298 genera. There were only four dominant fungal phyla (relative abundance >1%) in groups, which included Ascomycota, Mortierellomycota, Basidiomycota, and an unclassified_k__Fungi (Figure 6). It is worth noting that Mortierellomycota was not detected in groups K3_0 and K8, although it was more abundant in groups K3 and K4. At the genus level, the number of dominant fungal genera (relative abundance >1%) was 46. Similar to the bacterial results, the fungal genera in groups displayed the remarkable differences between groups. In group S_0, the top fungal genera (relative abundance >2%) were genera unclassified_f__Nectriaceae (28.92%), Neocosmospora (22.23%), Microascus (5.76%), Nectria (4.31%), Fusarium (3.97%), Penicillium (2.91%), Ilyonectria (2.32%), Purpureocillium (2.06%), and Thelonectria (2.02%). The top fungal genera in group K3 were genera Mortierella (16.93%), Simplicillium (11.33%), Purpureocillium (10.40%), unclassified_f__Nectriaceae (10.96%), Rhodotorula (7.78%), Fusarium (6.61%), Robbauera (5.72%), Cladosporium (4.52%), Aspergillus (4.05%), Cutaneotrichosporon (2.80%), Vishniacozyma (2.79%), and Penicillium (2.35%), while the different fungi such as genera Cephalotrichum (18.19%), unclassified_f__Cordycipitaceae (12.71%), Infundichalara (7.00%), Engyodontium (6.28%), Dichotomopilus (5.87%), Xenochalara (4.85%), and Alternaria (2.68%) were found in group K3_0. In group K4, the top fungal genera were more concentrated in genera Ilyonectria, Mortierella, Simplicillium, Cladosporium, Acremonium, and Lecanicillium, with the relative abundance of 34.29, 24.87, 10.29, 10.20, 7.70, and 4.51%, respectively. The control group K4_0 shown the accordant fungi with group K4 in addition to genera Cephalotrichum (42.38%) and Aspergillus (3.45%). In group K8, genera Purpureocillium (31.76%), Clonostachys (14.96%), Rhodotorula (6.65%), Cladosporium (4.63%), unclassified_f__Plectosphaerellaceae (3.95%), Simplicillium (3.61%), Penicillium (3.45%), and Aspergillus (2.41%) were the top fungal genera. Apparently, the types and relative abundance of different fungi in group K3 and group K8 differed obviously from group K3_0 and group K8_0, respectively. In contrast, the groups K4 and K4_0 displayed a certain similarity in fungal community composition.

The similarities and differences of bacterial and fungal communities were evaluated via clustering using Bray–Curtis distance-based principal co-ordinate analysis (PCoA) on the OTU level. The PCoA result of bacteria (Figure 7A) revealed that the sacrificial pits of K3, K4, and K8 shown significant difference in taxa biodiversity. Especially, group K4 exhibited the evident differences from groups K3 and K8. Group K4 also shown a dispersed distribution, which was closer to the control groups. As for fungi (Figure 7B), the group K4, as well as group K4_0, also performed significant difference with other groups, whereas it is hard to distinguish the samples of groups K3 and K8, suggesting that the fungal community between group K3 and group K8 was similar. The microbial distribution and structure were also performed by hierarchical clustering analysis (HCA, Supplementary Figures S5, S6), which were consistent with the PCoA results. Compared to group K4, group K3 and group K8 show higher similarity in bacteria and fungi communities and taxa biodiversity.

In addition, it should be noted that the location of samples and the material of artifacts around the samples, including pottery and bronze, also affect the microbial diversity. As shown in Figures 7C,D, the samples near pottery and bronze displayed the differences in bacterial biodiversity, but the distinction in fungi was not observed.

Based on the microbial results above, the differences in community and composition were further investigated. The top 20 bacterial and fungal genera are exhibited in Figure 8. As for bacteria, there were 15 genera manifesting the highly significant difference (p < 0.001) in different groups. The genus norank_f__norank_o__norank_c__norank_p__GAL15 was dominant in groups K3, K4, and K8. The genera Candidatus_Methylomirabilis, Sphingopyxis, and Cupriavidus were dominant in group K3. The genera Limnobacter, Pseudomonas, and Bacillus were more abundant in group K4. The genera norank_f__Gemmatimonadaceae and Candidatus_Methylomirabilis were more abundant in group K8.

Similarly, 15 genera of fungi performed the highly significant difference (p < 0.001). The genera Robbauera, Fusarium, and Rhodotorula were more abundant in group K3. The genera Ilyonectria and Mortierella were dominant in group K4. In group K8, the abundance of genera Purpureocillium and Clonostachys increased markedly. From the results of difference analysis, it could be seen that there were statistical difference in the abundance of different genera in each sacrificial pit.

A great number of bacteria and fungi exist in soil (Janssen et al., 2002). It is one of the most crucial aim to identify and compare the microbial community and composition of soil in different sacrificial pits. In recent years, several investigations have been conducted in the human archaeological remains (Philips et al., 2017). Some of the most common soil bacteria are considered to be Firmicutes, Actinobacteriota, Actinobacteria, Proteobacteria, Acidobacteriota, Methylomirabilota, Chloroflexi, Bacteroidota, and Nitrospirota (Kazarina et al., 2019; Wei et al., 2022). The fungal phyla of Ascomycota and Basidiomycota are dominated in human archaeological remains, followed by Mortierellomycota, Chytridiomycota, or Zygomycota (Margesin et al., 2016; Siles et al., 2018; Voyron et al., 2022). All bacterial and fungal communities collectively reflect the serious impact of ancient anthropogenic activities on soil environment.

In our results of bacteria, Proteobacteria were the most widespread phylum in the test groups, followed by Actinobacteriota, GAL15, Chloroflexi, Acidobacteriota, etc. (Figure 5). Proteobacteria are the largest and phenotypically most diverse phylogenetic lineage, which contain strictly anaerobic and aerobic species, facultative aerobes, and microaerophiles to participate the carbon, nitrogen, and sulfur cycles on Earth (Kersters et al., 2006). In Sanxingdui sacrificial pits, Proteobacteria were dominated via Alphaproteobacteria and Gammaproteobacteria. The former included the genera Caulobacter, Hansschlegelia, Sphingomonas, and Sphingopyxis. The latter contained genera Cupriavidus, Limnobacter, Variovorax, Methylotenera, Pseudomonas, and Luteimonas. Among them, genus Hansschlegelia, the most abundant in group K8_0, represented a novel lineage of autotrophic methanol-utilizing bacteria as the carbon and energy source in plant and soil (Ivanova et al., 2007; Wen et al., 2011). Genus Sphingopyxis has been reported in catacombs (Alias-Villegas et al., 2013), which is also widespread in diverse ecological environments, including soil, sediment, water, and heavy metal contaminated sites (Sharma et al., 2021). It has attracted considerable attention not only for excellent ability to survive under extreme environments but also for the ability to degrade organic matter. Therefore, although the content of copper ion was high in the soil of sacrificial pits, genus Sphingopyxis was still found dominantly in K3, K4, and K4_0, probably indicating the presence and biodegradation of organic matter. Genus Limnobacter is an obligately aerobic and grew on various organic substrates, but it has been designated chemolithoheterotrophs, which is able to grow by oxidation of thiosulfate to sulfate (Spring et al., 2001). The capabilities might indicate the presence of sulfide or degradation of organic matter in sacrificial pits. Genus Variovorax generally inhabits soil and water and plays an important role in the regulation of plant growth and development (Finkel et al., 2020). Meanwhile, the genus is found in the ancient apple seeds (Milanesi et al., 2019) and decaying woods (Ghio et al., 2012). Genus Variovorax, hence, is considered as a possible source of hydrolytic enzymes to degrade cellulose. Moreover, it should be noted that the genus is often detected after bushfire (Smith et al., 2008). Therefore, the presence in all groups except group S_0 might indicate the impact of high temperature processes in the ash layer on bacteria.

Actinobacteriota, the second abundant bacterial phylum in our results, are generally distributed across both terrestrial and aquatic environments, as well as in the microbiome of higher eukaryotes. The phylum shows the unrivalled metabolic versatility, producing highly diverse bioactive natural products, but it must be noted that there were lots of unclassified genera with high abundance in Actinobacteriota in our results, such as genera g__norank_f__norank_o__norank_c__MB-A2-108 and g__norank_f__norank_o__Gaiellales. Therefore, it is difficult to analyze their properties and functions in the ecosystem. Furthermore, the genera Streptomyces, Rhodococcus, Nocardioides, and Pseudonocardia were identified in sacrificial pits, which are found in ancient wall painting, ancient seeds, wooden shipwreck, tomb, or cave (Stakhov et al., 2008; Juanjuan et al., 2016; Liu et al., 2018; Sakr et al., 2020). Among them, genus Streptomyces can cause the deterioration of complex natural polymers via producing a great deal of enzymes (Sakr et al., 2020). The genus was dominant in groups K4 and S_0, probably indicating the presence of natural polymers. The phylum Methylomirabilota plays a key role in global carbon and nitrogen cycle as the methane oxidizing bacteria. The genus Candidatus_Methylomirabilis participates in the cycle by oxygenic denitrification and aerobic methane oxidation pathways (Sun et al., 2024). The dominant abundance in groups K3 and K8 indicated that a certain amount of organic matter probably was present in sacrificial pits.

In the part of fungi, Ascomycota were the most abundant fungal phylum, with the genera Cladosporium, Aspergillus, Penicillium, Simplicillium, Purpureocillium, Lecanicillium, and Acremonium. Among them, the genera Cladosporium and Simplicillium, belonging to family Cordycipitaceae, were dominant in all groups expect for group S_0, suggesting the effect of human burial behavior and buried artifacts. The genus Cladosporium is isolated in ancient wall painting and beeswax drops on ancient papers (He et al., 2021; Pavlović et al., 2022), showing the proteolytic and lipolytic abilities in the process of organic degradation. In addition, previous studies have indicated that the genus Cladosporium is capable of hydrolyzing cellulose by the production of cellulases (Leung et al., 2016; Li et al., 2019). The genus Simplicillium is identified on the polychrome pottery figurines (Zhang et al., 2023), resulting in the fading of the blue (azurite) and green (malachite) pigments through the absorption of copper (II) ion. Moreover, the genus has been proved to degrade tempera paintings and artifacts made with dammar varnish and rock-art paintings (Jurado et al., 2008). So genus Simplicillium is destructive to both organic and inorganic materials. The genera Lecanicillium and Acremonium belong to the order Hypocreales. Genus Lecanicillium contains some entomopathogenic species, and some are capable to degrade organic compounds (Fenice, 2016; Pozdnyakova et al., 2019). For instance, Lecanicillium muscarium is a potent producer of extracellular cold-resistant, chitin-hydrolyzing enzymes for the degradation of chitin-rich materials (Fenice, 2016), but the degradative properties of the genus Lecanicillium have been poorly studied. The genus Acremonium is the environmental saprophytes in soil and decaying plant material and is able to secrete a lot of cellulolytic enzymes with high activity (Amiri et al., 2011; Hu et al., 2015). The genus is found in ancient wall paintings (He et al., 2021), and the dominant proportion in group K4 probably implied the degradation of cellulose organic matter.

The genera Robbauera, Rhodotorula, Vishniacozyma, and Cutaneotrichosporon, belonging to Basidiomycota, were found in K3 and K8, especially in group K3 with the dominant abundance. Among them, genus Vishniacozyma has been identified as the core composition during the spontaneous fermentation of ice wine, probably relating to the fermentation process (Chen et al., 2020). As the pigmented yeast, genus Rhodotorula is strict aerobic yeasts with peculiarly metabolic characteristics. The genus is very important for food industries because it can produce glycogen, lipids, and carotenoid pigments during the growth phase (Hernández-Almanza et al., 2014), but it should be noted that the genus does not ferment carbohydrates (Dworecka-Kaszak and Kizerwetter-Świda, 2011). In addition, genus Mortierella, belonging to Mortierellomycota, was the core part in groups K3, K4, and K4_0. The genus is the widespread filamentous fungi that can decompose various organic compounds such as cellulose, hemicellulose, and chitin (Ozimek and Hanaka, 2021).

The problem of biocorrosion could not be ignored during thousands of years of burial. Considering that the bronzes were the most unearthed artifacts, the biocorrosion (also known as microbiologically influenced corrosion, MIC) of bronzes was the major concern. Some microbial taxa known to corrode metal material or solubilizing metals from metal-bearing substrates were found. As for bacteria, the genus Pseudomonas is repeatedly identified in archaeological remains and cultural heritage (Li et al., 2018; Siles et al., 2018; He et al., 2021; Pavlović et al., 2022). Numerous studies have shown that the genus Pseudomonas contains nitrogen-fixing bacteria and nitrate-reducing bacteria (NRB) (Dou et al., 2022; Wu et al., 2023). The NRB can reduce nitrate under biocatalysis by transporting the electrons into the cell cytoplasm and further cause the biocorrosion. The genus Arthrobacter is an aerobic culturable bacteria in soil and is also thought to be able to decompose various organic matter (Yao et al., 2015). More importantly, the genus also contained some species with excellent tolerance to excess heavy metal, which formed biofilms and colonized on the surface of copper, resulting in the pitting corrosion (Swaroop et al., 2016). The Variovorax sp. has also been isolated from bacterial biofilms of copper pipes, probably leading to the biodeterioration (Reyes et al., 2008), which might be the reason why the genus was significant abundance in all sacrificial pits and almost absent in control group S_0. In addition, the genus Cupriavidus also shows the prominent resistance to high concentrations of copper ion, and its growth initiation is strongly stimulated by copper in some instances (Makkar and Casida, 1987; Gillan et al., 2017). So it was reasonable that genus Cupriavidus was present in groups K3, K4, and K8 in spite of the high concentration of copper ion from bronzes.

Many fungal genera are also capable of corrosion metals. Recent study has demonstrated that the conidia and hyphae of some fungi adhere to the surface of the metal. They penetrate into the surface layer of metal and go deep into the bulk, finally causing corrosion damage with pitting, ulcers, and cavities (Belov et al., 2023). The genus Penicillium was found in all groups and dominant in groups K3 and K8, which has been proven to be able to solubilize several metals including zinc, stannum, aluminum, plumbum, and copper (Siegel et al., 1983). The genus Aspergillus exhibits the marked tolerance to copper and also shows the evident biocorrosion of copper metal through proton-and ligand-mediated dissolution mechanisms, causing obvious mass loss and surface etching (Zhao et al., 2020). This was a good explanation for why the genus was the dominant fungi in sacrificial pits. Genus Rhodotorula, dominant in groups K3 and K8, unexpectedly, was also identified in biofilms that causes MIC in copper pipes, probably meaning that the genus was involved in the corrosion of bronzes (Reyes et al., 2008) in sacrificial pits. In addition, the genus Cladosporium was also involved in the biodeterioration of metal (Iverson, 1987).

As we all know, the MIC process of metals is related to the formation of microbial biofilms, which can form on any metal surface. The bacterial and fungal colonization in biofilms can modify the localized environment at the metal surface and cause corrosion. Generally, there are two typical MIC processes, namely, the extracellular electron transfer MIC (EET-MIC) and metabolite MIC (M-MIC) (Dou et al., 2022). The former mainly utilizes EET to transport the electrons in the corrosion process of sulfate-reducing bacteria (SRB) and NRB because metal oxidation occurs extracellularly while sulfate or nitrate reduction occurs intracellularly with biocatalysis (Di, 2022). The genus Pseudomonas and genus Arthrobacter are usually considered to corrode the copper by EET-MIC. For example, Pseudomonas aeruginosa utilizes copper as an electron donor, and the terminal electron acceptor is nitrate, causing and accelerating the corrosion of bronzes (Dou et al., 2022). The Arthrobacter sulfureus is electrochemically active and is able to transfer electron through the membrane-associated C-cytochromes or through conductive pili (Swaroop et al., 2016). As for the other biocorrosion mechanism, the M-MIC often is caused by corrosive metabolites such as organic acids secreted by bacteria and fungi. The genus Variovorax in biofilm may destroy copper by releasing acid metabolites. Moreover, more and more research studies (Rao et al., 2002; Anjum et al., 2009; Amiri et al., 2011) also demonstrate that the fungi such as genera Penicillium and Aspergillus can generate varieties of organic acid metabolites including citric acid, lactic acid, oxalic acid, malic acid, and gluconic acid to dissolve and leach metals effectively, especially copper.

By comparing the results of microbial diversity of different groups, it was clear to observe that the microbial composition and structure of raw soil samples exhibited an significant difference from the test samples in sacrificial pits K3, K4, and K8, meaning the microbial community could be modified in the archaeological site because of the buried man-made structure and artifacts. The results were consistent with previous studies (Junco et al., 1992). Compared to the control group S_0, there were many dominant bacteria and fungi, such as genera Caulobacter, Sphingomonas, Sphingopyxis, Limnobacter, Cladosporium, Acremonium, Simplicillium, and Mortierella in test samples in sacrificial pits. Some of them were related to the degradation of organic matter, indicating that there might be a certain amount of organic matter ever buried in the sacrificial pits, such as proteins, lipids, and carbohydrates. Unfortunately, as for organic residues, only some textile traces were found on the surface of the bronzes. Although a large amount of ivory was preserved, the collagen in it was almost completely degraded (Honglin et al., 2022). More investigations are needed to identify and demonstrate the presence of different organics in sacrificial pits in the further. The other dominant bacterial genera Pseudomonas, Arthrobacter, and Variovorax and fungal genera Aspergillus and Penicillium in sacrificial pits were in connection with the biocorrosion of bronzes, also revealing the impact of the buried bronzes on microbiome. In addition, the phylum GAL15 was the primary bacteria in all sacrificial pits. The abundance of bacteria GAL15 in different groups followed the distribution pattern: raw soil outside pits (group S_0) < raw soil in pits (group K3_0, K4_0, and K8_0) < test soil in artifact layer (groups K3, K4, and K8). These results demonstrated that the abundance of phylum GAL15 might be related to the buried artifacts, especially for the bronzes in sacrificial pits, which were consistent with previous findings in ivory layer of Sanxingdui site (Sun et al., 2024). Unfortunately, none of family or genus was matched in this phylum to in-deep analysis.

In addition, the proportion of buried artifacts of different materials also affected the microbial community. The results of β-diversity (Figure 7) indicated that the bacterial and fungal communities of group K3 and group K8 were much closer, and both differed significantly from those of group K4. This phenomenon might result from the material of the buried artifacts. Based on the excavation brief of sacrificial area of Sanxingdui site (Honglin et al., 2022), there are 2,686 numbered artifacts in K3, of which bronzes are the largest in number at 1,171, accounting for 43.60%. As for K8, there are 5,210 artifacts unearthed as the publication of the brief report, with 3,822 bronzes, accounting for 73.36%. Furthermore, the brief excavation report of K4 has been published separately (Danyang et al., 2024), which records the exact number of artifacts in artifact layer. There are 1,118 artifacts found in artifact layer of K4, with 468 pottery sherds and 348 bronzes. The proportions of pottery and bronzes were 41.86 and 31.13%, respectively. The large number of bronzes in K3 and K8, therefore, had a significant impact on the microbial composition and structure, causing the evidence differences from K4. Recent studies have revealed that the copper contents in soil near bronzes in sacrificial pits of Sanxingdui site increased several times (Jianguo et al., 2023). Many metals are essential for life, such as copper, sodium, kalium, zinc, and iron, while all metals can exert toxicity when presented above certain threshold concentrations (Gadd, 2010). The high content of copper ion in soil could inhibit the growth of certain microorganisms, which led to the significant difference of bacterial and fungal communities between different sacrificial pits, also between the sacrificial pits and soil outside pits. The PCoA results of bronzes and pottery sherds (Figures 7C,D) suggested the impact of copper ion was more obvious on bacteria than fungi, with the more concentrated distribution of the same type samples. Meanwhile, the bacterial community of all raw soils either inside pits or outside pits was closer to group K4, because of the lower proportion of the bronzes in K4.