Bone loss markers in the earliest Pacific Islanders

Kingdom of Tonga in Polynesia is one of the most obese nations where metabolic conditions, sedentary lifestyles, and poor quality diet are widespread. These factors can lead to poor musculoskeletal health. However, whether metabolic abnormalities such as osteoporosis occurred in archaeological populations of Tonga is unknown. We employed a microscopic investigation of femur samples to establish whether bone loss afflicted humans in this Pacific region approximately 3000 years ago. Histology, laser confocal microscopy, and synchrotron Fourier-transform infrared microspectroscopy were used to measure bone vascular canal densities, bone porosity, and carbonate and phosphate content of bone composition in eight samples extracted from adult Talasiu males and females dated to 2650 BP. Compared to males, samples from females had fewer vascular canals, lower carbonate and phosphate content, and higher bone porosity. Although both sexes showed evidence of trabecularised cortical bone, it was more widespread in females (35.5%) than males (15.8%). Our data suggest experiences of advanced bone resorption, possibly as a result of osteoporosis. This provides first evidence for microscopic bone loss in a sample of archaeological humans from a Pacific population widely afflicted by metabolic conditions today.

www.nature.com/scientificreports/ composition through carbonate and phosphate content ( Fig. 1), averages of which we hypothesised to differ between the estimated female and male sex groups.

Results
We found evidence for bone loss, which differed between the estimated sex groups in our samples. Porosity which manifested as cortical bone trabecularisation was particularly advanced in the females. As expected, due to sexual dimorphism, males in our sample had larger femora when compared to the females (Table 1, Fig. 1). The average midshaft circumference was 92.25 mm (standard deviation (SD) = 4.79) in males, but 79.75 mm (SD = 7.09) in females (U = 15.5; p = 0.029). The diameter of femur midshaft was also larger and more pronounced in males in both the antero-posterior (A-P), and medio-lateral (M-L) planes. The A-P diameter was 30.75 mm (SD = 2.87) in males, but 27.26 mm (SD = 2.15) in females, albeit this difference was not statistically significant. The sexual dimorphism was likely also reflected in the 2D histological measurements of bone Haversian canal density collected from the cortical area of bone unaffected by porosity producing trabecularisation, which was higher in males (17.445/mm 2 ) than in females (15. 1,2,4). However, the area of bone samples showing porosity producing trabecularisation which represented as 'giant' coalescing pores was more evident in females than in males (Fig. 1). The average porosity producing trabecularisation was 35.5% (SD = 16.40) in females, but only 15.8% (SD = 24.17) in males. In fact, all four female samples were affected by a high degree of porosity producing trabecularisation that was widespread from the most inner bone layer (endo-cortical) through the mid-section region (intra-cortical), reaching the bone areas immediate to the most outer bone layer (periosteum) (Figs. 1, 2). This led to cortical wall thinning 88,89 and was a consistent pattern throughout the sample as supported by a statistically significant negative correlation between percent areas of trabecularisation and cortical wall area (Rho = − 0.762, p = 0.028; Fig. 5). However, comparing only the porosity and Haversian canal variables between the two sex groups did not result in statistical significance as males were clearly impacted by intra-cortical trabecularisation as well. In fact, one of the male samples (ID: Sk9.3, Figure S1) showed an estimated 51.8% of trabecularisation of intra-cortical bone. This data point is maximum across our entire sample of eight bones. The remaining three male samples showed an overall consistent appearance of porosity whereby the bone surfaces were dense, showing a limited range of variably sized porosity regions and their occurrence (one complete absence, Figs. 1, 3). On the contrary, the female samples exhibited much more Figure 1. Summary of methods and key findings in the present study. Posterior view of two right archaeological femora from Talasiu individuals estimated as female (♀, ID: Sk3.1) and male (♂, ID: Sk3.2) shows the sectioning location (black dashed line) and approximate sample removed (red dashed box). Methodological steps included synchrotron sourced infrared microspectroscopy to measure bone mineral composition, histomorphometry to estimate Haversian canal densities (red dots), and laser confocal scanning of bone topography to provide a qualitative illustration of intra-cortical bone porosity producing trabecularisation.  Table 2, S3-S6) (p < 0.05 in all sFTIRM variables except for integrated phosphate area). The average peak height of phosphate (from 848 spectra; U = 1,118,955.000; p < 0.013), and v 2 (from 1676 spectra; U = 3,767,893.000; p < 0.0001) and v 3 carbonates (from 2159 spectra; U = 4,978,269.000; p < 0.0001) were lower in females. While the phosphate difference of 2.6% in the means seems marginal, it was more substantial when considering v 2 (30.3%) and v 3 (32%) carbonates. We note that more spectra were successfully measured in the samples from males (2496 phosphate spectra, minimum 3329 v 2 and 3264 v 3 spectra) ( Table 2, S3-S5). Compared to the male samples, the average integrated areas of phosphate and two carbonate peaks were also lower in females. The phosphate area in females was only lower by 1.1% (Table S3). However, the v 2 (U = 3,765,291.500;   Table S6). This was despite females recording a maximum ratio spectrum of 0.143 contrasted with a much lower maximum spectrum of 0.038 in males, and the higher total number of spectra collected in males (2495) contrasted with 846 spectra in females.

Discussion
We report the first microscopic record of bone loss characteristics in archaeological femora representing some of the earliest Pacific Islanders. On the basis of how commonplace contemporary metabolic issues, including obesity and diabetes, are in Tonga, we interpret our evidence to indicate that similar problems that might have led to osteoporosis occurred there ca. 3000 years ago. Two key implications are presented. Firstly, our small sample of archaeological humans from the Pacific appears to be afflicted by cortical bone porosity producing trabecularisation despite the notion that osteoporosis and related conditions are modern-day diseases. This confirms that the occurrence of widespread metabolic disease in this region today may have roots in the past. It agrees with bone loss experiences reported for archaeological collections from around the world (see reviews 15,62-64 ),  www.nature.com/scientificreports/ and archaeological metabolic bone disease indicators reported in other parts of the Pacific 11,12 . Secondly, the extensive cortical bone porosity leading to trabecularisation in our sample of archaeological Tongan females suggests experiences of oestrogen loss, mirroring modern day female osteoporosis incidence. This may further suggest at least the females represented in our sample lived to menopause age. The occurrence of trabecularisation in some of the male samples further points to advanced age despite widely held assumptions that past human longevity was shorter than today. We are unable to ascertain that archaeological Tongans specifically suffered from obesity, but we explain our findings through a differential diagnosis considering various aetiologies that might have impacted these individuals-complementarily or in an alternate fashion.

Occurrence of metabolic abnormalities as a result of lifestyle and environmental factors.
The examination of human remains from archaeological backgrounds, such as tracing the origins of tuberculosis or leprosy 90 , has revolutionised our current understanding of disease in the modern world. Osteoporosis is often considered a modern disease because it mostly manifests in the elderly and is strongly linked to modern sedentary lifestyles. However, its incidence was previously confirmed in various archaeological European 52-57 , North American 58 , and African 59-61 osteological assemblages. We now report the first microscopic record of abnormally porous human bones from the Pacific region, adding to the growing body of evidence that modern metabolic conditions in the Pacific may be rooted in the past 11,12 . Talasiu people are some of the earliest representatives of the Polynesian society emerging after the first settlement of archipelago at c. 2850 cal BP 84 . The estimated Tongan skeletal fragility presented here is consistent with a historic basis for other current highly prevalent widespread metabolic problems such as obesity and type www.nature.com/scientificreports/ 2 diabetes in the region [3][4][5] . Published evidence also exists for other types of skeletal abnormalities reported from different archaeological sites in Tonga 91-94 , and Vanuatu 11,12,95 . For example, scurvy and hypervitaminosis A, and iron-deficiency anaemia were discussed amongst several possible conditions arising in archaeological Tongan communities possibly as a result of poor nutrition 91 . This came from skeletal evidence in 17 children aged 6 months to 3 years old suggesting a combination of infections and metabolic bone disease at the ' Atele pre-European burial mounds dated to AD 1100-1250 from Western Tonga 91 . Further, poorly mineralised 21 archaeological deciduous teeth obtained from infant skeletons excavated at ' Atele in Tonga also indicated developmental disturbances in dental enamel linked to physiological stress arising from possible infections and nutritional deficiencies 92 . Stable isotope data extracted from archaeological human Tongan bone suggested diets to be predominantly based on starchy and marine food sources 83,93 . Selection of a wider variety of food sources would have been limited by the island size in addition to norms arising from social organisation 83   www.nature.com/scientificreports/ disruption indicators in 33 adult and 11 juveniles 94 . Lower spine degeneration (spondylosis) was also observed on archaeological Tongan lumbar vertebrae 94 . While spondylosis occurring as a result of bone degeneration due to age can be accelerated through weight bearing from active lifestyles 94 (combat sports at archaeological Tonga mounds were previously also discussed 96 , and see our next discussion section about physical activity links to age-related bone loss), clinical studies have also linked it to increased body weight in obesity 97 . The experiences of degenerative bone conditions in our sample of the Tongan individuals can be inferred from joint lesions indicating osteoarthritis (Table S1). In Vanuatu, an archipelago western to Tonga, previous reports of most likely diagnoses of scurvy due to vitamin C deficiencies 95 , gouty arthritis and DISH 11,12 , point to significant metabolic and nutritional issues afflicting Pacific Islanders thousands of years ago. While we cannot provide one aetiological explanation for the occurrence of bone metabolic conditions in our sample, we propose that a combination of lifestyle factors such as nutrition, and genetic predisposition played a role in the poor Tongan bone metabolism. Theoretical attempts to explain the high incidence of metabolic disease in the Pacific have predominantly focused on genetics (see 98,99 for discussion). "The Thrifty Genotype" hypothesis proposes that thrifty genes, which would have been advantageous during famine events in the human past, drive diabetes and obesity in modern environments 99 . However, considering our data, and the aforementioned archaeological skeletal data from other sites in the region 11,12,[91][92][93][94][95] , it is clear that genes alone cannot elucidate the human obesity trends in the Pacific. Additionally, as noted previously by Gosling et al 98 , thrifty genes cannot alone account for a range of selective pressures that characterised the Pacific islands. Alternative, or complementary, interpretations to consider include early childhood exposure to other diseases 91,92,100 , compromising adult immunity which can equally result in altered bone turnover and poor skeletal mineralisation 100,101 . Ongoing macroscopic pathology examination of the Talasiu sample will provide further data and help shed light on possible incidence of conditions such as DISH or gout, expanding our reported microscopic bone loss markers. Table 2. Histomorphometry and synchrotron source Fourier-transform infrared microspectroscopy descriptive data per estimated sex (four males and four females). SD: standard deviation, min.: minimum data, max.: maximum data, A-P: antero-posterior, M-L: medio-lateral, H.Dn: Haversian canal density per mm 2 , %Po.Ar: percent area of sample impacted by porosity producing trabecularisation. www.nature.com/scientificreports/ Sex-specific and ageing driven bone fragility. Osteoporosis has a long history of afflicting females more than males due to menopause-driven loss of estrogen 31,32 . Oestrogen plays a key role in inhibiting osteoclast-mediated bone resorption, by regulating osteoclast apoptosis 31 . After menopause, prolonged bone loss occurs leading to increasingly porous and weak bones. Females are biologically disadvantaged because of this phenomenon, with modern postmenopausal women experiencing four times the level of osteoporosis than men 102 . Our data for a sample of archaeological Tongan females appear to match this sex-specific difference in bone fragility, providing another line of interpretation as an alternative or complementary aetiology. All our female samples showed a trabecularisation effect whereby the compact bone exhibited trabeculae-like architecture accompanied by cortical wall thinning (Figs. 1, 2). Female samples exhibiting higher cortical bone loss than male samples as a result of sex differences is consistent with prior archaeological 58 and clinical reports [103][104][105] . The cortical wall bone not impacted by abnormal/trabecularised porosity in the female samples showed a lower density of Haversian canals, which can be interpreted as a proxy for the amount of remodelled bone 58,85 . This can be explained by males possibly experiencing higher mechanically stimulated bone remodelling than females, body size differences, and/or within-sample age differences as noted in prior studies 73,103,104,106 .
Inter-woven with the sex-specific differences in bone microstructure is ageing 65,72,[103][104][105] . It is well understood that the variability of age impacts on bone structure with sex manifests substantially in the femoral cross-section (and in other lower limb bones) 107 . Ageing of bone tissue results in a greater resorption on the endocortical surface, leading to modified long bone cross-sectional geometric properties 107 . In a classic study examining this in a sample of US human cadavers 107 , age-related endocortical resorption manifested both in males and females. However, males appeared to exhibit a simultaneous formation of bone sub-periosteally and resorption endo-cortically, which did not drastically impact bone strength 107 . In females, while the medullary cavity expanded, there was no associated expansion of the sub-periosteal bone, meaning that their cross-sectional properties weakened with age 107 . When considering preindustrial humans such as the Pecos who undertook high levels of physical activity 108 , sex-differences in bone loss while present were not as extensive as those seen in the aforementioned cadavers 107 . In the Pecos, sub-periosteal bone and biomechanical properties of the femur (and tibia) followed a general pattern of increase with age in both sexes 108 . We were not able to undertake a biomechanical analysis of cross-sectional geometry, but the substantial thinning of cortical wall in the Tongan female samples, along with lower Haversian canal densities, could suggest not enough mechanically stimulated bone remodelling around the time of their peak bone mass accrual phase of the life-course (third life decade) 109 . Indeed, sedentary lifestyles are so widespread in Tonga today that several modern intervention efforts targeting adolescents have been unsuccessful 110 . This raises an important consideration for future studies of archaeological bone loss comparing populations from different parts of the world, whereby environmental and lifestyle factors impacting bone building in the early ontogeny will differ in accordance to genetic and cultural determinants 109 .
One of the negative repercussions of overall 'improved' longevity and mortality across contemporary human populations is the advanced age of soft and hard tissues that results in multiple degenerative diseases experienced by the elderly 111 . Bone fragility, increased fracture incidence, and hip-replacement surgeries, are some of the most common issues for the elderly of significantly deteriorated skeletal health. One of our female individuals (ID: Sk3.1) showed highly advanced skeletal deterioration characteristics (e.g. loss of all mandibular teeth antemortem, Figure S2) so she was classified as the oldest, and likely the only elderly, individual in our sample. This implies she might have survived into her sixth life decade (and possibly beyond). The combined age-at-death data and her abnormal intra-cortical porosity, are evidence that, at least some, archaeological Tongan females might have well surpassed the commonly assumed short longevity of humans in the distant past 112,113 . The trabecularisation of the cortical bone in our samples is remarkably similar to histology and microradiography images of midshafts examined in modern Australians aged 89 year old 88 , and 67, 78, 90 year old 114 . This encourages further research combining bone biology, skeletal anatomy, archaeology, and social structures of Tonga to elucidate aspects of care of elderly in past Polynesia 115 .
Archaeological context-specific and methodological limitations of our study mean we cannot provide a single diagnosis and aetiology of the bone loss characteristics reported. Access to hundreds of well-preserved skeletons, as has been the case in some prior archaeological osteoporosis studies 54,55 , is not possible at Talasiu as the site is on a remote island constrained by its land and population size. Given the archaeological age of the Talasiu samples, the preservation of bone is not comparable to modern or post-mortem tissue, and thus cannot be experimentally examined to the same level. Future microscopic research applied to archaeological human remains from the Pacific will hopefully generate more comparative data, ideally using complementary 3D and 2D methods where feasible. While our study is limited by the broad age-at-death ranges, this is an issue impossible to overcome in biological anthropology as a narrow and exact chronological age for archaeological individuals cannot be ascertained from gross osteological analysis alone 116 . The assignment of biological sex is also a probability estimate based on well-established sexually dimorphic features of the human skeleton, which, without future aDNA validation, will be the best sex estimate possible.

Conclusions
Our results are the first microscopic record of cortical bone loss in a sample of archaeological humans from Tonga. Given the small sample size, we suggest caution in the generalisation of our results in regards to the wider archaeological Tongan societies. With larger archaeological samples from across the Pacific islands, a pattern in bone loss may be shown in the future. Nevertheless, our results are evidence that a possible occurrence of bone metabolic conditions in a sample of archaeological Pacific individuals can be detected microscopically. We think this is an important step forward for discussions about metabolic diseases in the past and present Pacific. We discussed several explanations for the observed bone loss markers including: sex-specific driven bone microarchitectural deterioration with the possibility of our sample of Tongan females experiencing hormonal changes www.nature.com/scientificreports/ due to menopause; age-related bone loss similar to that seen in modern aging populations; and lifestyle factors such as poor physical activity and nutrition. Equally, other diseases in the region compromising immunity might have contributed to abnormal bone physiology in our sample. Given how commonplace contemporary metabolic issues, including obesity and diabetes, are in Tonga, our key interpretation is that similar problems might have led to osteoporosis in our sample there ca. 3,000 years ago.

Methods
We implemented an invasive methodology and so were restricted to eight femora (seven right and one left) ( Table 1). The choice of femur side was determined on the basis of macroscopic preservation and availability of midshaft bone for extraction. Brief summarises of gross anatomically visible skeletal lesions per individual are reported in Table S1, with several individuals (6/8) showing evidence for osteoarthritis. While limited, this sample size is comparable to previous research implementing microscopic methods of bone assessment in archaeological samples 77,85 . The biological sex of the individuals represented by the femora had been previously estimated 82,83 following established methods 117 . Four males and four females (including one probable male and two probable females, Table 1) were estimated. The inclusion of 'probable' sexes is standard practice where archaeological human remains that do not show a directional consistency in sexually dimorphic skeletal features 119 . The adult age-at-death was assigned based on morphological features of the pelvic auricular surface 118 . The true chronological age of these individuals cannot be ascertained from bone morphology alone. The skeletal remains represented by the specimens reported here were excavated and analysed with the permission from the Lapaha Community and Nobles, and the Ministry of Internal Affairs (Government of Tonga). The samples are curated at the Australian National University, Canberra until further repatriation notice. Talasiu on Tongatapu in the Kingdom of Tonga is one of the most archaeologically significant sites in Asia-Pacific. It represents some of the earliest occupations of the Neolithic people migrating into the Pacific alongside the Lapita culture some 3200-2850 cal BP 120 . It is a shoreline site located to the north of Lapaha, and is mainly comprised of shell middens and ceramic deposits 115 . The first excavation of Talasiu had taken place in 1957 which was subsequently followed by dating of the recovered shellfish to 2800 + /− 70 BP 120 . Excavations from 2008 onwards revealed burials containing human remains, including burnt remains that provided an insight into early mortuary behaviours in the region 81 . At least 19 late-Lapita/ immediately post-Lapita single to multiple burials have been reported, detailing a possible total number of 45 individuals represented and complex mortuary practices 81,82 . Shell midden samples from Talasiu indicate a sedentary human occupation dated to approximately 2700-2650 cal BP 120 , with more recent calibrations of 2650 BP for the burials 82 . Twenty-one of these individuals were recently examined as part of a multi-isotopic analysis reconstructing the first Polynesian diet 83 . A sub-sample of these Talasiu human individuals with well-preserved femora was studied here to conduct the analysis of bone loss markers. Applying the microscopic methods reported here to bones previously sampled for isotopic analysis, and limiting sample extraction to a cortical quadrant (rather than a complete long bone cross-section) ensures minimal invasion of the archaeological human material as per ethical recommendations 76 .
Samples were collected from the midshaft femur due to its biomechanical versatility, and prior published data reporting abnormal porosity occurring there 103,104 . Prior to sampling, each femur was photographed and measured at midshaft using standard WorkZone calipers and a measure tape to obtain A-P and M-L diameters (in mm), and the midshaft circumference (in mm) ( Table 1). As the femora were fragmented on the distal and proximal ends, we could not measure maximum femoral lengths (Fig. 1). However, the determination of the midshaft location was easily achieved by locating the linea aspera, which is also the region from which samples were extracted (Fig. 1). This followed prior methods examining femur bone histology in archaeological specimens 75,85 , and reports of increased porosity on the posterior aspect of the femur in modern cadavers 103 . Approximately 1 cm thick cortical quadrants were removed using a Dremel 200/2-30 Two-Speed Rotary Tool equipped with a rotary blade. Cutting was performed as per prior methodologies 121 . The samples detached loosely after the blade had reached the medullary cavity having made two parallel longitudinal and transverse cuts on the bone exterior.
The preparation for three different microscopic analyses (2D histology, 3D confocal laser topography scanning, sFTIRM, Fig. 1) included sequential partitioning of bone samples into slices. Using the Dremel blade, each sample was further cut in half to designate an approximately 0.5 cm thick bone section for confocal topography scanning. The remainder of the sample was embedded in Buehler epoxy resin to impregnate the internal bone structure for histology and sFTIRM. Using a Kemet MICRA CUT precision cutter equipped with a 150 mm diamond blade, a ~ 150 μm thin slice of bone was removed from the embedded block and designated for sFTIRM. The remaining portion of the sample was set aside for histological preparation.
The preparation of thin sections followed standard methods applicable to archaeological human bone 75 . The histology surface revealed by cutting on the low speed cutter was glued to a microscope glass slide using Stuk epoxy glue. It was further trimmed on the saw and ground using a series of silicon carbide pads until ~ 100 μm thickness was reached. The sections were then polished using Buehler MicroPolish alumina powder and cleared in an ultrasonic bath. This was followed by dehydration in ethanol, clearing using xylene, and mounting with cover slips. Imaging of the thin sections was undertaken using a high powered Olympus BX53 microscope, a DP74 camera, and associated Olympus cellSens Life Science Technology software. Images were scanned under a 40 × total magnification (17 mm working distance) and 'auto-stitched' using the "Multiple Image Alignment-MIA" tool available from the Olympus cellSens software. Each image was then imported into Adobe Photoshop CC 2014 (replicated recently in Photoshop CC 2020), greyscaled (Black & White), converted to type 16 bit, and thresholded so that Haversian canals throughout cortical bone were 'enhanced' . This essentially converted cortical bone areas into black pixels and porous bone space into empty/white areas of the image (Fig. 2). Thresholding greyscale images for microscopic quantitative analysis is common practice 122,123 and can be applied to threshold In order to differentiate between porosity producing trabecularisation and Haversian canal density the sections had to be manually segmented. We followed descriptions in 88,89 where a delimitation between 'dense' cortical bone and porosity producing trabecularisation cortical space can be estimated by visually separating the two bone matrices (see red dashed line in Fig. 2). As such, five variables as reported in previous bone porosity studies [123][124][125][126] (some acronyms were modified following the nomenclature standards by Dempster et al 127 ) were collected. Total Bone Area (T.B.Ar in mm 2 ) 123 was measured using the ImageJ vol. 1.52 "Polygon" tool by tracing the outer outline endosteal and periosteal borders of the section. Porosity Area (Po.Ar in mm 2 ) 124 was measured using the same tools, but by manually selecting intra-cortical bone regions characterised with abnormal 'giant pores' that originated from within the endo-cortical section areas, which we here define as porosity producing trabecularisation. Cortical area (Ct.Ar) was the cortical bone region comprising cortical walls where no porosity producing trabecularisation (Po.Ar) was noted (Ct.Ar = T.B.Ar -Po.Ar). From these measurements, we calculated porosity producing trabecularisation (%Po.Ar = P.Ar/T.B.Ar × 100). Therefore, if no trabecularised regions are consistently observed, this method allows for a null result of Po.Ar and %Po.Ar. From within the Ct.Ar, Haversian canal number (H.N) was counted manually 125,126 (using the "Multi-Point" tool in ImageJ vol. 1.52 see red dots in Fig. 1). To estimate Haversian canal density (H.Dn) 125,126 , H.N was divided by Ct.Ar to obtain a value per mm 2 . Therefore, we took into consideration two types of bone porosity measures-H.Dn as a proxy for the amount of remodelled bone present in cortical bone, and %Po.Ar which estimates the area of intra-cortical bone affected by trabecularised cortex. Through H.Dn, we worked with an assumption that one Haversian canal represents one secondary osteon. Because cement lines of secondary osteons were not consistently preserved in these archaeological samples, this technique does not account for fragmentary osteons (Fig. 2). The canals that were counted were identified as Haversian canals to the best of our expertise. This excluded micro-features that resembled other pores, which might have occurred as a result of diagenesis. That way, %Po.Ar is exclusively composed of 'giant' and coalescing pores, occurring consistently intra-cortically, originating on the endo-cortical part of the sample. These regions should otherwise be filled with dense cortical bone if the individual did not experience significant bone loss. Both measures can provide an insight into a Bone Multi Cellular unit (BMU) activity tunnelling through cortical bone 128 , whereby H.Dn approximates the number of BMUs that once existed per mm 2 , and %Po.Ar indicates prolonged bone resorption through osteoclast-mediated activity.
To provide a qualitative illustration of bone surface topography in relation to the porosity producing trabecularisation, we scanned two contrasting male and female samples (IDs: Sk3.1 and Sk3.2) using an OLS5000 3D laser confocal microscope. Confocal laser scanning microscopy is a recommended technique for characterising porous structures such as bones 129 . The associated OLS5000 2017 LEXT data acquisition and data analysis application software (Olympus LEXT, Japan) ( Fig. 1) was then used to apply a heat map of false colours that ranged from red to blue indicating highest to lowest depth, respectively (Fig. 3, S10). This resulted in yellow to green colours indicating low topography (Fig. 3, S11) in the male, and red to blue marking high topography in the female (Fig. 3, S10). We used a 5 × LEXT short working distance objective (20 mm) with a 405 nm violet laser that scans 4,096 pixels along the x-axis, with the zoom as at 1.0x. We used the LEXT automated stitching tool to collectively scan six regions of bone located on the mid-line of the cortical area of the posterior bone quadrant. The scanned area of the female sample was 7192.648 μm long (y-axis) and 4872.699 μm wide (x-axis). The area scanned on the male sample was 7185.926 μm long (y-axis) and 4876.763 μm wide (x-axis). The z-plane depth (height of laser reaching the bone surface) was approximately 3741.925 μm.
Each sample was examined for phosphate and carbonate content using sFTIRM 86,[130][131][132][133][134][135][136] . Phosphate was selected because it is necessary for bone metabolism, and occurs in skeletal tissue as part of hydroxyapatite crystals 134 . Carbonates (v 2 and v 3 ) substitute calcium apatite and influence the properties of crystal in bone, and thus influence bone function at the macroscopic scale 133 . Carbonate in bone is mainly A-type whereby it substitutes for phosphate, but it can also be naturally accompanied by B-type (substituting for hydroxide) 133 . Assessing total (A-and B-type combined) carbonate content, in addition to phosphate, and v 2 carbonate and phosphate ratio of integrated areas under trace, can thus provide an insight into the extent to which calcium phosphate has been resorbed through osteoclast-mediated activity 86,135,136 .
The samples were scanned at the IRM beamline at the Australian Synchrotron facility in Melbourne (Victoria). The scanning technique followed settings reported by Vrahnas et al. 86 with the exception that we used an attenuated total reflectance (ATR) attachment that allowed a direct contact with bone surface 85,87 . The synchrotron light source provides a highly intense infrared beam that was used to analyse mineral content of bone in situ and the ATR attachment allows the collection of infrared (IR) data from sample sections that are otherwise too thick for conventional IR transmission analysis 87 . The sFTIRM measurement was performed using a Bruker V80v FTIR spectrometer and a Hyperion 3000 IR microscope, which produced high quality spectra in terms of signal-to-noise ratios at 1-2 μm spatial resolution when the synchrotron IR beam coupled to the ATR crystal 87 .
Each sample had four regions of interest (ROIs) identified on the sub-periosteal area of bone, avoiding the endocortical surfaces particularly that they are so extensively affected by abnormal porosity in the Talasiu samples (Fig. 1). Before placing the samples on the microscope stage, the ROIs were identified from images captured at 2 × magnification under a basic dissecting microscope (Fig. 1). We measured carbonates v 2 (890-850 cm −1 ) and v 3 (1500-1400 cm −1 ), and phosphate (1180-916 cm −1 ), which were then used to calculate carbonate(v 2 ):phosphate ratios (890-850 cm −1 :1180-916 cm −1 ) 85,86,132,133 . Data analysis was undertaken in OPUS 7.2 and 8.0.19 (Bruker Optik, Germany) by creating integration files and extracting peak height (absorbance/AU) and area under the trace values in each spectrum. Each ROI was scanned for 220 spectra, which totalled 880 spectra per sample, totalling 6820 spectra in the entire sample (this excludes the second region of interest, which was unsuccessfully scanned, therefore deducting 220 spectra in data for individual Sk 3.1). www.nature.com/scientificreports/ Once data were inspected, it became apparent that not all spectra were suitable for analysis. This might have occurred as a result of the ATR attachment not making contact with all scanned bone regions. Additionally, as our samples derive from an archaeological context, and are thus impacted by diagenetic processes, diagenetic calcite deposited post-mortem can contribute to carbonate sites of the spectra. To account for these issues, the spectra were inspected and those of poor quality were removed. Our supplemental file (Dataset S1) with raw data reports all spectra collected during the scanning, as well as the datasets following inspection. The inspection involved removing phosphate peak values (and corresponding integrated areas under the peak) of < 0.2. All carbonate spectra values in the negative range to 0.001 were also removed. To further assess for possible diagenesis impacting the carbonate data (sites where calcite can occur), statistical correlations were performed to check how well carbonate v 2 and v 3 data aligned within each specimen (Table S6, S7). Where no strong correlations were identified, this was taken as an indication that diagenesis might have impacted the data as v 3 and v 2 values are scattered randomly. We also visually examined the sites of v 3 for a strong peak at 1427 cm −1 indicating presence of calcite, which should otherwise be replaced by a trough in apatite spectra (creating a doublet) 137 ( Figure S3-S9). Overall, the inspection reduced the number of suitable phosphate spectra to 3344 in males and 844 in females. Carbonate spectra were minimum 1676 (v 2 ) and 2159 (v 3 ) in females, and 3339 (v 2 ) and 3329 (v 3 ) in males ( Table 2). The carbonate v 2 :phosphate ratios calculated from integrated areas under the trace were possible to compute for 846 in females and 2495 in males. We note that all data for Sk12 appeared unsuitable, and so this specimen was excluded from data comparisons.
All data obtained in our study were firstly summarised descriptively. Descriptive comparisons were made within-sample by contrasting results between the sexes, and identifying whether either of the groups showed relatively higher or lower values of data. Next, as our sample size is small (n = 8), we were limited in the choice of statistical analyses to conduct inferential tests. On this basis, we applied non-parametric tests when comparing the sex groups (using a Mann Whitney U test 138 ) and a Spearman's Rho correlation to test if %Po.Ar was significantly associated with cortical thinning. All inferential statistical testing was conducted in IBM Statistical Package for Social Sciences 26 (SPSS) software. For the sFTIRM spectra inspection through correlations, data normality Kolmogorov-Smirnov tests were performed to check whether non-parametric or parametric correlations tests should be applied (Table S7). This step informed the use of Spearman's Rho correlations as the spectra values were not normally distributed (Table S7, S8). Strong correlations (Rho = 0.68-1.00) 139 were taken as an indication that data can be interpreted to draw conclusions for our research question.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files). www.nature.com/scientificreports/