Unique Chemical Parameters and Microbial Activity Lead to Increased Archaeological Preservation within Vindolanda, UK.

Waterlogged burial conditions impact upon artefact preservation. One major determinant of preservation is presence and behaviour of microorganisms, however, unravelling the mechanisms is challenging. In this study, we analysed elemental composition, bacterial diversity and community structure from excavation trenches at the Roman Site of Vindolanda, Northumberland, UK, using pXRF and 16s rRNA gene amplicon sequencing. Excavation trenches provide layers which can be classed as phosphorus sinks and which are strongly linked to the abundance of vivanite. The results indicated that microbial communities were dominant in Firmicutes, Bacteroidetes and Proteobacteria at a phylum level. Samples which also had visible vivianite presence showed that there were marked increases in Methylophilus. Methylophilus might be associated with favourable preservation in these anaerobic conditions. More research is needed to clearly link the presence of Methylophilus with vivianite production. The study emphasises the need for further integration of chemical and microbiome approaches, especially in good preservation areas, to explore microbial and chemical degradation mechanisms.


Introduction
Vindolanda, a Roman auxiliary fort situated south of Hadrian's Wall near Bardon Mill, Northumberland UK, is a site known for its excellent preservation of leather and wooden artefacts 1 .This is typi ed by the writing tablets showing the correspondence of Julius Verecundus, Prefect of the First Cohort of Tungrians, dated to between 1st and 2nd centuries; an extensive assemblage of leather artefacts; and the discovery of the oldest Roman boxing gloves c.a. 120 AD.The fort is situated upon the Alston formation, which is calcium carbonate sedimentary bedrock, overlaid by loamy and clayey soils.However, this alone does not explain the exceptional level of artefact preservation 1 .Between Roman occupation periods, wooden and stone buildings were destroyed, sealed with thick layers of clays and then re-built upon, forming layers in which oxygen was excluded from the decomposing material underneath.This lack of oxygen led to the formation of anaerobic layers which are ideal preservation environments 2 and created areas of waterlogging on top of dense clay layers.Waterlogged environments have unique and complex hydrological chemistry, hence they are often too complex to determine individual indicators of good presevation 3 .A general lack of understanding of waterlogged environments has led English Heritage to adopt a reburial policy for in-situ preservation as is typi ed by the Rose and Globe Theatres in London 4 .Preservation indicators in waterlogged environments can include and be in uenced by i) redox reactions, such as Fe 3+ reducing to Fe 2+ , and formations of chemical complexes like vivianite through combination of Fe 3+ and phosphate 2 , ii) microbiological species, such as decreasing abundance of Proteobacteria, Actinobacteria and Firmicutes, iii) hydrological indicators, such as acidity (pH values), conductivity and oxidation-reduction potential, iv) the decomposition material of surrounding highly organic matter, including those rich in phosphates 3 .
Redox reactions are one of the most important and visually impactful processes present at Vindolanda.Chemically vivianite (Fe 3 (PO 4 ) 2 .8H 2 O) can form, when there is an availability of iron and phosphorus in the soil, lack of oxygen, low sulphur content and low pH values 2,5,6 .The phosphorus released from the decaying organic matter, such as oor material, animal faeces and plants is retained, through sorption to iron oxides and as such vivianite acts as a phosphorus sink.Upon excavation and exposure to oxygen, Fe 3+ is converted to Fe 2+ , which produces the unique blue colouration of vivianite 2 .Interestingly, vivianite is associated with other instances of good preservation including that of human remains: For example human DNA of increased quality can be extracted from samples located within the Brisbane burial grounds from underneath vivianite crusts 5 .The formation of vivianite is also thought to be microbially mediated with studies showing a range of microorganisms capable of producing the substance in anaerobic culture when exposed to the correct conditions 6-8 .Thus, despite vivianite acting as a visible preservation indicator, the exact chemical and microbiological formation, production and generation mechanisms are not completely understood.
Microorganisms play a key role in decomposition, with above ground artefacts degraded by a range of bacteria, fungi and insects relatively quickly 9 .The speed of this degradation is impacted by oxygen availability, sample age and material which in turn impact upon the microbial community capable of carrying out decay 10 .Upon burial, particularly under anaerobic conditions, many fungal species (for example, soft rot fungi) typically involved in degradation cannot act.This means that the majority of degradation in anaerobic conditions is carried out by bacteria 11 .There is also evidence in marine sediments that key microbial communities, linked to anoxic environments can be central to increased preservation 12 .Ultimately, the mechanisms responsible for the excellent preservation observed at Vindolanda are unknown and will be explored within this study.Our assumption is that the exceptional levels of preservation seen at Vindolanda are brought about by a combination of environmental factors speci cally linked to the unusual anaerobic and high iron conditions observed, through an interaction between local geochemistry, hydrology and past human activity at the site (clay sealing).Such conditions in turn impact on the microbial community within the soil, producing vivianite and limiting growth of degrading organisms.
The initial aim of this study was to characterise the microbial community within different occupation layers of archaeological soil at Vindolanda.Bacterial communities within the different occupation layers which exhibit various degrees of preservation were compared with non-occupation environmental soil layers.The overall aim was then to correlate changes in environmental variables within the site and begin to unpick whether the microorganisms, environmental conditions or a combination of the two may be responsible for the good preservation observed at Vindolanda.As the factors in uencing preservation and decomposition are poorly understood it is hoped that this study will help to shape current thinking of factors responsible for improved preservation at archaeological sites, to allow strengthening of in-situ preservation techniques.To our knowledge, this is the rst reported study of a Roman burial environment using both chemistry and microbiological data, obtained using next generation sequencing techniques.
Elemental analysis obtained from pXRF analysis showed signi cant changes in elemental composition between the differing locations.D1p and d2p, the sites of best preservation, displayed different elemental composition with respect to phosphorus, sulphur and iron (P < 0.001).Pairwise comparisons using Dwass-Steele-Critchlow-Fligner (DSCF) showed that the interaction between phosphorus, sulphur and iron in d1p and d2p was not signi cantly different to each other, but were signi cantly different to d3i, d4t, c1 and c2.Elemental composition for major and minor elements are provided in Fig. 1.Phosphorus has a range of Max = 1.094;Min = 0.512; Range = 0.582; Mean = 0.813; Median = 0.806 in d1p and d2p.

Microorganisms indicative of preservation layers
LEfSe analysis of the microbial community revealed speci c community structure associated with preservation layers characterised by Firmicutes, Epsilonproteobacteria, BRC1 and Chloro exi with no phyla indicative of non-preservation layers due to varied communities.Tentative organisms which may be indicative of preservation layers are identi ed as: the Firmicute Turicibacter; AlphaproteobacteriaFilomicrobium and Sphingobium; Betaproteobacteria Pusillimonas and Vogesella; Epsilonproteobacteria Arcobacter; and Gammaproteobacteria Thiovirga (Figure 4) Composition of samples that contain vivianite.

Discussion
Elemental composition within the soil layers Overall, there were signi cant shifts in elemental composition between soil samples at Vindolanda speci cally with increases in phosphorus, sulphur and iron within d1p and d2p preservation locations.
Non-preservation locations d3i, d4t, c1 and c2 showed consistent elemental composition regardless of the different trench locations.Whilst preservation of artefacts is poor in d3i and d4t there is evidence of Roman occupation.However, there is no such evidence in c1 and c2 suggesting they sit outside of the settlement.Interestingly, there was little observable difference within pH values between any of the samples.
Iron and sulphate reduction, denitri cation and nitrate respiration are associated with anaerobic environments 13 , with dissimilatory Fe 3+ reduction being the most abundant anaerobic metabolism 14 .
Anaerobic soil conditions are typically characterised by the accumulation of iron cations (Fe 2+ and Fe 3+ ) as soil-available Fe is reduced.In aerobic conditions oxidation of Fe 2+ to Fe seems to correlate to a speeding up of microbial metabolism and a decomposition of organic matter both within eld and pot experiments 15 .However, pXRF analysis used in the current study cannot differentiate between the different oxidative states of iron; it does however, suggest that there is an increase in the amount of iron present within these locations not just a change in redox state.
Phosphorus release is associated with increased iron reduction within anaerobic soil environments with the dynamics of phosphorus release and Fe 3+ reduction closely linked 16 .However, there is also a relationship between the cycling of iron and phosphorus with the concentration of sulphate found within waterlogged environments.Within such conditions sulphate reduction results in increases in phosphorus release 17 .The iron and phosphorus concentrations are higher at Vindolanda within locations showing preservation and could, therefore suggest why increased vivianite are observed here 2 .Ratios of less than 1.5 sulphur:iron are needed for vivianite formation within marine sediments 18 ; all soils used within this study met these conditions.However, an excess of phosphorus is also needed which is only seen in d1p and d2p.
Certain soil types are more likely to result in good archaeological preservation than others with anaerobic soils rich in organic material being more likely to show reduced biological oxidation of wood, leather and soft tissue 19 , partly due to the microorganism which ourish in such environments 10 .Reduced forms of phosphorus, sulphur and iron are more likely to be found within anaerobic soils compared with aerobic.
Furthermore, the changes in moisture conditions at Vindolanda are linked to the periodic release of agricultural leachate which would be expected to increase soil degradation but may allow reaction with organic material within the soil, thus facilitating a natural 'tanning' process 19 .

The Microbial community within the Vindolanda soil layers
Preservation and interface locations d1p, d2p and d3i are distinct from other samples in terms of composition of phyla and genera.Shifts in bacterial phyla are correlated to location of sample but not pH values or elemental composition.Preservation locations are dominated by Firmicutes, Bacteroidetes and Proteobacteria at a phylum level with LEfSe analysis suggesting that Firmicutes, Epsilonproteobacteria, BRC1 and Chloro exi may be indicative of preservation.Non-preservation and control locations were typi ed by increased abundance of Acidobacteria, Actinobacteria and Planctomycetes.Shannon diversity remained high within preservation layers and was comparable to the surrounding soil.Maintenance of diversity and evenness within good archaeological preserving environments has been shown previously in No. 1 Wangshanqiao Chu Tomb in Jingzhou, China 20 .
Previous studies have shown that at the phylum level soils within the top approximately 20 cm are likely to be dominated by Acidobacteria, Actinobacteria and Alphaproteobacteria with Firmicutes and Spirochaetes being more abundant below 50cm 21 .Acidobacteria and Actinobacteria are increased within d4t, c1 and c2 compared with d1p, d2p and d3i.Actinobacteria are widely recognised as being crucial to aerobic degradation of organic matter particularly that which is cellulose based 22,23 .Acidobacteria are a diverse phylum which are mostly heterotrophic in metabolism and can degrade a wide range of carbohydrate sources 24 .Members of the Acidobacteria group 1-4 have been linked with degradation of cellulose based material 25 and are seen in signi cant abundance within layers 4-6 of this study.Interestingly they are greatly reduced within the preservation layers despite the pH values remaining consistent.This is probably due to the anaerobic conditions present within these layers as the majority of Acidobacteria are aerobic 24 .
Firmicutes, Bacteroidetes and Proteobacteria were signi cantly increased within the preservation layers.However, this is most likely a result of the anaerobic conditions found rather than these phyla not being decomposers.Proteobacteria, Firmicutes and Bacteroidetes are the most dominant phyla in solid waste decomposition within anaerobic phases 26 with Firmicutes and Bacteroidetes proposed as the main microorganisms involved in cellulose degradation 25 .There are very few studies assessing microbiology within archaeological sites.However, those available correspond with our ndings of consistent microbial diversity between layers which show preservation and those which do not 20,27 .Siles (2018) also found varied soil pro les that were common to soil environments and recorded no dominant 'rare' microbial taxa as clear indicators of preservation 27 .
At the genus level preservation locations showed increased abundance of Sul curvum, Flavobacterium and Methylophilus with additional genera Turicibacter, Filomicrobium, Sphingobium, Pusillimonas, Vogesella, Arcobacter and Thiovirga being identi ed as indicative of preservation by LEfSe.These microorganisms are not typically indicative of depth or oxygen availability as changes at phylum level were.Instead Sul curvum, Flavobacterium and Methylophilus are all capable of degrading polycyclic aromatic hydrocarbons (PAHs) and are found to be encouraged by increased concentrations of iron and sulphur 28,29 .Flavobacterium and Methylophilus are aerobic, whilst Sulfuricurvum are facultative anaerobes which are major contributors to the sulphur cycle given their abilities to oxidise sulphur and are observed in the presence of excess methane [30][31][32] .This coupled with the changes in soil chemistry involved suggests an unusual soil environment within the preservation layers at Vindolanda.

Microbial community with vivianite
Interestingly, further investigation of samples with visible vivianite showed that these soils contained marked increases in Methylophilus compared with adjacent soils.Methylophilus are methylotrophic Betaproteobacteria commonly occurring in activated sludge, mud, and aquatic habitats, and are capable of utilizing methanol or methylamine as a sole source of carbon and energy 33 .Members of the genus are thought to be strictly aerobic and have been extracted from various soils 32 .Usually Methylophilus would be expected to be found in low abundance within soil environments.However, previous studies have found increased amounts of CO 2 within soil samples can lead to proliferation of Methylophilus presumably due to an increased abundance of methane 34 .
Methylophilus is capable of solubilizing inorganic phosphorus 35 and produces siderophores and indole acetic acid 32 .Studies have shown that Methylophilus species are capable of reducing iron species such as ferrihydrite within conditions of increased methanol with evidence that the ferrihydrite was transformed into crystalline magnetite, a mixed Fe 3+ oxide 35 .Furthermore, other microorganisms capable of producing magnetite can also produce vivianite in phosphate rich environments 36 .
Studies in Brazil have shown that, with the exception of environments that are high in sulphur, phosphorus and iron the production of iron species such as magnetite and haematite are relatively common while the production of vivianite is more restricted 37 .These conditions are similar to those we have identi ed within the preservation layers at Vindolanda.Additionally, analysis within the South China Sea showed that concentrations of iron-bound phosphorus and organic-phosphorus were in uenced by sulfate-driven and metal-driven anaerobic oxidation of methane resulting in the release of phosphates and Fe 2+ and a subsequent signi cantly increased formation of vivianite 38 .

Conclusion
Our ndings taken together suggest that the soil at Vindolanda contains distinctive chemical and microbial pro les associated with layers that show increased preservation of organic/inorganic artefacts.These shifts are not simply the result of changes in depth and oxygen availability but suggest that the speci c conditions potentially allow the microbial production of vivianite which may lead to enhanced preservation of artefacts.The role of the occupation layers as phosphorus sinks is strongly linked to the abundance of vivanite.More research is needed to clearly link the presence of Methylophilus with vivianite production.

Sampling and archaeological context
Soil was sampled from two trenches within the Vindolanda site (54.9911°N2.3608°W).The external surface was scraped back by approximately 5cm.Samples were then taken in triplicate at four equal depths (d1p, d2p, d3i and d4t), each representing a separate archaeological context (see supplementary Fig. 2).The top 40cm of the trench had was exposed for approximately 3 months prior to the sampling taking place down to the interface level of d4t.The lower 1m section was excavated to the level of d1p four weeks prior to the sampling taking place.D1p and d2p correlate to the deepest layers and are associated with the best quality preservation.D3i and D4t are associated with lower quality preservation while d3i forming an interface (i) layer.Soil was sampled in triplicate (A-B) from a control (c) trench showing no evidence occupation or archaeological preservation, at two depths (c1-c2) (see supplementary Fig. 3).
In addition, further samples were taken from a rampart section, as detailed in Taylor et al 2019.The rampart sample was sampled in triplicate from the vivianite visible layer (blue) as well as triplicate samples within 10 cm which showed no visible vivianite formation.

Moisture
Triplicate soils were weighed before and after drying in an oven overnight at 105°C to determine the moisture content of the original sample using Eq. 1.
Equation 1:  ℎ () −  ℎ () /  ℎ () × 100 =   % pH Dried soils (10g) were placed into plastic cups with distilled water added (20ml) for a 1:2 proportion.The solution was stirred with a glass rod and settled for 30 minutes.pH values was analysed using an Advanced Jenway 3510 pH meter calibrated with commercial buffers of pH 4, 7 and 10.The electrode probe was rinsed between each sample.pXRF Soils were analysed following the method provided by Williams (2020) 39 .The dried soil samples were homogenised with mortar and pestle for 140 seconds and sieved to 2mm.Soil were loaded into XRF sample cups (SPEX CertiPrep™ 3529) and covered with 5 µm polypropylene thin-lm (SPEX™ SamplePrep 3520 window lm).A Thermo Niton™ XL3t GOLDD + pXRF with an Ag anode (6-50kV, 0-200 µA max X-ray tube) was used to analyse the soil.The pXRF was warmed up, system checked against the internal 1¼ Cr-½ Mo coupon, and scanned against a blank standard and NIST 2709a standard (y = 0.9674x -0.0089, r 2 = 0.9998).The NIST 2709a is a San Joaquin soil, used because it was certi ed for most elements of interest within the single calibration sample.The pXRF was periodically reset and system checked to account for drift.Soil samples were analysed using Mining mode (fundamental parameters), with 30second scans for the main lter (50 kV, ≤ 50 µA), low lter (20 kV, ≤ 100 µA) and high lter (50 kV, ≤ 40 µA), and a 60-second scan for the light lter (6 kV, ≤ 200 µA).

DNA extraction
DNA was extracted from 0.25 g of each soil sample using the Qiagen DNeasy PowerSoil Kit following manufacturer's instructions.DNA quality and yield were measured using NanoDrop 1000 spectrophotometer (V3.8.1, ThermoFisher) and agarose gel electrophoresis.DNA was stored at -80°C prior to further analysis.

Next Generation
Bacterial community DNA sequencing was made (NU-OMICs, Northumbria University, Newcastle Upon Tyne, U.K.) by targeting the V4 region of the 16S rRNA gene according to Kozich (2013) 40 .The raw sequencing reads were processed in FASTQ format and analysed with Mothur software package (version 1.36.1)(University of Michigan, U.S.A.).UCHIME was used to quality check and lter the FASTA formatted sequences for chimeras.These were aligned to the SILVA reference, and taxonomic identi cation of the reads was assessed by assigning sequences to operational taxonomic units (OTUs) using Ribosomal Database Project (RDP) classi cation.PCR negative controls were run and sequenced in parallel to the experimental samples.The OTUs recorded in negative controls and samples were excluded from further analysis.Non-bacterial sequences (Archaea/Eukaryota/Unclassi ed) were discarded.OTUs less than 3% were classi ed as rare taxa with these and the unclassi ed OTUs omitted from the plots.

Statistical analysis
Data were analysed in R 4.0.2 41using the car, DescTools and PMCMR packages [42][43][44] .Normality and variance failed the Q-Q plot, Shapiro-Wilk's and Levene's assumption tests.Data were analysed using Kruskal-Wallis rank sum tests, followed by Nemenyi tests with Chi-Square distribution to identify