Improvement of sample preparation and DNA-extraction methods on challenging bone samples for a high-throughput workflow

(1)

Improvement of sample preparation and

DNA-extraction methods on challenging bone samples

for a high-throughput workflow

M

ASTER THESIS

Tom Hopman // 11038691 MSc of Forensic Sciences

36 ECTS

September 2020 // April 2021 Date of submission: April 19th₂₀₂₁

Supervisor: T.J. Parsons, PhD

International Commission on Missing Persons Assessor:

Prof. L.M.T. Sijen, PhD University of Amsterdam

(2)

2

Abstract

As large-scale disaster victim identification responses are regularly occurring and challenging events, forensic scientists and DNA specialists continually work to improve the DNA identification processes from challenging bone samples. Therefore, the International Commission on Missing Persons (ICMP) have developed a powerful SNP assay to identify victims from casualties with the use of MPSplex. This study assesses successful DNA extraction methodologies from ancient and forensic fields to improve the recovery of short DNA fragments for SNP assays with the aim for automated DAN extraction protocols for a high-throughput workflow. Extractions are quantified with real-time PCR and protocols are assessed based on their efficiency of time, costs, throughput, and contamination risks. For downstream analysis, STR profiles were obtained and compared between the methods. The aDNA extraction protocol of Dabney et al. (2013) resulted in overall higher DNA recovery per gram on single extracts. Implementing this method in the workflow of the ICMP would reduce costs, improve the workflow and likely improve the recovery of short DNA fragments suitable for MPS testing. To obtain comparable DNA concentrations as with the current ICMP method, pooling of the aliquots could be considered but increases contamination risks and exclusion of automation. The protocol by Rohland et al. (2018) based on silica bead purification also showed increased DNA yields per gram and is due to the small volumes most suitable for automation purposes. Both extraction methods resulted in partial to full STR profiles which could be used for kinship analysis. To conclude, this study has shown insights in the ability to combine the extraction methodologies that are used in archaeology in a forensic context. DNA processing and extraction efficiencies on the recovery of small DNA fragments should benefit the entire identification process related to mass casualties, especially where large-scale DVI is required.

(3)

3

1. Introduction

Over the last decades man-made or natural mass casualty incidents are responsible for the disappearance of many people such as the crash of MH17, the 9/11 attacks, the wars in the Middle-East or several tsunamis and natural fires (1). These events have led to the development of large-scale victim identification (DVI) procedures within the forensic community to react to mass fatalities in disasters, to investigate mass graves, and to identify full or partial human remains in forensic casework (2). The identification of those that are missing after disasters is crucial for the recovery and closure for relatives and families which is emphasized by organisations such as the International Commission on Missing Persons (ICMP) (3). The ICMP was established in 1996 by the G7 and is since 2004 extended to a global mandate which focusses on identifying missing persons all over the world regardless of the cause. Due to partnerships with multiple governments, the involvement of civil society organizations and the assistance to law and justice, the ICMP has become a driving force in the forensic society on DVI.

The identification of the victims after these mass-casualty incidents is often based on traditional identifiers such as dental records, fingerprints or physical appearance, and certainly this was the basis for historical identifications (4). Although these identifiers were initially used in many DVI processes, the quality and quantity of the human remains are limiting factors for this type of identification, as well as the sometimes limited availability of antemortem records for comparison purposes (5). Due to the challenges of highly degraded and/or fragmented human remains from mass casualty events and the identification based on circumstantial evidence, several cases of misidentification have been reported such as in Chile where multiple instances of misidentified bodies being returned to relatives have subsequently been discovered (6,7). Facing these limitations during the excavations in the Western Balkans from the conflicts of the 1990’s, the ICMP have developed a successful DNA-led process for bone samples on a large scale to identify the missing through kinship analysis with reference samples from family members (3). This approach has led the ICMP to report more than 20,000 missing individuals worldwide based on DNA matches from their database of family reference samples.

In many cases of missing persons, bone samples have been used successfully as a source of DNA as bone preserves and protects the DNA for a long time against degradation and fragmentation. But often samples that are used for identification of missing persons are not ideal due to the presence of inhibitors as a result of environmental exposure or traumatic processes, such as bone samples recovered from a crash site, due to fire, environmental influences and long post-mortem times (8,9). Bone samples recovered from soil often contain humic acids that are carried over into the DNA extract and inhibit amplification, decreasing the success of gaining a useful DNA profile with PCR amplification (10). Especially in bone samples that have been exposed to challenging environmental conditions, several oxidation processes are also likely to damage the DNA structure (11). Therefore there is a high importance, in extraction methodologies used for challenging bone and teeth samples, to remove as many inhibitors as possible, as well as to obtain the maximum yield of preserved DNA (12). Besides inhibition, fragmentation of DNA in a sample, i.e. degradation, is a challenging factor in the success of useful DNA recovery from bones. Fragmentation results in shorter lengths of DNA which are more difficult to recover and can be too short to be used as effective template material for profiling and kinship analysis (13).

However, these physical and chemical barriers in bone that protect the DNA from environment and microbial assault also hinder the access of reagents in the extraction process (14). Bones are constructed from organic and inorganic chemical bonds that are responsible for calcification, the strong structures and long-term DNA survival within these structures. Studies have shown that DNA is preserved in the mineral structures of hydroxyapatites and in collagen spaces in between these mineral structures (15). A study of Salamon et al. (2005) demonstrated that indeed DNA is well preserved in these crystal aggregates of human bones and extracted relatively more endogenous DNA from these niches than when all bone powder was used (16). These physical barriers make it challenging for forensic investigators to breakdown the chemical structures of the skeletal remains without damaging the DNA. Chemically mediated bone demineralisation and protein lysis is required to make the DNA accessible and successfully extract it from bones. A commonly used reagent for decalcification is EDTA (ethylene diamine tetra acetic acid) which chelates the Ca2+_ions

(5)

5 (17). Secondly, it inhibits DNase and RNase by binding the Mg2+_{ions which are required for those enzymes}

to break down the DNA and RNA. Decalcification reagents are often combined with a proteinase K buffer to convert the collagen into gelatin to make it soluble in the buffer and to promote DNA release (16).

Over the years, different methods, techniques and modifications have been developed to enhance DNA extraction from challenging skeletal remains (18). Extraction plays a key role in DNA recovery and generally consists of the following steps (19). First, the breakdown of the collagen matrix during demineralization, with an EDTA-based buffer, which allows the DNA to enter the solution and makes it accessible for purification. Secondly, the lysate will be incubated with a highly salted binding buffer that facilitates binding of the DNA to a silica medium, such as coated magnetic beads or within a spin column. The sample is then washed using an ethanol-based buffer to minimize inhibition carry-over alongside the extracted DNA and to remove cellular debris. Finally, the sample is eluted from the binding medium used in a low TE buffer. This low TE buffer inhibits DNases and RNases to suppress DNA and RNA degradation which makes it suitable for long-term DNA storage.

Fine bone powders are traditionally used to maximize the surface area and enhance access of extraction chemicals (20,21). At the ICMP the bone samples are powdered after manual sanding, several washings and overnight drying. These steps during sample preparation are performed to minimize contamination from exogenous DNA and inhibitory components and allows maximum decalcification of the bone. As these processes depend on time-consuming manual handling of the sample, a few studies have reported different approaches to minimize this labour-intensive process as desired in a high-throughput workflow. A study by Harrel et al. (2018) showed the success of DNA extraction from bone chips instead of bone powder to minimize sample preparation time and avoid bone pulverization (22). An alternative approach is a two-step drilling process with a dental drill, which is less invasive than sanding, and could minimizes intensive manual labour and hands-on time per sample (23,24). Regardless of the preparation method, DNA extraction is a time-consuming and laborious processand the extraction procedure itself takes several hours to days when full demineralization is involved (25–27). Therefore, several studies have reported automated or semi-automated protocols on DNA extraction. One of these protocols was recently published by the Netherlands Forensic Institute (NFI) on a rapid DNA extraction method for bone samples with the Maxwell FSC extraction robot and Promega Bone DNA Extraction kit (21). They developed a semi-automated protocol based on partial demineralization to shorten the process of DNA profiling from bones without losing amplifiable DNA and with successful autosomal short tandem repeat (STR) typing as a result. Here, samples from different burial times (4-44years) were assessed. With this approach, DNA was extracted from bone powder within 1 hour, limited hands-on steps and allows the possibility of scalable automation. Another attempt to shorten bone preparation time, and overall DNA extraction, was reported by Turingan et al. (2020) where partially demineralized hammered bone chips were analysed on the ANDE Rapid DNA Identification system (28). This study showed successful DNA extraction and profiles from fresh bone samples, but more outdated samples (>1 year) were subjected to overnight incubation.

Automation has taken an important stand in recent developments around forensic DNA analysis, as it is time efficient by reducing manual handing, subsequently reducing the costs and minimizes contamination and human errors (18). At high-throughput work environments for DVI, like the ICMP, a transition to automated extraction and DNA analysis would be preferred. This requires major changes in the way DNA is extracted from bone samples now. Currently the in house full demineralization protocol from the ICMP extracts DNA usually from 1g bone powder in 16ml of extraction buffer (25). These large volumes are needed to successfully demineralize this 1 gram of bone powder and ensuring the ratio between EDTA and bone powder to successfully chelate calcium ions. This protocol has been working successfully for many years, but automated systems are not designed to work on these large volumes (21,29,30). Downscaling these volumes also require a substantial reduction in bone powder input to achieve sufficient demineralization with EDTA-based buffers. As automated systems are designed to extract DNA from volumes of lysates and reagents between the 1-2ml, demineralization has to occur within these volumes. For that, bone powder input should be reduced to 50-100mg for successful demineralization in a 0.5-1ml volume of extraction buffer. This reduction results in a 90-95% loss of potential DNA compared to the current amount of bone powder input but insurmountable for the transition to automated systems. To

(6)

6 compensate for this loss, the ICMP is currently assessing different approaches to achieve DNA concentrations that could be used for autosomal DNA profiling which are discussed in this study.

A generally accepted method of DNA profiling is autosomal STR typing which has become the gold standard for human individualization (31). This approach has been very successful in comparing STR-profiles from the missing and close family members on a large scale using bone samples from the former Yugoslavia (3) and for the identification of unknown WWII soldiers (32). STRs can allow identification but require relatively long fragments of DNA which is not always recovered in challenging samples (33). The majority of the commercially available STR amplification kits are designed to amplify DNA fragments of 70 to 500bp (34). Additionally, due to the time elapsed since some disasters, the lack of available close relatives for comparison purposes makes identification with STRs problematic.

A way to improve DNA-led identifications is to utilise alternative approaches to DNA profiling, such as DNA sequencing of single nucleotide polymorphisms (SNPs). As SNPs are single base pair differences, shorter fragments of the genome (<50bp) are utilised which is beneficial in cases that deal with degraded bone samples (35). A study of Xavier and Parson (2017) observed a higher recovery of successful SNP calls compared to STR calls in ancient samples (36). The ICMP has been developing a powerful method with the use of SNPs. Using massively parallel sequencing (MPS), SNPs all over the genome are tested and used for human identification. The MPSplex (referring both to Missing Persons and Massively Parallel Sequencing), has been specifically designed by the ICMP to obtain profiles by using over 1200 SNPs in a high-throughput workflow (37). With the MPSplex, the casualties could be identified from highly degraded bone samples using kinship analysis with high statistical discrimination. This is emphasized by several other studies which have shown the potential for MPS as a powerful tool for the identification of human remains on degraded DNA in forensic applications (38,39). This transition to MPSplex assays requires a different approach to DNA extraction from bones. As for STR analysis, in which the longer DNA fragments were of interest, MPS testing requires shorter DNA fragments during the analysis. To improve SNP assays on degraded samples, effective recovery of shorter DNA fragments is critical for the ICMP which may be achieved with a different extraction method.

The fields of ancient DNA (aDNA) analysis have developed over the last few years several successful DNA extraction methods and techniques which involve SNP analysis (39). Several studies were able to recover short fragments (30-200bp) of genomic DNA from severely degraded ancient samples (e.g., human, megafauna, and pathogen) dating as far back as the Middle Pleistocene (∼700,000 to 126,000 years ago) (14,40–45). Van der Valk et al. (2021) extended this boundary for the oldest genomic DNA recovered past 1 million years while following the DNA extraction protocol from Dabney et al. (2013) (23,46). This protocol is specifically designed for the recovery of shorter DNA fragments, making it particularly suitable for use with highly degraded samples and assays capable of targeting small fragments. An overload of binding buffer causes a shift in the pH to overcompensate for the EDTA in the lysis buffer and facilitate an increased efficiency of DNA binding to the silica membrane compared to other extraction methodologies (47). The results showed a significant increase in DNA yield for fragments between 35-150bp (23). More interestingly for the use on MPSplex, this study showed the ability to recover 94% of the DNA fragments <50bp which previous methods were not able to recover efficiently. Although the methods presented by Dabney et al. (2013) are of high interest for the ICMP, the large volumes (13ml) of binding buffer are a limiting factor if automation comes at hand.

The study of Glocke & Meyer (2017) on aDNA recovery, based on the methods published by Dabney, did show with silica-based protocols that they were able to retrieve DNA fragments for sequencing with the use of two different binding buffers (48). Short (≥35 bp) DNA fragments were recovered by increasing the isopropanol concentration to 40% and in addition the isopropanol content was increased up to 70% to achieve higher DNA recovery of ultrashort (≥25 bp) DNA fragments. Both buffers were assessed by Rohland et al. (2018) on silica membranes and silica beads to study the DNA recovery with different methods (30). Silica beads were found more suitable for the extraction of smaller DNA fragments (20-30bp) than silica membranes (30-50bp). This study showed protocols that are also optimized for automation on liquid-handling systems due to reduced volumes of reagents compared to the original Dabney protocol, provide high DNA yields, and require not more than ~15 min of hands-on time per sample.

(7)

7 In this study the extraction methods from the field of archaeology and forensics are studied to assess and compare the differences in DNA extraction methodologies on degraded bone samples at the ICMP. The aim is to optimize short DNA fragment recovery from highly degraded samples and improve MPSplex analysis and with that, optimize the reagents and conditions that are involved. Secondly, besides the purpose for DNA profiling, protocols are assessed on their adaptability and potential to be implemented in automated DNA extraction to reduce costs, handling time and contamination risks. Other potential benefits include the decrease of co-extracted inhibitors and co-extracted exogenous DNA as abovementioned studies did conclude. Although this occurs in small numbers at the ICMP, the reduction in bone powder input could be beneficial in cases where bone samples are limited. This study contributes in the transition to automated DNA extraction and the improvement of SNP analysis with the MPSplex in a high-throughput environment. These improvements to DNA processing and extraction efficiency should benefit the entire identification process related to mass casualties, especially where large-scale DVI is required.

2. Materials and Methods

2.1. Sample and extraction method selection

In this study three different femur bone samples were used which were previously quantified at the ICMP using Quantifiler (Thermofisher), and then subsequently processed through MPS-testing (MPSplex) and/or STR-profiling (Powerplex 16) (Table 1). The samples originate from excavations performed in the Western Balkans as a result of the conflict during the breakup of Yugoslavia during the 1990’s, circa 20-25 years ago. The samples selected for this study are not yet identified, and additional data obtained from this work could assist in the identification process.

Table 1. Sample selection, representing the original extraction year and results from previous analyses. Sample _numberSample Extraction Year STR results _{(# of SNPs)}MPS results Quantification _{results (ng/μl)}

A 9145021 2013 PP16 (full profile)

Partial profile

(1039/1190) 0.0873

2012 PP16 (full profile) n/a 0.0674

B 9148766 2015 PP16 (full profile) Partial profile _(994/1190) 0.264

C 9149690 2019 n/a Partial profile (1067/1190) 0.0875

2016 n/a n/a 0.332

The samples were cleaned prior to this study, by first sanding them with a mechanical Dremel-drill. Then, chemically cleaned by washing them twice with commercial bleach (5% sodium hypochlorite), twice with deionized water and twice with ethanol before placed in the oven to be completely dried at 50-60°C. The samples were ground into powder using a Waring blender with steel blender cups. The bone powder and fragments were stored at room temperature prior to processing and sufficient homogeneous bone powder was available for all the experiments on the three different samples.

The experiments is this study are based on some previous mentioned established protocols which will be compared with the current in-house full demineralization ICMP method (25). The experiments performed in this study are summarised below (Fig. 1). A downscaled version of the ICMP protocol was studied besides three other methods (i) the original Dabney protocol with MinElute silica membrane purification (23), (ii) an adapted version of the Dabney protocol published by Rohland et al. (2018) with magnetic bead purification (30), (iii) the Promega Maxwell method similar to the publication of Duijs & Sijen (2020) (21). Replicates were performed for each of the extraction methods tested. For the current ICMP method 500mg of bone powder was processed in duplicate, for the other experiment’s aliquots of 50mg bone powder were processed in triplicate. (Detailed protocols are set out in Supplementary Information P1-P6 and Supplementary Table S1).

(8)

8

2.1.1 Current ICMP protocols

Previous studies at the ICMP had assessed the in-house full demineralization protocol with 500mg of bone powder in combination with 30kDa Amicon filters, instead of the initially validated 1g and 100kDa Amicon filter combination(25), and shown some success in reducing inhibition. This approach referred as ICMP-500mg was adopted in this study. To break down the chemical structures of the bone powder, 15ml demineralisation buffer was added (0.5 M EDTA pH 8.0, 1% N-laurylsarcosinate) and 1ml proteinase K (10mg/ml) to 500mg of bone powder and incubated horizontally at 56°C overnight on a rotation platform (90RPM) to reach full demineralization of the sample.

Fig. 1 | An overview of the experimental setup and sample preparation. For each bone 500mg bone powder was assessed

with the current ICMP method ICMP-500mg, 50mg bone chips were assessed with the downscaled ICMP method (ICMP-50mg bone chips), 50mg bone powder was assessed with the downscaled ICMP method incubated at 56°C (ICMP-50mg-56°C) and 3 x 50mg pooled (ICMP-3x50mg), 50mg bone powder was assessed with the adapted ICMP method and incubated at 37°C (ICMP-50mg-37°C), 50mg bone powder was assessed with the Dabney protocol (Dabney-50mg and Dabney-3x50mg), 50mg bone powder was assessed with the Rohland protocol (Rohland-50mg) and 50mg bone powder was assessed with the protocol on the Maxwell FSC (Maxwell-50mg). The Rohland-50mg and Maxwell-50mg protocols utilised silica-coated magnetic beads as the binding medium; all other protocols used MinElute silica columns (QIAGEN).

Fig. 2 | The purification process of the current ICMP method with the Amicon-Ultra-15 filter. First the entire

(9)

9 After incubation, the full volume of lysate was transferred to an Amicon-Ultra-15 tube which concentrates the sample by filtration over a 30kDa membrane and purifies the lysate to reduce the potential co-extracted inhibitors (Fig. 2). During this filtration step, the majority of DNA fragments smaller than approximately 50bp pass through the filter and are discarded as their molecular weight falls below the limit of the filter (49). After this filtration step, 300-450μl concentrated lysate was transferred to a fresh 2ml Eppendorf tube and 1.5ml binding buffer (QIAGEN) was added and mixed. For purification, the mixture was spun through a MinElute spin column with a silica membrane in a three-step process (3x 630μl). The membrane was washed three times with 750μl wash buffer (PE buffer, QIAGEN) and spun dry afterwards. As a final step, the DNA was eluted in a two-stage elution by adding 25μl elution buffer (QIAGEN) to the membrane, incubated for 5 minutes before spun down and the elution steps repeated once for a final eluate volume of 50μl.

2.1.2 Adapted ICMP extraction methods

One of the starting points in the development of an adapted protocol was to reduce the volumes of the demineralization buffer to make it suitable for automation. Based on similar efforts to establish methodologies with smaller amounts, this study will use 50mg bone powder and 400μl of extraction buffer per sample. 2ml tubes with 50mg bone powder and 400μl ICMP extraction buffer were incubated overnight at 56°C (ICMP-50mg-56°C) and at 37°C (ICMP-50mg-37°C) on a rotation platform (90RPM) to assess the performance of the buffer and demineralization at different temperatures (Table 2). The tubes were incubated in two different orientations to assess the influence of movement on the demineralization. Samples were incubated

vertically in a tube-rack, or horizontally in which the tube-rack was rotated by 90° to improve movement of the lysis buffer and bone powder in each sample tube on the rotation platform (Fig. 3).

Table 2. The composition of the reagents in the extraction buffer for ICMP methods.

ICMP-500mg ICMP-50mg Total 16000μl 100% 400μl 100% 0.5 M EDTA pH 8.0 15000μl 94% 375μl 94% Proteinase K (10mg/ml) 1000μl 6% 25μl 6% 1% N-laurylsarcosinate 150mg 3.75mg

Table 3. Overview of the incubation conditions for the adapted ICMP experiments. Experiment Amount of sample Extraction buffer

volume

Incubation

temperature Incubation condition

ICMP-500mg 500mg bone powder 16000μl 56°C horizontal

ICMP-50mg-56°C 50mg bone powder 400μl 56°C vertical

ICMP-3x50mg* 50mg bone powder 400μl 56°C horizontal

ICMP-50mg- bone chips** 50mg bone chips 400μl 56°C horizontal

ICMP-50mg-37°C 50mg bone powder 400μl 37°C vertical

* three aliquots through one MinElute column ** cleaned and uncleaned bone

Fig. 3 | Incubation positions. The left shows the vertical

incubation on the tube rack with limited movement of the contents of the tube. The right the horizontal incubation with improved movement. The arrows indicate the movement direction and space.

(10)

10 After demineralization and incubation, 1600μl binding buffer (QIAGEN) was added to the 400μl lysate in the 2ml tube without any filtration steps required before purification is performed. After visual assessment of the demineralization, the 2ml Eppendorf tube was spun down and the lysate was transferred to a new tube before adding the binding buffer, excluding any leftover bone powder. The remaining binding, purification and elution steps were performed as described in the previous section (2.1.1. Current ICMP protocols). As an additional experiment (ICMP-3x50mg) the lysates of three aliquots were pooled into one MinElute spin column through repeated centrifugation steps (Fig. 4).

Furthermore, the extraction of DNA from bone chips, rather than bone powder, was assessed on two parts of the same bone sample (sample A and C). One part was sanded, the other part kept in the original state and neither sample underwent any additional washes before processing. The samples were wrapped in a polyester cloth and crushed with a hammer on a block made of polypropylene and 50mg of bone chips were collected in triplicate. The samples were processed in the same way as the samples presented with ICMP-50mg but incubated horizontally at 56°C for 18 hours to stimulate movement in the Eppendorf during incubation. To summarize the adapted versions of the ICMP protocol; an overview of the amount of bone powder input, buffer volumes and incubation conditions is presented below (Table 3).

2.1.3 Adapted Dabney protocol

Triplicates of 50mg bone powder were demineralized in 1ml of the Dabney extraction buffer (0.45 M EDTA, 0.05% Tween-20, and 25μl proteinase K (10mg/ml)) prepared fresh on the same day as demineralisation setup. The samples were incubated overnight at 37°C in a vertical orientation for Dabney-50mg-A. After incubation, the samples were visually assessed on demineralization. Then 950μl of lysate was transferred to a 50ml Falcon tube containing 10ml of binding buffer (5 M guanidine hydrochloride, 40% (vol/vol) isopropanol, 0.05% Tween-20) and 400μl 3 M

sodium acetate as in (19). A combination of a Zymo-Spin reservoir of 15ml (Zymo Research) and a MinElute silica spin column (QIAGEN) was constructed (Fig. 5) as described in the original protocol (23). The assembly was placed in a 50ml Falcon tube and the binding buffer plus lysate was transferred to the spin apparatus and spun down for 4 min at 1500RPM, rotated through 90 degrees and then centrifuged at 1500RPM for an additional 2 minutes. The Zymo-Spin reservoir was removed after the entire volume has spun through the column and the MinElute column was transferred to a clean 2ml collection tube. The MinElute membrane was washed twice with 750μl of PE buffer (QIAGEN) and spun dry. A two-step elution was performed, with 25μl of EB buffer (QIAGEN) applied to the MinElute membrane, incubated for 5 minutes before centrifugation to release the eluate from the membrane. The elution steps were repeated once for a final eluate of 50μl. Dabney-50mg-A was repeated with some alterations made to the incubation conditions to increase the level of demineralisation. For experiments 50mg and

Dabney-Fig. 4 | Pooling of three aliquots into one MinElute column.

ICMP-3x50mg, where three aliquots were pooled into one MinElute spin column for a final eluate of 50μl.

Fig. 5 | Spin apparatus. On the top the Zymo

reservoir where a MinElute spin column is attached at the bottom (purple) before constructed into the 50ml Falcon tube.

(11)

11 3x50mg, incubation was performed at 56°C with the sample tubes in a horizontal orientation to increase the movement of bone powder and lysis buffer during the incubation as discussed previously (see Fig. 3). In experiment Dabney-3x50mg, three 50mg aliquots of sample A were demineralized, and the lysates were spun through one Zymo reservoir-MinElute column apparatus, comparable with ICMP-3x50mg (Fig. 6).

2.1.4 Adapted Rohland protocol

An alteration of the Dabney protocol (23) was published by Rohland et al. (2018) (30) where they studied the extraction and binding buffers and subsequently reduced the volumes of the buffers and reagents to allow for a silica magnetic bead purification that has the potential to be automated. The composition of the extraction and binding buffers are similar as described in the previous section (2.1.3 Adapted Dabney protocol). To reduce the volumes, only a proportion of the lysate for the subsequent purification steps was utilized (Fig. 7). In brief, 150μl lysate was transferred to an Eppendorf tube with 1.56ml binding buffer D and 10μl silica bead suspension. The tubes were incubated for 15 minutes at room temperature to allow DNA binding to the silica beads, spun down and washed three times with 250μl PE buffer (QIAGEN) with the removal of the supernatant from the magnetic beads at each step. After the washes, the magnetic beads were air dried for 20 minutes to aspirate any leftover wash buffer. Elution was performed to resuspend the beads with 25μl EB buffer (QIAGEN) and then a 5-minute incubation. The supernatant was transferred to a fresh tube and repeated once for a final eluate of 50μl. In experiment Rohland-50mg-A, the samples were incubated at 37°C vertically as for Rohland-50mg samples were incubated at 56°C horizontally to encourage demineralisation.

2.1.5 Maxwell-based method

A protocol previously studied by Duijs & Sijen (2020) has been adapted to the use of the Maxwell FSC Instrument and Promega protocols according to the manufacturer’s instructions. This protocol (Maxwell-50mg) is based on magnetic silica-bead purification which is comparable to Roland (Rohland-(Maxwell-50mg) but performed with the Maxwell FSC Instrument instead of manually (Fig. 8). In this study 50mg of bone powder was used to allow comparison between the different protocols whereas the manufacturer recommends the use of 100mg bone powder.

Fig. 7 | Purification steps on the Rohland protocol. At first 150μl of lysate will be mixed with 1.56ml of binding buffer D

before adding 10μl of silica bead suspension. The beads will be washed 3-times with PE buffer, afterwards 2x 25μl EB buffer was used to form a final volume of 50μl.

Fig. 6 | Pooling of three aliquots into one MinElute column.

Dabney-3x50mg, where three

aliquots were pooled into one MinElute spin column for a final eluate of 50μl.

(12)

12 In short, 400μl lysis buffer A (Promega) was added to the 50mg bone powder in the Eppendorf tube, vortexed briefly and incubated at 56°C for 2.5 hours. Next, the tubes were vortexed and spun down before the lysate was transferred to fresh tubes as only partial demineralization occurred in this timeframe. 900μl of lysis buffer B (Promega) was added to the lysate to reach a total volume of approximately 1.3ml and mixed briefly. This volume was transferred to the Maxwell cartridge and DNA was extracted using the Maxwell FSC DNA IQ™ Casework Kit (Promega) on the Maxwell FSC Instrument (Promega) according to the manufacturer’s instruction. For the final extraction, an elution volume of 50μl Elution Buffer (Promega) was used.

2.2. Quantification

All extracts were quantified in duplicate by real-time qPCR using the Investigator Quantiplex Pro (QIAGEN) quantification kit (50). Quantiplex Pro (QPP) uses a standard curve from 50ng/µl to 2.5pg/µl. This kit contains region specific amplicons for detection of the multiple PCR products in different dye channels on the Applied Biosystems 7500 Real-Time PCR instrument (51). TaqMan® probes are used which contain fluorescent reporters to detect the amplification and quantities of the targets. For reference, the fragments analysed in QPP are displayed in Table 4. The 91bp human quantification target region is present on several autosomes of the human genome. The male quantification target region is detected as an 81bp fragment which is specific for regions on the Y-chromosome. In addition to these two markers, the kit contains a 434 bp internal positive control (IPC) to detect levels of inhibition within a sample. This IPC is used to compare the standard curve samples, which are non-inhibited, with the sample and calculates an inhibition index based on cycling times. According to in-house ICMP protocols, the threshold for inhibition is set on 22 cycles for the IPC and an inhibition index <1. The last target present in the kit is the degradation target of 353bp targeting the same locus as the 91bp autosomal target. Due to the different sizes of the human and degradation targets, the degradation status of each sample could be assessed as the longer degradation target is more susceptible to degradation than the shorter human target. The degradation index (DI) is calculated by dividing the human DNA marker quantity by the quantity of the degraded DNA marker. According to the manufacturers and in-house ICMP protocols, a DI ≤ 10 is set as a default as it allows differentiation between DNA fragments larger or smaller than 300bp which is the average fragment size of a successful STR profile.

Table 4. Targets and fragment sizes on the Investigator Quantiplex Pro

Target Fragment Size (bp)

Human 91

Male 81

Degradation (same target as human) 353

IPC 434

Fig. 8 | Automated Maxwell protocol. At first 400μl of lysate will be mixed with 900μl lysis buffer B (Promega) before

(13)

13 All QPP quantification performed in this study was done in accordance with ICMP in-house protocols. The result provides the DNA concentration (ng/μl) and the number of cycles for each target undergoes before reaching their plateau phase (52). This information was used to determine the DNA yield (ng DNA/g bone powder) for each DNA extract, this was used to compare the efficiency of the different DNA extraction methods as it is corrected for the amount of bone powder input. Also, the quality of the sample, including the presence of inhibitors and the degree of degradation of the sample was assessed.

2.3. Nuclear DNA STR profiling

Based on the results from the quantification, a set of samples (N=10) was amplified in duplicate using a standard CE-based STR system according to ICMP in-house protocols (Table 5). For the ICMP-500mg, Dabney-50mg and Rohland-50mg for each sample one replicate was selected with the highest DNA concentration to maximize the success of STR profiling. As for Dabney-3x50mg only sample A was extracted using this method so therefore this extract was used for STR amplification. An extraction negative from one of the batches processed was amplified and analysed. Promega’s PowerPlex 21 STR amplification kit was used with 32 amplification cycles according to the manufacturer’s protocol with a PCR volume of 25μl and standardized DNA input (5μl of master mix, 5μl of primer mix, 5μl of PCR-grade water and 10μl of normalized DNA). Normalized DNA concentrations for amplification were aimed on 500pg DNA to prevent overamplification. PCR settings were as follows: initial denaturation at 96 °C for 1 min, 32 cycles of 94 °C for 10 sec, 59 °C for 1 min and 72 °C for 30 sec, final elongation at 60 °C for 10 min. Fragments were separated and detected with the use of an Applied Biosystems 3130xl Capillary Electrophoresis instrument. Afterwards the results were analysed using GeneMapper ID-X.

Table 5. Sample selection per method and absolute concentrations per sample (pg/μl).

Extraction method Sample-ID DNA concentration (pg/μl)

A_2 43.35 ICMP-500mg B_1 420.00 C_1 109.00 A_3 12.15 Dabney-50mg B_3 93.00 C_3 29.65 Dabney-3x50mg A_3 29.10 Negative control 0 Rohland-50mg A_2 1.97 B_3 9.27 C_1 4.02

2.4 Measures against contamination

During this study ICMP-guidelines according to contamination were followed as described in Amory et al. (2019) (20). The extraction experiments were performed with precautions to contamination with modern DNA; appropriate protective gowns, face masks, hair net, latex gloves, foot covers, and sleeves were worn. The equipment and surfaces have been decontaminated with 10% bleach and UV irradiated before use. Bone samples were prepared as described in previous section (2.1. Sample and extraction method selection). Locations of pre- and post-PCR laboratories are separated with different air handling systems. Sample and liquid transfers were performed in separate UV irradiated hoods dedicated for bone sample processing. For each batch, negative controls were included during extraction and PCR-negative controls in amplification reaction.

(14)

14

2.5 Statistical analysis

Averages and standard deviation (SD) are calculated based on the duplicates from the qPCR which were extracted from the triplicates per sample for each method (n=6). To assess significance differences within the same samples but between methods, a One-Way ANOVA has been performed. Where methods were compared per sample, a two-sided paired t-test was performed, while a two-sided unpaired t-test was performed to assess significance between methods for all samples combined. All tests were performed with a significance level of 0.05.

3. Results

This study has processed three degraded bone samples from the Western Balkans and a total of 90 extractions have been performed on different methods. All samples were quantified in duplicate and selected samples were used for STR-profiling (N=10) which was also performed in duplicate.

After extraction, samples were quantified, and the DNA concentrations assessed. All quantification data and comparisons performed were based on the Human marker in the QPP-kit, unless otherwise stated. The original QF quantification results for each sample (Table 1) were obtained with 1g of bone powder input, the other experiments as described in previous sections.

3.1 Overall DNA recovery

The quantity of DNA that is yield from a sample is the most important factor in successful DNA extraction procedures (Table 6). The results from the original QF and ICMP-500mg extractions show that absolute DNA concentrations from single extracts are the highest obtained through any method, especially sample B provided relatively high DNA recovery (420.00pg/μl) (Table 5). The concentrations obtained with the ICMP-500mg experiments, which obtained on average 179.50pg/μl ± 129.26 of DNA, formed the basis to compare the experiment with 50mg of bone powder input with (Fig. 9).

The downscaled ICMP methods and aDNA protocols were expected to show lower absolute DNA concentrations as lower amounts of bone powder (50mg) were used, approximately 10% of the ICMP-500mg. The concentrations differed significantly between the samples due to different sample qualities and degradation but showed overall a comparable pattern (Table 6 and Supplementary Table S2). On average 14.63pg/μl ± 10.23 DNA was recovered with the ICMP-50mg-56°C protocol, which is consistent reduction for all three samples. Less successful results were obtained with the ICMP-50mg-37°C protocol; on average 9.59pg/μl ± 7.22 DNA was recovered which correlates with a 95-96% of DNA loss compared to the ICMP-500mg (90% was expected). As higher DNA concentrations were obtained with incubation at 56°C, during experiment ICMP-3x50mg this temperature was used, and the bone powder was incubated horizontally. The results showed an increase in DNA concentration as sample input has been tripled to 150mg in total

Fig 9. | Average DNA concentrations per sample in pg/μl. The average DNA concentrations from the original quantification (Original QF),

the ICMP-500mg, ICMP-50mg-56°C, ICMP-50mg-37°C, Dabney-50mg, Rohland-50mg and Maxwell-50mg. Error bars indicate the standard deviation (SD) on the three extractions.

(15)

15 (44.08pg/μl ± 2.55). This equals a DNA recovery between 7-12% of the DNA that was recovered in the ICMP-500mg experiment while using 30% percent of the bone powder input (3x50mg).

Among the other 50mg extractions, the aDNA extraction methods from Dabney et al. (2013) and Rohland et al. (2018) displayed higher yields. The results presented and discussed here for the Dabney and Rohland protocols are from the repeated experiments with modified incubation conditions. With the 50mg 41.62pg/μl ± 27.54 DNA was recovered, the pooled results for sample A with Dabney-3x50mg showed a higher concentration 29.10pg/μl. This method recovered 72% of the DNA compared to the ICMP-500mg with only 30% bone powder input (3x50mg). The DNA concentrations obtained from Rohland-50mg resulted in smaller absolute quantification results (4.51pg/μl ± 2.4) which was expected as only 15% of the lysate was processed. The Maxwell-50mg method resulted in even less DNA recovery (0. 8pg/μl ± 0.4), while using 50mg of bone powder input, and showed a recovery of <1% compared to the ICMP-500mg.

3.2 Efficiency of the extraction methods

To assess the efficiency of the methods, normalized DNA yield (ng/g) was calculated from the absolute DNA concentrations and corrected for the bone powder input per sample (Table 6 and Supplementary Table S3). A consistent pattern is shown for all three samples regardless of the level of preservation of DNA in the sample, i.e., sample quality (Fig. 10). The three current ICMP methods, which were tested with One-Way ANOVA, showed a similar DNA yield per gram for sample A and C (p-value >0.05), but for sample B a significant difference due to the low quant scores of the original QF (p-value <0.05). Additionally, a two-sided unpaired t-test was performed to assess the significance of the differences between the extraction methods for each sample. For sample B, no significant difference was shown between ICMP-500mg and ICMP-50mg which shows that the downscaling of the protocol does not affect the efficiency of DNA recovery. The incubation at 37°C did show significant differences for all three samples (p-value <0.05) which makes the ICMP-50mg-37°C less efficient (9.59ng/g) than the same protocol with an incubation temperature at 56°C (14.63ng/g). A similar pattern was shown with the repeated experiments with Dabney-50mg-A and Rohland-50mg-A. By repeating the experiments with a 56°C incubation temperature significantly more DNA was recovered per gram (p-value <0.05) (Supplementary Table S4).

The results from the Dabney and Rohland protocols show significantly superior results when compared with the current ICMP methods (p-value <0.05). For all three samples more than a 2-x proportional gain was observed for the Dabney-50mg and Dabney-3x50mg protocol. These results emphasize the efficiency of the Dabney method which recovered at least twice as much DNA from the same amount of bone powder input. The bead-based protocol from Rohland performed less reproducibly within each sample as shown

Fig 10. | Average normalized DNA yield per sample in ng/g. The average DNA concentrations from the original quantification

(Original QF), the duplicates with ICMP-500mg and triplicates with ICMP-50mg-56°C, ICMP-50mg-37°C, Dabney-50mg, Rohland-50mg and Maxwell-50mg. Normalized DNA yield was calculated from the quantification results from Fig. 9 and corrected for the amount of bone powder input per sample and the error bars indicate the standard deviation (SD) on the three extractions.

(16)

16 by higher SD values. Even though a small amount of lysate is purified, the method shows its ability to obtain higher efficiency than the current ICMP methods for all three samples and shows similar DNA yields per gram as the Dabney methods based on silica membrane purification for sample A and C. The Maxwell-50mg resulted in lower DNA yields for all three samples comparing to the other protocols.

Table 6. Results and averages from quantification with QPP and calculated normalized DNA yield. Averages per sample are based on the duplicates per extraction, averages per method on the three samples. Samples were processed in triplicate unless otherwise stated.

Extraction method Sample concentration of Avg. DNA samples (pg/μl) Avg. DNA concentration of method (pg/μl) Avg. Normalized DNA yield of samples (ng/g) Avg. Normalized DNA yield of method (ng/g) Original QF-1g* A 77.35 183.70 3.87 9.19 B 264.00 13.20 C 209.75 10.49 ICMP-500mg** A 40.25 179.50 3.98 17.94 B 386.50 38.65 C 111.75 11.18 ICMP-50mg-56°C A 3.53 14.63 3.53 14.63 B 31.00 31.00 C 9.37 9.37 ICMP-3x50mg-56°C A 14.25 44.08 4.71 14.53 B 84.20 27.59 C 33.80 11.28 ICMP-50mg-37°C A 1.78 9.59 1.78 9.59 B 21.13 21.13 C 5.85 5.85 Dabney-50mg A 10.87 41.63 10.83 41.73 B 85.42 85.75 C 28.60 28.61 Dabney-3x50mg A 29.10 29.10 9.72 9.72 Rohland-50mg A 1.60 4.51 10.64 29.92 B 8.34 55.22 C 3.59 23.89 Maxwell-50mg A 0.48 0.84 0.48 0.84 B 1.46 1.45 C 0.58 0.58

(17)

17

3.2.1 Bone chip preparation

As an additional experiment, the success of DNA extraction from bone chips was assessed from sample A and C to improve sample preparation methods. Inconsistent results were shown during the DNA quantification of these experiments (Supplementary Table S2). It was expected that bone chips would provide lower DNA concentrations as the reagents for extraction are limited in their ability to demineralize and lyse the bone as equally as with bone powder due to the reduction in surface area of chips relative to powder. For sample A and C, equal DNA yields were observed for the original, uncleaned, bone chips compared to the ICMP-50mg demineralization experiment. For the cleaned bone chips, sample A resulted in a ~7-x gain in DNA yield whilst sample C resulted in a ~2-x loss. Also, the SDs indicated that there was increased inconsistency of results with bone chips, it is unclear what factors could have caused these results.

3.3 DNA quality and STR profiling

To get a better insight into the quality of the samples, the level of degradation was assessed. According to the Quantiplex Pro validation reports from Qiagen, a DI ≥ 10 indicates that the sample is degraded and could reduce the possibility of obtaining a full STR profile (51). All three samples showed different levels of degradation, which was expected (Table 7). According to the results from QPP, sample Bwas most degraded (DI of 104.87 ± 71.43). Sample A (DI of 20.90 ± 10.21) and C (DI of 10.12 ± 4.87) showed lower DIs which indicate that less degradation has occurred. No differences in DI were observed between methods, but due to the fragment sizes in the QPP, it was unable to assess if degradation is independent of the variation in recovered DNA (Supplementary Table S5).

Table 7. Averages of the degradation indexes per sample.

Average DI ± SD

Sample A 20.90 ± 10.21

Sample B 104.87 ± 71.43

Sample C 10.12 ± 4.87

Next, the level of inhibition was assessed as this is not only sample related but also method related. This inhibition index provides valuable information about the carryover of inhibitors per method and the success for downstream STR amplification. The results show that for none of the samples the threshold of 22 cycles for the IPC was exceeded (Supplementary Table S5). At this point it correlates with the previous Quantifiler results (Table 1) which also showed inhibition below the threshold set in the in-house ICMP protocol. These consistent results show that there are no meaningful differences in the extraction methodologies in terms of the level of inhibition. For that, the elimination of the filtration step with the Amicon-Ultra-15 filter did not lead to an increase in the retention of inhibition. The aDNA methods and the ICMP-500mg showed no level of interference for quantification based on the inhibitory substances that could have been carried over. None of the samples which were processed for STR amplification using PowerPlex 21 demonstrated obvious indications of inhibition, consistent with the IPC Ct data.

For downstream analysis, STR profiles were amplified on previous mentioned samples (Table 5). The results from the STR amplification are shown below where the average loci count, and peak height are stated per extraction method (Table 8). As a control, an extraction negative from Dabney-3x50mg was amplified which resulted in an empty profile (Supplementary Figure 4). The difference in sample degradation is noticeable in the obtained STR profiles. Since sample B is most degraded overall more unbalanced profiles are shown with the peak heights noticeably lower for the longer DNA fragments. The overall higher standard deviation for the peak heights for sample B support this visual assessment but did not result in less loci present. For sample A and C, in general more balanced profiles were obtained for all the extraction methods due to the higher quality of the samples.

(18)

18 ICMP-500mg method performed particularly well as expected; for all three samples full profiles were obtained (21 ± 0 loci). The profiles were overall balanced and showed equal heterozygote peaks. Reasonable profiles were obtained with the Dabney-50mg protocol despite the small input of bone powder. In sample A and B there is some dropout on the side of the longer DNA fragments which is usual from degraded samples and therefore some loci are not complete. Also, some low homozygote peaks and imbalanced heterozygote peaks are shown within these two samples. With sample C a full profile was obtained with more balanced peaks due to the higher quality of the least degraded sample. For the pooled Dabney-3x50mg a quite similar profile was obtained as with the ICMP-500mg while using the same volume of DNA extract input for the STR amplification reactions (5μl and 10μl). Finally, STR profiles obtained with Rohland-50mg from the magnetic beads were analysed. Based on the quantification results obtained with this method, lower-level profiles were expected as only 15% of the lysate volume was used for purification. This has resulted in lower peak heights, allelic and locus dropouts, and an average of 17.67 loci per profile despite the maximum input into the STR reactions.

Overall, the different extraction methodologies provided reasonable partial and mostly qualitatively full profiles which could be used for STR kinship analysis (Supplementary Figures 1-3). To illustrate the obtained STR profiles the results of GeneMapper ID-X from sample A will be shown for each method (Fig12A-12D).

Table 8. Results from analysis with GeneMapper ID-X per extraction method. Avg. # Loci ± SD Avg. Peak height ± SD

ICMP-500mg (n=6) _{21 ± 0} _{3351 ± 2717}

Dabney-50mg (n=6) _{20.67 ± 0.47} _{1525 ± 1805}

Dabney-3x50mg (n=2) _{21 ± 0} _{2384 ± 2418}

(19)

19

Fig. 12B | STR profile of sample A extracted with the Dabney-50mg protocol. This graph shows the peaks on 21 loci with the Powerplex 21

and is anlayzed with GeneMapper ID-X with an input of 10μl of DNA eluate.

Fig. 12A | STR profile of sample A extracted with the ICMP-500mg protocol. This graph shows the peaks on 21 loci with the Powerplex 21

(20)

20

Fig. 12D | STR profile of sample A extracted with the Rohland-50mg protocol. This graph shows the peaks on 21 loci with the Powerplex

21 and is anlayzed with GeneMapper ID-X with an input of 10μl of DNA eluate.

Fig. 12C | STR profile of sample A extracted with the Dabney-3x50mg pooling experiment. This graph shows the peaks on 21 loci with the

(21)

21

4. Discussion

4.1 Comparisons of the methods on DNA recovery

The concentration of DNA in the eluate determines the input for STR and SNP profiling and with that the success of kinship analysis. Preferably for successful STR testing, the concentration of the DNA ranges from 50 up to 500pg/μl, as this is normalized before amplification. Extracts with lower concentrations have been amplified but the risk of peak imbalance, allele and locus dropout and peaks below interpretation thresholds increases with reduction of input. For MPSplex testing, 16μl of normalized DNA could be added to the amplification reaction and contains preferably between the 1-5ng of amplifiable DNA (37). To obtain these reasonable SNP assays, DNA concentrations of eluates are preferably between 75-300pg/μl to achieve this. These concentrations are challenging to obtain from degraded bone samples and limiting the success of kinship analysis. As for most cases at the ICMP, sample quantity is no limitation, DNA extractions were usually performed with 0.5-1g of bone powder to support concentrations suitable for STR and SNP profiles.

As expected, a reduction in bone powder input have led to a drastic loss in amplifiable DNA for all the 50mg methods. The adapted ICMP-50mg protocols failed to obtain 50pg at incubation temperatures at 56°C and at 37°C for all samples. Although these protocols have shown an increase in reagent volumes and the exclusion of the Amicon-Ultra-15 filtration, the concentrations are not favourable for DNA extraction and further STR or SNP kinship analysis. As the protocols are similar in time and handling steps, these protocols provide no improvement on time efficiency or contamination risks. When the three aliquots are pooled in the ICMP-3x50mg56°C the DNA concentration was tripled, as expected, but still provided insufficient results compared to other methods. Besides the low quantification, pooling aliquots requires 3-x the time for purification as the three aliquots were spun through one column. The results obtained with the Maxwell protocol were the lowest for DNA recovery on all samples. Although this semi-automated protocol was easy to perform and extracted DNA within the hour from the lysate, these concentrations are not suitable for downstream STR or SNP profiling. A reasonable explanation for the low DNA concentrations for the ICMP-50mg and Maxwell-50mg experiments could be found in the low volumes of extraction buffer. Within these experiments 50mg of bone powder was demineralized in 400μl of extraction buffer. For all these experiments partial demineralization has been observed, also after 18 hours of incubation at 56°C with the ICMP-3x50mg experiment. The underlying cause of this insufficient demineralization is probably due to the small ratio (8:1) of 0.5 M EDTA to mg of bone powder. After a period of time, these small volumes lack unbound EDTA in the buffer to demineralize hydroxyapatite of the bone powder. Longer incubation times or incubation conditions are not of relevance if the EDTA is saturated and unable to chelate any more Ca2+_ions.

The Dabney protocol obtained the highest absolute concentrations and DNA yield per gram for all 50mg experiments on single extracts. With this protocol reasonable amounts of DNA were extracted from the 50mg samples (Table 6). The efficiency of the Dabney protocol is confirmed by the results from STR profiling, full profiles for sample B and C, and 20/21 loci for sample A were recovered while only processing 50mg of bone powder input (Supplementary Table S6). Also, the results from Rohland-50mg with silica bead purification showed high DNA recovery per gram of bone powder input. This protocol extracted DNA from the smallest amount of input, 15% of the lysate was used, which roughly corresponds to 7.5mg of bone powder. STR profiles obtained with this method showed lower peak hights compared to the other methods and resulted in partial profiles ranging from 13/21-20/21 loci. Even though these results were expected to obtain lower quality profiles, this method is supremely suited for automation purposes due to the low volumes and silica-bead based extraction. The efficiency of this method is comparable to the extractions performed with Dabney, due to similar extraction and binding buffers.

The success of these two aDNA methods could be explained by a set of variables during demineralization and extraction steps. Firstly, the incubation conditions and buffer volumes during demineralization play an important role in the success of DNA extraction. As the composition of the lysis buffers used in all of the protocols tested are similar, the volume of the extraction buffer influences the success of demineralisation and DNA extraction as mentioned before. The ICMP-50mg and Maxwell-50mg protocols were incubated in 400μl of extraction buffer, whereas the Dabney-50mg and Rohland-50mg

(22)

22 samples were incubated in 1ml. Due to this higher volume the ratio of 0.5 M EDTA and mg bone powder is increased to 20:1. This increase facilitates more binding of Ca2+_{ions which results in higher decalcification}

of the matrix which increases DNA release. By repeating the experiments of Dabney-50mg and Rohland-50mg where samples were horizontally incubated at 56°C visual assessment concluded full demineralization while incomplete demineralization was observed when samples were incubated vertically at 37°C. Besides this visual assessment, overall DNA recovery increased with the adapted incubation conditions (Supplementary Table S4). Similar results were shown on the ICMP-50mg experiments where incubation at 56°C recovered higher DNA concentrations than with 37°C. Besides temperature, horizontal incubation stimulates the movement of the bone powder and lysis buffer mixture which increased extraction efficiency which was supported by the results on ICMP-3x50mg horizontal incubation (Fig. 3).

Another factor affecting the efficiency of DNA extraction is that after demineralization, different volumes of binding buffers were added to the lysate (Table 9). In the Dabney-50mg and Rohland-50mg methods, the binding buffer was added in a higher 1:10 ratio to the lysate, which may have increased binding efficiencies as in studies by Dabney and Rohland (23,30). This is also in line with the study of Gamba et al. (41) where they obtained a significant higher DNA recovery with this method in comparison with the silica column based extraction methods of Rohland & Hofreiter (27) and Yang et al.(53).

Related to the binding buffer, is the reported efficiency of these methods on the recovery small DNA fragments which are particularly suitable for SNP profiling with the MPSplex at the ICMP. As the protocols from Dabney in previous studies recovered 94% of the DNA fragments <50bp, or even fragments from 35bp, more DNA may have been recovered but due to the limitation of the Quantiplex only DNA fragments from 91bp and up are quantified. Although this study was not able to measure the smaller DNA fragments, according to literature it is likely that the range of these fragments were also recovered during the extractions in this study.

4.2 Implementation and automation

This study aimed to assess the possibilities of implementing a different bone preparation and DNA extraction method that was suitable for automation without losing the efficiency of DNA recovery and the potential to recover small DNA fragments. The aDNA protocols performed well in DNA recovery in smaller volumes of buffers, these protocols are compared to the current ICMP protocols to assess the implementation based on time efficiency, throughput, human handling, costs, contamination avoidance and the amenability to automation.

4.2.1 Sample preparation of bone chips

To shorten and improve sample preparation, the use of bone chips was assessed. Besides the inconsistent results, the methods of hammering the bone and selecting 50mg of bone chips with tweezers was a challenging procedure. Due to the impact of the hammer on the bone, bone chips get easily spread if not paying extreme attention on the procedure. Inter-bone variation was also noticed; sample A was much harder to crush than the bone of sample C due to the denser structure. This made the hammering of the bone an intensive and time-consuming. Overall, this way of sample preparation and DNA extraction is labour intensive and increases the risk of contamination due to the intense fragmentation of the bone. Hammering the bone takes even more time as grinding the bone and weighing out the bone chips was more

Table 9. The ratios between the lysate and binding buffer volumes.

ICMP-500mg ICMP-50mg Dabney-50mg Rohland-50mg Maxwell-50mg

Volume of Lysate 300-450μl* 400μl 1ml 150μl 400μl

Volume of

Binding buffer 1500μl 1600μl 10.4ml 1560μl 900μl

Ratio 1: ~4 1:4 1:10 1:10 1:2.3

(23)

23 time consuming than bone powder. Implementing this method of sample preparation is not recommended in a high-throughput workflow wherein minimizing contamination risk is of extreme importance.

4.2.1 Implementation of the adapted Dabney and Rohland protocol

As current ICMP protocols are successful in the recovery of DNA suitable for STR and SNP analysis, the implementation of the aDNA protocols have to reach similar levels of success or must be suitable for automated extraction. When performing the Dabney-50mg, several benefits arise during the procedure. As a start, bone is successfully demineralized in 1ml of extraction buffer. This indicates that a 20:1 ratio of 0.5 M EDTA to mg bone powder is sufficient to fully demineralize the bone, similar to current the ICMP methods. This reduction is cost effective as less demineralization buffer is needed per gram of bone powder. Besides the reduction in volume, this protocol used in-house made extraction and binding buffers and therefor the exact composition of these buffers is known. If during validation adjustments are preferred, this could easily be done in contrast to the commercially bought buffers from Qiagen. Next, the filtration with the Amicon-Ultra 15 filter was avoided, which minimizes time and the purchase of these supplies. However, Zymo reservoirs and Falcon tubes are required to spin down the 11.4ml of lysate/binding buffer mixture. But considering all costs, Dabney-50mg reduces the costs to €11.34 per extract while the ICMP-500mg costs €25.33 which is a decrease of €13.99 per extraction. To obtain similar DNA concentrations with the Dabney method, theoretically 200mg of bone powder input and 4ml of binding buffer is needed which could be spun though one MinElute column. These manual handlings are comparable to the current ICMP method where 16ml of lysate is concentrated through the Amicon filter and then the resulting 3ml of lysate/binding buffer mixture is spun through a MinElute column in three steps. In this way an equivalent absolute DNA concentration could be obtained but even higher concentrations of small DNA fragments are extracted with the Dabney approach which drastically should improve the success in SNP assays. The costs for a 4ml with 200mg bone powder are similar to current ICMP extractions, €25.50. Besides costs, the Dabney protocol uses one step less in the washing of the MinElute column which slightly reduces handling time and volumes but excludes an entire step in which contamination could occur. At last, currently batches are processed with the QIAcube but are limited to 11 samples and 1 negative control. Batch sizes could be increase up to 19 samples and 1 negative control as one centrifuge could take up to 20 falcon tubes. This increase in batch size, decreases the costs per negative as it is spread over more samples and increases the throughput and effectively saves time. The biggest limitation of the Dabney protocol is the large volume (10ml) of binding buffer which makes it unfit for automated systems, that is why the Rohland protocol is more suitable.

This protocol requires the same binding buffer as with Dabney purification on silica membranes but reduces the volumes and is based on magnetic bead purification which is most suitable for automation purposes. The efficiency of this method has been proven to be successful although low absolute quantification values were obtained due to the small lysate volume that can be purified. Demineralization was performed similar to the previous described Dabney protocol in which volumes and costs were reduced. Due to the small volumes of reagents, overall costs for the buffers range between €4-€8 per extraction which is a significant decrease compared to silica column purification. In this study purification was performed manually to visually assess the samples in between steps and study the workflow. This protocol required several washing steps where extreme care is needed to prevent disturbance of the pellet of magnetic beads. The workflow for manual silica-bead purification is labour intensive, many human handling steps are required and therefore sensitive for contamination. Besides these considerations, DNA loss, because of bead loss, is inevitable during every step of extraction and purification whatever these steps are performed manually or automated. However, the standardisation gained from automation improves consistency and most robots use ring-based magnets so the beads form round the whole edge of the tube allowing the tip to pass through the middle to the bottom of the tube which reduces the risk of pellet disturbance. Overall is this the only protocol, beside the Maxwell-50mg extractions, which is suitable for automation due to the small volumes. The limiting factor is the low absolute DNA concentrations that were obtained with this method. Low quality STR profiles were obtained from the amplifications from these extracts but with a reasonable number of loci correctly called. To improve DNA concentration, pooling the lysates from

Improvement of sample preparation and DNA-extraction methods on challenging bone samples for a high-throughput workflow