EXPERT COMMITTEE ON BIOLOGICAL STANDARDIZATION Geneva, 12 to 16 October 2015

Transcription

1 ENGLISH ONLY EXPERT COMMITTEE ON BIOLOGICAL STANDARDIZATION Geneva, 12 to 16 October 2015 Collaborative Study to establish the 1 st WHO International Standard for BKV DNA for nucleic acid amplification technique (NAT)-based assays Sheila Govind, Jason Hockley, Clare Morris and the *Collaborative Study Group Division of Virology and Biostatistics. National Institute of Biological Standards and Control, Blanche Lane, South Mimms, Potters Bar, Hertfordshire. EN6 3QG. United Kingdom. *See Appendix I NOTE: This document has been prepared for the purpose of inviting comments and suggestions on the proposals contained therein, which will then be considered by the Expert Committee on Biological Standardization (ECBS). Comments MUST be received by 14 September 2015 and should be addressed to the World Health Organization, 1211 Geneva 27, Switzerland, attention: Technologies, Standards and Norms (TSN). Comments may also be submitted electronically to the Responsible Officer: Dr M Nübling at nueblingc@who.int World Health Organization 2015 All rights reserved. Publications of the World Health Organization are available on the WHO web site ( or can be purchased from WHO Press, World Health Organization, 20 Avenue Appia, 1211 Geneva 27, Switzerland (tel.: ; fax: ; bookorders@who.int). Requests for permission to reproduce or translate WHO publications whether for sale or for noncommercial distribution should be addressed to WHO Press through the WHO web site: ( The designations employed and the presentation of the material in this publication do not imply the expression of any opinion whatsoever on the part of the World Health Organization concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Dotted lines on maps represent approximate border lines for which there may not yet be full agreement. The mention of specific companies or of certain manufacturers products does not imply that they are endorsed or recommended by the World Health Organization in preference to others of a similar nature that are not mentioned. Errors and omissions excepted, the names of proprietary products are distinguished by initial capital letters. All reasonable precautions have been taken by the World Health Organization to verify the information contained in this publication. However, the published material is being distributed without warranty of any kind, either expressed or implied. The responsibility for the interpretation and use of the material lies with the reader. In no event shall the World Health Organization be liable for damages arising from its use. The named authors alone are responsible for the views expressed in this publication.

2 Page 2 Summary An international collaborative study was conducted to establish the 1 st WHO International Standard for use in the standardisation of Polyoma virus BKV nucleic acid amplification (NAT) technology assays. Two candidate samples of freeze-dried whole BKV virus preparations (subtype 1b-1 and 1b-2), formulated in 10mM Tris-HCl ph 7.4, 0.5% Human serum albumin (HSA), 0.1% D-(+)-Trehalose dehydrate, were analysed by 33 laboratories from 15 countries, each using their routine NAT-based assays for BKV detection. Of the two candidate samples 14/202 and 14/212 the latter was found to be most suitable for use as an international standard. The results from the accelerated thermal degradation stability studies performed at 3 months have demonstrated that the candidate material is stable at temperatures used for storage (- 20 C) and laboratory manipulation (4 C to 20 C), as well 37 C to 45 C reflecting ambient temperature fluctuations encountered during global shipment. Further real-time stability studies will ensue to assess the long-term stability of the candidate. Based upon the conclusion from the dataset received, it is proposed that the candidate sample (14/212; 4092 vials) be established as the 1st WHO International Standard for BKV DNA for nucleic acid amplification technique (NAT)-based assays with an assigned potency of 7.0 log 10 IU/ ml per ampoule.

3 Page 3 Introduction BK Virus is a member of the polyomaviridae family of double stranded DNA viruses. Primary infection is acquired in early childhood and in the majority of cases is asymptomatic. Consequently seropositivity across the adult population is as high as approximately 90% [1]. Following primary infection the virus establishes latency in the kidneys and urinary tract with intermittent reactivation throughout life, where virus is detectable in <5% of healthy individuals [2, 3]. There are 4 main BKV subtypes I, II, III and IV based on nucleotide variation of the VP1 gene that encodes for the viral capsid protein 1. Subtype I is prevalent in most geographical regions with a prevalence of 46-82% throughout the world, whilst subtype IV, the next most prevalent subtype is more commonly associated with East Asian populations, and subtypes II and III are rare [4]. Subtype I can be further divided into sub-groups 1a, 1b-1,1b-2, and 1c on the basis of DNA sequence variation [5]. Each have been reported to be traceable to a unique geographical location where 1a is most prevalent in Africa, 1b-1 in South-east Asia, 1b-2 in Europe and 1c in North-east Asia [6]. The clinical sequelae of BKV reactivation is confined to the immunocompromised state, such as in renal transplantation and haematopoietic stem cell transplantation (HSCT) patients. Under immunosuppression, latent viral reactivation can result in BKV-associated nephropathy (BKVAN) characterized by interstitial nephritis and/or urinary tract stenosis, affecting up to 10% of patients. This can cause allograft loss in up to 60% of affected kidney transplant recipients [7]. In HSCT patients BKV reactivation can present with haemorrhagic cystitis that can be associated with significant morbidity and mortality [8]. Guidelines for the management of BKVAN in renal transplant patients recommend rigorous viral monitoring for BKV reactivation using quantitative PCR of urine and plasma posttransplantation for specified time points. Viral reactivation is detectable in the urine several weeks before viremia is detectable. A BKV viral load (BKVL) of 4 log 10 /ml (in plasma/serum) for >3 weeks is presumed predictive for BKVAN, upon which a reduction in immunosuppression is recommended. This threshold value has been recommended by the UK Renal Association and by the American Society of Transplantation (AST). An international group convened in 2006 and 2008 to discuss the requirement for the international standardisation of JCV NAT assays, alongside which BKV NAT assay standardisation was also discussed. This included participants from academia (Professor Hans Hirsch, Dr Manfred Weidman), pharmaceutical industry (Biogen Idec, GSK, Millennium Pharmaceuticals), the External Quality Assessment provider QCMD (Quality Control for Molecular Diagnostics) and governmental institutions (NIBSC). Evaluations from QCMD proficiency panels in 2007 and 2008 for both BKV and JCV highlighted large variability in NAT-assay quantitative data, underscoring the need for greater standardisation and the availability of international standards that could be used to calibrate the different working standards used by individual laboratories. The merits of four source materials were considered for use as candidate materials for the preparation of IS candidates. It was concluded that purified virions from BKV infected cell cultures should be used for the preparation of a BKV candidate international standard (Manuscript in preparation Paola Cinque, Proceedings of the Second meeting on JCV PCR standardisation: Towards the establishment of International standards). A publication by Hoffman et al further reiterated the need for standardisation. The absence of standardized BKV assays constitutes a major obstacle to the comparison of BK titers measured at different institutions and may severely

4 Page 4 limit the usefulness of generalized quantitative viral-load cutoffs in screening and diagnostic protocols. [9]. The current study describes the preparation and evaluation of two BKV candidate materials, intended for use as primary international standards for NAT-detection assays. They are referred to in the text by the assigned NIBSC codes 14/202 and 14/212, as well as by the alphabetic code given as part of the collaborative study test panel, Sample B and D respectively. Aim of the study The purpose of this study is to establish the 1st WHO International Standard for BK virus DNA for use in NAT detection assays primarily for use in clinical diagnostics. Two candidates were evaluated in a multicentre collaborative study, and the results obtained has enabled a potency estimation to be assigned to the proposed candidate based on the range of NAT assays that are currently in use, represented by the collaborative study datasets. The collated data has also been used to determine the suitability of the candidate material for use as a primary reference material in the calibration of secondary reference materials. Furthermore the evaluation conducted provides a preliminary assessment of the commutability of the candidate formulation when used as a reference standard for a range of matrices following reconstitution in nuclease-free water. Bulk Material and Processing Candidate Standards Two materials were evaluated for consideration as proposed candidates for BKV IS production. The first candidate 14/202 was sourced from Dr JL Murk MD, Medical Microbiology Department of Virology, University Medical Centre Utrecht, The Netherlands. The second candidate 14/212 was sourced from NIBSC, which comprised a virus stock archived in Both proposed BKV standard formulations were cell-free, live virus preparations from productively infected cell culture. The candidate standards 14/202 and 14/212 have both been formulated in universal buffer comprising 10mM Tris-HCl ph 7.4, 0.5% Human serum albumin (HSA), 0.1% D-(+)-Trehalose dehydrate, to permit dilution into a sample matrix pertinent to the end user. Both preparations have been freeze-dried to ensure long term stability of the product. The donated cultured viral stock was obtained from a viral isolate from the urine of a bone marrow transplant (BMT) recipient diagnosed with BK-cystitis. As the donated BKV stock had not been genotyped the whole viral genome was sequenced at NIBSC using Sanger sequencing to confirm the genotype of the isolate as well as assure the sequence integrity of the viral genome post propagation. Primer pairs were designed to amplify ~1000bp regions of the entire viral genome. PCR was performed on viral DNA extracted using the QIAamp Viral RNA mini kit (QIAGEN, Germany) from the donated cell culture supernatant. This extraction kit is recommended for the extraction of all viral nucleic acids. PCR was performed on the Veriti 96 well thermal cycler (Applied Biosystems) using Platinum Taq DNA Polymerase (Life Technologies). Sequencing reactions were performed on purified PCR templates on the 3130 XL Genetic Analyzer (Applied Biosystems). The sequence was identified to be BKV1b- 1 subtype, based on base pair comparison within the VP1 region and showed a 98.9%

5 Page 5 sequence identity with the BKV 1b-1 sequence (NCBI GenBank Accession AB301095). The NIBSC BKV stock material was also subject to full sequence analysis as described above and identified to be BKV1b-2 which showed 98.4% sequence identity to (NCBI GenBank Accession AB301093). Further sequencing is in progress to verify the subgroup allocation. Culture of bulk material A total volume of 100ml (2 x50ml) of viral supernatant was provided by the donating lab, shipped on dry-ice, and stored at -80⁰C until required. Briefly the propagation of BK virus was achieved as summarised by the donating laboratory. BKV culture was performed at 39⁰C without CO 2 using the human foetal lung fibroblast cell line MRC-5 cultured in MEM-Eagle with Hanks salts, 0,084% Na-bicarbonate, 200 mm L-Glutamine and 3% FBS. Culture supernatant was harvested when maximal cytopathic effect (CPE) was visible (after 4 weeks) and cleared from cellular debris by centrifugation for 5 minutes at 380 g (1316 rpm). Subsequently FBS was added to the BKV viral stock to obtain a final concentration of 10% and then stored at -80⁰C. The in-house BKV culture was also performed using MRC-5 cell cultures grown at 37⁰C with 5% CO 2 using MEM (Sigma Cat# G8644), 10% FCS, 2% 200mM L-Glutamine (Sigma Cat# G7513), 1.5% 1M HEPES (Sigma Cat# H0887). Culture supernatant was harvested when maximal CPE was visible, approximately after ~4 weeks and cleared from cellular debris by centrifugation for 15 minutes at 150 g. 70ml of culture supernatant was harvested and FCS was added to the cleared BKV viral stock to obtain a final concentration of 10% FCS which was then stored at -80⁰C until required for bulk preparation. Pre-fill testing The concentration of the BK viral stocks were determined at NIBSC using a CE marked IVD kit, BKV ELITE MGB kit (ELITech, Torino, Italy). Briefly nucleic acid extractions were performed using 140 µl of BKV sample using the QIAamp Viral RNA mini Kit (QIAGEN, Germany). Extractions were performed using the QIAcube, an automated extraction platform (QIAGEN, Germany). Extractions were performed with the inclusion of an internal control (CPE) which was included as positive control for DNA extraction from non-cellular biological samples (ELITech, Torino, Italy). 20µl of the 80µl purified nucleic acid sample was amplified by qpcr using the probe-based quantification BKV ELITE MGB kit (ELITech, Torino, Italy) on the ABI 7900HT Real-time PCR instrument (Applied Biosystems, California USA). Viral quantification was achieved with the inclusion of 4 plasmid based quantification standards in the amplification reaction, to generate a standard curve with a dynamic range of quantifiable geq/ml (ELITech, Torino, Italy). The donating laboratory provided an estimation of ~3x10 8 copies/ml for the viral load of the sample. Using the method described above the donated viral stock was determined to contain a copy number of 3.24 x 10 8 geq/ml (Range 2.89 x x 10 8 geq/ml). The NIBSC cultured BKV was estimated to have a viral copy number of 2.55 x10e8 geq /ml (Range 2.38 x x 10 8 geq/ml). (1 genome equivalent/ ml (geq/ml) is equivalent to 1 copy/ml according to ELItech kit instructions). Preparation of bulk material and evaluation of materials The production dates of each of the BKV candidates were 20/10/14 and 10/11/14. However both candidate bulks were prepared in an identical fashion. The universal buffer 10mM Tris- Cl ph 7.4, 0.5% Human serum albumin (HSA), 0.1% D-(+)-Trehalose dehydrate was prepared at NIBSC. The HSA used in the production of the candidate standards was derived

6 Page 6 from licensed products that was further screened at NIBSC to be negative for anti-hiv-1, HIV-2, HBsAg, and HCV. Frozen aliquots of the viral supernatants were thawed using a 30⁰C water bath prior to dilution into universal buffer. A 100 fold dilution was made in order that the bulk preparation should contain approximately 3.24 x 10 6 copies/ml in the case of the first candidate and 2.55 x10 6 copies /ml in the case of the second candidate, in a final volume of 4.5L of universal buffer mL of each of the liquid bulks was divided into aliquots of 0.25ml, 0.55ml, 0.75ml and 1.1ml in 2ml screw cap Sarstedt tubes and stored at -80⁰C for viral copy determination as well as for inclusion in the collaborative study panel to be tested alongside the equivalent lyophilised preparation. The remaining bulk volume in each case was processed for lyophilisation which was designated NIBSC product code 14/202 for the first candidate and 14/212 for the second candidate. Filling and lyophilisation of candidate standard Lyophilisation The filling and lyophilisation of both bulk materials was performed at NIBSC, and the production summary is detailed in Table 1 for 14/202 and Table 2 for 14/212. The filling was performed in a Metall and Plastic GmbH (Radolfzell, Germany) negative pressure isolator that contains the entire filling line and is interfaced with the freeze dryer (CS150 12m2, Serail, Argenteuil, France) through a pizza door arrangement to maintain integrity of the operation. The bulk material was kept at 18 C throughout the filling process, and stirred constantly using a magnetic stirrer. The bulk was dispensed into 5 ml screw cap glass vials in 1 ml aliquots, using a Bausch & Strobel (Ilfshofen, Germany) filling machine FVF5060. The homogeneity of the fill was determined by on-line check-weighing of the wet weight, and vials outside the defined specification were discarded. Filled vials were partially stoppered with halobutyl 14mm diameter cruciform closures and lyophilized in a CS150 freeze dryer. Vials were loaded onto the shelves at +4 C, then cooled to -50⁰C and held at this temperature for 2 hours. A vacuum was applied to 270 µb over 1 hour, followed by ramping to 100 µb over 1 hour. The temperature was then raised to -15 C, and the vacuum maintained at this temperature for 31 hours. The vacuum was lowered to 30 µb and the shelves were ramped to 25 C over 10 hours before releasing the vacuum and back-filling the vials with nitrogen, produced by evaporation of liquid nitrogen with an analysis of % purity. The vials were then stoppered in the dryer, removed and capped in the isolator, and the isolator decontaminated with formaldehyde before removal of the product. The sealed vials are stored at -20 C at NIBSC under continuous temperature monitoring for the lifetime of the product. Post-fill testing Validation of study samples The freeze-dried candidates (14/202 and 14/212) were tested to determine the homogeneity of the viral contents of the lyophilised material post-production. Briefly lyophilised samples were reconstituted in 1 ml of nuclease-free water (QIAGEN, Germany), mixed gently on a vortex and left for 20 minutes. 140 µl of reconstituted sample was used for extraction and the extracted DNA used for amplification as described for Pre-fill testing. The assessments of residual moisture and oxygen content are critical parameters when considering the stability and shelf life of lyophilized products. Non-invasive moisture and oxygen determinations were made as follows. Vials of excipient only formulations for the

7 Page 7 proposed BKV standards were prepared to be used to compare between destructive and noninvasive moisture analysis by near Infra-Red reflectance (NIR, Process Sensors MT 600P, Corby, UK). Results obtained from the non-infectious samples by NIR would then be correlated to coulometric Karl Fischer (KF, Mitsubishi CA-100, A1 Envirosciences, Cramlington, UK) to give % w/w moisture readings. Moisture determinations were compared against values from a standard curve, which was made using 10 vial replicates of noninfectious excipient only samples, which were subjected to varying exposure times (0, 5, 10, 15, 30, 45, 60 and 90 minutes) to atmospheric air, by removing screw caps and raising the stopper to the filling position for the designated period of time before the stoppers were fully re-inserted and the caps re-sealed. Subsequently, several vials of each time point were tested by coulometric Karl Fischer and the standard curve generated. Then 12 vials of the definitive batch containing lyophilised BKV in excipient formulation were tested by NIR and their moistures assigned based on the calibration curve generated from the data from the noninfectious excipient vials. Oxygen headspace content is an indicator of the success of the nitrogen back-filling process in the dryer and subsequent integrity of the seal on the vials. Oxygen headspace was measured using non-invasive headspace gas analysis (FMS-760 Lighthouse, Charlottesville, VA). This correlates the NIR absorbance at 760nm (for oxygen) based on a NIR laser source and is calibrated against equivalent vials, sealed with traceable oxygen gas standards. Calibration of the unit was achieved using 5ml screw capped vials containing oxygen standards 0% and 20%. Stability assessment of candidate product To predict the stability of the freeze-dried materials, vials of the proposed BKV IS candidates 14/202 and 14/212 are subject to accelerated degradation studies. This entails the storage of multiple vials of each candidate post production at -70⁰C, -20⁰C, +4⁰C, +20⁰C, +37⁰C and +45⁰C for up to 10 years. Periodically 3 vials are removed from each temperature and tested for viral potency using the real-time PCR method described above, to provide an indication of stability at the storage temperature of -20⁰C. 3 tests are performed in the first year and annually thereafter. Study samples A total of 7 study samples coded A-G, were prepared for evaluation in this study (Table 1). Participating laboratories were sent a questionnaire (Appendix II) prior to sample dispatch, to ascertain the types of clinical samples routinely assayed in their laboratory and to determine the quantity of sample required for their extraction methodology. Sample sets were thereby customised to each participant based on the responses received. All participants received the candidate BKV materials in both the lyophilised and liquid state (Candidate 14/202; B and C, Candidate 14/212; D and E). In addition to the proposed IS candidates 3 other samples were also included for testing alongside the candidate materials. These were a plasmid construct encoding the BKV genome excluding a section of the non-coding regulatory region (Sample A); kindly donated by the National Institute of Standards and Technology, Gaithersburg, USA. The donating laboratory provided an estimation of x 10 6 genome copies/ µl. Two patient samples were kindly donated by the Rigshospitalet, Department Clinical Microbiology, Copenhagen, Denmark. 3ml aliquots of urine and plasma obtained from a bone marrow transplant patient were received on dry-ice. The viral copy numbers of the neat

8 Page 8 samples were estimated by the donating laboratory to be 2.3 x 10 7 copies/ml (urine) and 6.5 x 10 5 copies/ml (plasma). Each 3 ml aliquot was diluted into 107 ml of BKV negative urine and plasma respectively and aliquoted into smaller volumes (0.25ml, 0.55ml, 0.75ml and 1.1ml in 2ml screw cap Sarstedt tubes) commensurate to the volumes required by participants to perform duplicate nucleic acid extractions in 3 separate runs as part of the collaborative study. Plasma was obtained from the National Blood Service and urine was obtained from an in house donor. Samples were tested prior to dispatch to study participants to confirm the viral load and homogeneity of the aliquots. All liquid samples were stored at -80⁰C until required for shipment. Design of the study Participants were shipped 6 vials of each sample on dry-ice and asked to perform duplicate analysis of each sample using a fresh vial of each sample for each data point in 3 independent runs. This was with the exception of sample A, where participants received 3 vials, and participants were recommended to use 1 vial per run. (In some instances where insufficient volumes of samples remained, participants received fewer vials, but quantities sufficient for 6 extractions). Study protocol Upon receipt participants were directed to store samples either at -70⁰C (C, E, F, G) or -20⁰C (A, B, D). Participants were directed to reconstitute the lyophilised sample (B and D) in 1ml of nuclease-free molecular grade water for a minimum of 20 minutes with occasional agitation before use. For liquid preparations participants were directed to thaw samples just prior to extraction. Participants performing quantitative analysis, were directed to test samples B and D for the first run, undiluted and in addition at a minimum of 3-4 serial ten-fold dilutions in a single sample matrix commonly used in their laboratory (e.g. urine, plasma etc.). For example, dilutions of 1/10, 1/100 etc. were suggested such that at least 1 of these dilutions should fall into the linear range of quantitation in their assay. For subsequent assays participants were requested to test a minimum of two serial dilutions of sample B and D that fall within the linear range of their assay. Those participants performing qualitative analysis were requested for the first assay to test samples undiluted and then an additional minimum of 7 serial 1:10 fold serial dilutions of Sample B and D in a single sample matrix commonly used in their laboratory (e.g. urine, plasma etc.) in order to determine the end point of detectable viral DNA. Participants were asked to select a single matrix for the dilution of both samples B and D such that the data would be comparable between the two lyophilised BKV candidates. Participants were requested to ensure their data included at least 2 dilution points at which a product was no longer detectable. For the 2 remaining qualitative assays, participants were requested to re-test the dilutions around the assay end point as determined in the first assay, and to include a minimum of two half-log serial dilutions either side of the determined end point dilution.

9 Page 9 Sample A was estimated to be 10 9 genome copies/ ml therefore participants were advised to perform serial dilutions that fell within the linear range of their quantification assay. Participants were asked to only use nuclease-free water for the dilution of this sample. For clinical samples F and G, participants were asked to test these samples neat. A results report form was provided to each participant. This included a sheet to provide details of extraction and amplification processes in the assays performed. Separate sheets were provided to submit data values of each assay performed. An example study protocol is shown in Appendix III. Participants 36 participants were recruited, however 3 of these were unable to proceed with the study due to import constrains. Study samples were sent to 33 participants representing 15 different countries (Appendix 1). Participants were selected from research and clinical laboratories based on recent peer reviewed publications on BKV NAT detection assays. Manufacturers of BKV NAT in-vitro diagnostic (IVD) kits were also included as well as reference and EQA laboratories. All participating laboratories were assigned randomly a laboratory code by which to reference their data thereby assuring laboratory anonymity. Where laboratories submitted more than one dataset they are referred to as, for example 27a, 27b etc. Statistical Methods Qualitative and quantitative assay results were evaluated separately. In the case of qualitative assays, for each laboratory and assay method, data from all assays were pooled to give a number positive out of number tested at each dilution step. A single end-point for each dilution series was calculated, to give an estimate of NAT detectable units/ml. It should be noted that these estimates are not necessarily directly equivalent to copies/ ml. In the case of quantitative assays, results were reported as copies/ml and were used directly in the analysis. For each assay run, a single estimate of log 10 copies/ml was obtained for each sample, by taking the mean of the log 10 estimates of copies/ml across replicates, after correcting for any dilution factor. A single estimate for the laboratory and assay method was then calculated as the mean of the log 10 estimates of copies/ml across assay runs. The overall mean estimates were calculated as the means of all individual laboratories. Variation between assays (intra-laboratory) and between laboratories (inter-laboratory) was expressed as standard deviations (SD) of the log 10 estimates and % geometric coefficient of variation (%GCV) of the actual estimates. The potencies of each sample (A, C, E, F and G) relative to sample B or D, the candidate International Standards, were calculated per laboratory as the difference in estimated log 10 units per ml (test sample candidate standard). Potencies were also estimated for samples F and G relative to sample A. A comparison of log 10 copies/ml results (after correction for dilution) was carried out in Minitab 17 [Minitab. Inc, State College, PA, USA] by fitting a general linear model with laboratory and diluent ( undiluted, plasma, urine, Blood etc.) as factors, with post-hoc Dunnett s test being used to compare results obtained using different diluents with those obtained for undiluted samples.

10 Page 10 Results and analysis Validation of study samples Stability studies Production data for the candidate standards sample B and D validated the CV of the fill mass and the mean residual moisture and oxygen content, which were both determined to be within limits acceptable s for a WHO International Standard (Table 2 and 3). The mean BKV viral nucleic acid load from 12 randomly selected vials containing the candidate IS preparations were tested in duplicate for homogeneity. Each vial was reconstituted using nuclease-free water for 20 minutes with gentle agitation. 140 µl of a 1:10 dilution from each vial was extracted as described above and the extracted DNA used for amplification. An average of each duplicate was then used to determine the average viral copy number (Table 4). A 3 month thermal accelerated degradation assessment was performed on the candidate standards by assessing the change in potency if any, of detectable BKV viral nucleic acid across the various storage temperatures. 1 vial was tested at each of the storage temperatures in 3 independent assays. Each vial was tested at a 10 fold dilution and a further 100 fold dilution. The values obtained at 3 months show no observable drop in potency at temperatures up to +45⁰C (Table 5). The stability analysis using Degtest-R (CombiStats, EDQM) was unable to show any predicted loss of potency. For candidate 14/212 the model showed an apparent upwards trend in activity (+0.006% per year when stored at -20⁰C). The reason for this is not clear. Further analysis are scheduled to be performed which will provide an additional evaluation on the stability and suitability of the candidates for long term use. Data Received Data were received from 33 laboratories from 15 different countries. They were allocated a participant number at random. From the 33 laboratories 35 quantitative datasets, and 3 qualitative datasets were analysed. For quantitative data, participants returned values as copies/ml or log 10 copies/ ml. 2 laboratories provided data in geq/ml which were assumed to be equivalent to copies/ml according to the manufacturer s protocol. Qualitative data was expressed as positive or negative detection. In general, participants performed their experimental runs using 1 assay method with the use of one matrix type for the dilution of Sample B and D. Exceptions were as follows: Participant 10 performed their analysis using two different versions of a commercial assay on a different amplification platform for each. They were given the designation10a and 10b. Participant 13 performed one quantitative assay and one qualitative assay which were given the designation 13a and 13b respectively.

11 Page 11 Participant 21 performed two sets of analysis, using 2 types of extraction kits and corresponding automated extraction platforms. Each was then performed on a different amplification platform and given the designation 21a and 21b respectively. Participant 27 performed amplification reactions using 4 different amplification instruments each of these methods were designated a-d. In addition dilutions of Sample B and D were made in 3 different matrices (Urine, Whole Blood, and Plasma) and assayed on each of the 4 amplification platforms. Therefore mean estimates of sample B or D from this laboratory is represented by the total dataset based on all 3 diluents, and represented as 27a, 27b, 27c or 27d to distinguish each method variation. Participant 35 did not receive sample C. A sufficient number of vials were not available for a comprehensive analysis comparable to other laboratories. Summary of assay methodologies Most participants used commercially available nucleic acid extraction kits, which included: BioMerieux: NucliSENS easymag. MACHEREY-NAGEL: NucleoSpin Blood. Lifetechnologies: ChargeSwitch gdna Serum Kit. Anatolia Geneworks: Magrev Viral DNA/RNA Extraction Kit. ELITechGroup S.p.A: ELITe GALAXY 300 Extraction Kit. Norgen Biotek Corp: Plasma/Serum DNA Purification Kit. Roche: MagNA Pure 96 DNA and Viral NA Small/Large Volume Kit, MagNA Pure LC Total Nucleic Acid Isolation Kit, MagnaPure LC Universal Pathogen kit. QIAGEN: (QIAsymphony DSP Virus/Pathogen Kit, QIAsymphony VIRUS-BACTERIA midi kit, EZ1 DSP Virus Kit, QIAamp 96 DNA QIAcube HT kit, MagAttract Virus Mini M48 kit, QIAamp DNA mini kit, QIAamp Viral RNA kit, QIAamp Viral RNA Mini QIAcube Kit, QIAamp DSP Virus RNA mini Kit, QIA amp DNA Blood Mini Kit. Only1 manual method was performed using Proteinase K/Chloroform-Phenol extraction. 15 of the 33 laboratories used QIAGEN kits, 7 laboratories using BioMerieux kits and 5 using Roche extraction kits. No two methodologies were alike. Of the 33 laboratories 10 performed manual extractions, whilst the remainder performed automated extractions on platforms including: BioMérieux (Nuclisens EasyMAG), QIAGEN (QIAcube, QIAcube HT, QIAsymphony SP, M48 BioRobot MDX, EZ1 Advanced XL), Roche (MagNA Pure LC, MagNA Pure 96), and ELITech Group (ELITe GALAXY). For PCR amplification 35 datasets were quantitative compared with 3 qualitative datasets. 15 laboratories generated data using in-house amplification methods, compared with 18 laboratories generating datasets using commercially available amplification assays. 8 BKV NAT commercial assays were represented in the study; Eurospital (Euro-RT BKV Kit), ELITEch (BKV ELITe MGB kit), Altona Diagnostics (Realstar BKV PCR Kit 1.0), Anatolia Geneworks (Bosphore BKV Quantification Kit v1), and Epoch Biosciences (MGB Alert BK Virus Primer mix ASR), Biomérieux (R-Gene BK virus quantification kit), Luminex (Multicode BK primers (ASR) and MultiCode associated ancillary reagents), and QIAGEN (Artus BK Virus RG PCR assay). In addition the Abbott Molecular Inc, IRIDICA Viral IC Assay Reagent Kit which combines PCR and electrospray ionisation mass spectrometry. Differences in the targeted PCR amplification region were as follows; the T-antigen (small and large) target gene region was used by 11 of the 33 participating laboratories, 9 laboratories used the VP1 region. 6 datasets were derived by amplification of VP2 and VP3 genes, and 1 using the NCCR non-coding region. 6 data sets did not disclose the region of amplification. A range of amplification platforms were also represented by returned datasets,

12 Page 12 which include Applied Biosystems instruments (7500 Fast, 7900HT Real-time PCR Systems), Bio-rad (DX-Real-time, I cycler, CFX connect), Corbett Research QIAGEN (Rotor-Gene 6000, Rotor- Gene Q), Roche (LightCycler 480 II, LightCycler 2), Eppendorf Mastercycler pro S, Anatolia Geneworks (Montania 4896 Real-Time PCR Instrument), Life Technologies (Stratagene MX3000P QPCR system), Cepheid (Smartcycler) and Focus Diagnostics (Integrated Cycler). Estimated potencies of study samples The laboratory mean estimates from 35 quantitative and 3 qualitative datasets are presented in Tables 6 and 7 respectively. Table 6 shows the mean estimates as log 10 copies/ ml from quantitative assays performed for each of the samples received by each of the corresponding laboratories. Samples A to E were assayed by all laboratories (with the exception of laboratory 35 that did not receive sample C, which is denoted by **). Where an * appears under sample F and G in Table 6, these samples were not received by these laboratories based on their response to the questionnaire regarding the types of samples processed routinely in their laboratory. The mean estimates are based on the collective dataset, including data points from undiluted and diluted samples. For table 7 qualitative datasets are presented only for sample B and D the two proposed BKV candidate standards. There is an overall broad range in the estimated viral load estimates across all the assay formats and for most of the samples assayed. The log 10 copies/ml range for sample A (BKV plasmid construct) is , a spread of 4.23 log 10. The log 10 copies/ml range for the first proposed BKV candidate IS (sample B) is combining both the qualitative and quantitative datasets. For the quantitative data alone the range is log 10 copies/ml. The log 10 spread increases from 3.15 to 3.41 when the quantitative data is combined with the qualitative data, a fractional increase as the difference within the qualitative data alone is only 0.78 log 10 copies. The corresponding liquid bulk sample (C), shows the highest log 10 spread of 4.43 (range ), of all the samples assayed by quantitative analysis alone. The difference in the quantitative mean estimate between the lyophilized and liquid equivalent for candidate 14/202 is 0.06 log 10 copies. The second proposed BKV candidate sample D shows a log 10 spread of 4.73 ( ) when combining quantitative and qualitative estimates. The quantitative mean estimates alone range between 5.69 and 8.33 with a log 10 spread of The 3 qualitative datasets alone show a log 10 spread of 2.59 NAT-detectable units/ ml. The liquid equivalent of the second BKV candidate shows a mean estimate range of , a log 10 spread of Here the difference in the quantitative mean estimate between the lyophilized and liquid equivalent for candidate 14/212 is 0.02 log 10 copies. The two clinical samples F and G show mean estimate log 10 ranges of and , equating to a spread of 3.82 and 3.36 log 10 copies respectively. Inter-laboratory variation The overall mean estimates and the inter-laboratory variation for both quantitative and qualitative assays are presented in Table 8. Column n defines the number of datasets used to derive each row of data. Quantitative log 10 copies/ml estimates are provided for all samples, and qualitative estimates for Sample B and D are provided in NAT detectable units/ ml. The overall mean estimate for sample B reported by qualitative assays is 2.06 log 10 lower compared with the overall mean estimate reported by the quantitative assays for sample B ( ). For sample D this difference is 2.50 log 10 copies/ ml between quantitative and qualitative mean estimates ( ). The highest standard deviation is seen with the

13 Page 13 qualitative data for sample D which also has the highest geometric coefficient of variation. This contrasts with the GCV% obtained for the qualitative estimates of sample B which is 152%, some 13 fold less. However between the quantitative mean estimates of the two proposed BKV standards the GCV% are 344% and 306% for B and D respectively. These represent the lowest variation of the whole dataset with the exception of the GCV% obtained for the qualitative assessment of B. The overall mean estimates of the liquid bulk samples C and E are within range of the lyophilized bulk mean estimates showing good agreement. For the remaining samples, sample A shows a mean estimates of 8.97 log 10 copies/ ml, sample F 2.67 log 10 copies copies/ ml and sample G 2.99 log 10 copies copies/ ml. However the standard deviations of the mean estimates are high and the degree of variation between laboratories is also reflected by the high GCV% values obtained. Intra-laboratory variation The intra-laboratory standard deviations of the quantitative log 10 copies/ ml estimates for all of the samples for each laboratory are provided in Table 9. Within most laboratories the SD values are low across the assayed samples. Sample F (clinical urine sample) gives the greatest proportion of the highest SD values across all the laboratories, with 18 of the 29 laboratories giving an SD > 0.2, which represents 62% of the dataset. For the other samples the proportion of laboratories with SD values greater than 0.2 were considerably lower (A: 7/35, B: 10/36, C: 5/35, D: 9/36, E: 8/36 and G: 8/29). The highest standard deviation of 2.25 log 10 copies/ ml is seen with sample E for laboratory 13a. They also have the highest SD value for sample F but the SD values across this laboratory for the remaining samples are far lower. Comparison of laboratory reported estimates Figures 1, 3, 5-7 show histogram representations of the quantitative laboratory mean estimates for sample A, C, E, F and G respectively in log 10 copies/ ml. For samples B and D, figures 2 and 4 show the histogram representations of the laboratory mean estimates represented in NAT detectable units/ ml and log 10 copies/ ml for qualitative and quantitative assays respectively. Each laboratory dataset is shown by the assigned laboratory number inside each box. Where laboratories have provided more than one dataset a further designation of a or b alongside the laboratory number has been given. The mean estimates of each sample are plotted as log 10 copies/ ml against the frequency of the estimated mean values. In Figure 2 and 4 that represent each of the BKV candidates quantitative assay estimates are shown in the unshaded boxes and the qualitative datasets in shaded (red) boxes. Each histogram provides a representation of where each laboratory lies in the distribution of the total dataset for each sample, highlighting positioning in relation to the consensus mean estimate. For sample A whilst there is a spread of 4.23log 10 in the mean estimates across the laboratories, 15 datasets lie at the overall mean estimate, representing 43% of the total datasets showing very good agreement (Figure 1). (Laboratory 32 is not represented in Figure 1 as their amplification targeted the NCCR gene target region which was omitted from the plasmid construct). For sample B, 47% of datasets lie at the overall quantitative mean estimate again showing very good agreement in almost half of the datasets (Figure 2). Datasets at the lower end of log 10 scale include data obtained from qualitative assays. Figure 3 shows the log 10 mean estimates for sample C the liquid bulk equivalent of sample B. 23 datasets out of 35 sit within the two highest peaks that are around the overall quantitative mean estimate of 6.71 log 10 copies/ ml for sample C. For the second BKV candidate sample

14 Page 14 D a broader distribution of mean estimates is evident, compared with sample B. 23 datasets out of 38 represented by the two peaks in the distribution, report mean estimates that are within ~0.50 log 10 copies/ ml of the overall quantitative mean estimate. The two lowest estimations are represented by datasets obtained by qualitative assays (Figure 4). For the liquid equivalent of the second BKV candidate (sample E) a similar two peak distribution is evident, with 13 datasets out of 35 in agreement, reporting to within 0.50 log 10 copies/ ml of the overall quantitative mean estimate. For the clinical samples, the mean estimates exhibit a much broader bell-shaped distribution of mean estimations, particularly for sample F (urine), compared with samples A-E (Figure 6). For sample G (plasma) 32% of datasets are in agreement, reporting mean estimates to within 0.2 log 10 copies/ ml of the overall mean estimate of 2.99 log 10 copies/ ml (Figure 7). Relative potency estimations Figures 8 and 9 show the laboratory mean estimates of potency relative to the proposed candidate standards sample B or D from quantitative assays, for each of the liquid bulk samples (C and E). These values were obtained by taking the difference between the laboratory derived mean estimates for sample B or D from each of the laboratory derived estimates for sample C and E. Figure 8 shows the mean estimate difference for sample C measured by each laboratory when expressed relative to their mean estimate for the proposed candidate standard B. There is very good harmonization of the dataset with 66% of laboratories in agreement after the relative potency assessment (23 of 35 datasets). 29% of the remaining laboratories fall to within 1 log 10 of the consensus after the potency assessment. Similarly for sample D, 66% of the datasets are in agreement, and 31% of the remaining datasets fall to within 1 log 10 of the candidate standard D, showing harmonisation after the relative potency assessment. Figures 10 and 11 shows the mean estimate differences for sample F (clinical urine sample) when expressed relative to the laboratory mean estimates of either candidate standard B or D respectively. There is a reduction the log 10 spread of the data from 3.82 (Figure 6) to 2.42, a difference of 1.40 log 10 when a relative potency estimation is made with the BKV candidate B. When a relative potency assessment is made using the second BKV candidate sample D there is a fractional change in the log 10 spread of the data from 3.82 ( ) to 3.16 (-6.08 to ). Overall there is some harmonisation of the datasets obtained for sample F where the harmonisation is fractionally better relative to sample D. In Figure 12 when a relative potency assessment of sample G (clinical plasma sample) is made with sample B there is a change from a 3.36 log 10 spread (1.24 to 4.60) (Figure 7) to 1.64 log 10 spread (-4.52 to -2.88). Good harmonization of the dataset is also seen when the mean estimates are expressed relative to sample D (Figure 13). A tighter harmonisation of the data is seen with the relative potency assessment with sample B compared with the relative potency assessment with sample D. As a comparison we also performed a similar analysis to see if agreement of the laboratory mean estimates for sample F and G would improve when expressed relative to sample A, the BKV plasmid construct. For both sample F and G, the potency assessment relative to sample A gave minimal harmonisation of the dataset when compared with sample B and sample D (Figure 14 and 15).

15 Page 15 Assessment of diluent effects Figure 16 shows the laboratory mean estimates for sample B in log 10 copies/ ml using data obtained following the dilution of the reconstituted lyophilised candidate. Both clinical samples and non-clinical samples were used to perform dilutions. Each diluent used is represented by a colour/shade of grey. At the consensus plasma is represented in 9 out of 20 values. Sample D also was subject to dilution into multiple diluents and the mean estimates from the diluted values are shown in Figure 17. At the consensus plasma is represented 12 out of 16 values. Since the proposed IS preparation is intended for use using multiple diluents, Figure 18 and 19 show a post-hoc Dunnett s analysis to establish the effect, if any, on the mean estimate of Sample B and D depending on the diluent used. The results obtained using different diluents are compared with those obtained for the undiluted sample (reconstituted in water). All diluents used for sample B or D are listed on the y-axis. For both candidates only PBS shows a significant difference from the mean of the undiluted sample. Plasma is represented by the greatest number of datasets showing the least difference from the mean of the undiluted sample for both B and D.

16 Page 16 Discussion The proposed candidate standards B (14/202) and D (14/212) comprise whole virus preparations of BKV type 1b-1 and 1b-2 respectively. Both preparations were derived from viral propagation in a permissive cell line. For both practical and ethical reasons, it would be impossible to acquire sufficient volumes of each clinical matrix relevant for BKV NAT detection, for the production of multiple formulations at the scale required for the production of an international standard. Furthermore there would be little or no consistency between batches of the standard. Therefore BKV viral preparations were grown from cell culture and formulated in universal buffer for further dilution by the end user, using a diluent pertinent to the clinical analyte under investigation. In addition the use of a whole virus preparation allows the candidate standard to be extracted alongside clinical samples. The inclusion of the candidate standard into the workflow of the detection assay thus allows for the standardisation of the entire process. The BKV preparations were not subject to viral inactivation methods, in order that viral particles remain as close to the native state as possible and potentially more comparable to viral particles observed in a clinical specimen. This was also implemented considering the possible changes introduced through lyophilisation of the candidate. The production data analysis of the residual oxygen and moisture content of the lyophilised formulations are both within the acceptable limits for long term stability. The results obtained from the accelerated thermal degradation studies at 3 months indicate that both candidate preparations are stable. Further analysis at future time points will provide an indication of continued suitability for long-term use. The viral copy values obtained for both candidates post-production, show good homogeneity across the vial contents. The mean copies/ ml obtained for 14/202 from the in house analysis (6.53 log 10 copies/ ml) is in agreement with the both the quantitative and combined mean estimate of this candidate (6.62 and 6.46 log 10 copies/ ml) obtained by the collaborative study data. The mean copies/ ml obtained for 14/212 from the in house analysis (6.50 log 10 ) is also in reasonable agreement with the combined mean estimate of this candidate (6.97 log 10 ) obtained from the collaborative study. The overall quantitative mean estimate of the freeze-dried material 14/202 sample B is 6.62 log 10 copies/ ml (SD 0.65 and GCV 344%), this compares well with the quantitative mean of the equivalent liquid bulk (Sample C) 6.71log 10 copies/ ml SD 0.77 and GCV 494%), indicating there was no significant loss in potency upon freeze drying of the candidate. For 14/212 (sample D) the overall quantitative mean estimate for the freeze-dried material is 7.17 log 10 copies (SD 0.61 and GCV 306%), this compares well with the quantitative mean of the equivalent liquid bulk (Sample E) 7.21log 10 copies/ ml SD 0.71 and GCV 415%), again indicating there was no significant loss in potency upon freeze drying of the candidate. Overall there was a good agreement between the quantitative laboratory mean estimates of the candidate materials, with the lyophilised samples showing better agreement in estimation compared with the liquid equivalents. Furthermore good agreement across laboratories was also observed with the plasmid construct. The viral loads of these samples were high and participants were advised to perform serial dilutions to identify the dilutions that fell within the linear range of their quantitative assays. Therefore the mean estimates for the lyophilised and the plasmid sample were obtained from multiple data points. For the liquid equivalents of the candidate standards most data points were obtained from undiluted samples, or where dilutions were performed they were not as comprehensive as