Journal of Immunological Methods 306 (2005) 1 – 15

www.elsevier.com/locate/jim

Standardization

Biological standardization of human interferon beta: Establishment

of a replacement world health organization international biological
standard for human glycosylated interferon beta
Anthony Meager a,*, Rose Gaines Das b
a
Division of Immunology and Endocrinology, The National Institute for Biological Standards and Control, Blanche Lane,
South Mimms, Herts EN6 3QG, UK
b
Biostatistics Group, The National Institute for Biological Standards and Control, Blanche Lane, South Mimms, Herts EN6 3QG, UK
Received 20 June 2005; received in revised form 17 August 2005; accepted 25 August 2005
Available online 26 September 2005

Abstract

Human interferon beta (IFN-h) has been developed as a major biotherapeutic agent for the treatment of multiple sclerosis.
Since World Health Organization (WHO) international standards (IS) for IFN-h were established several years prior to the
development of clinical grade IFN-h products, a number of scientific issues with regard to the biological standardisation of
natural and recombinant IFN-h products have emerged. In order to address these issues, an international collaborative study to
evaluate WHO IS and candidate international standards (CIS) of IFN-h was instigated by the National Institute for Biological
Standards and Control (NIBSC) in 2000 and was carried out in the succeeding year. Sixteen expert laboratories from 8 countries
worldwide participated in the study. They performed titrations on 8 different IFN-h preparations, including IS and new CIS, in a
variety of mainly antiviral- but also including some antiproliferative- and reporter gene-assays, and contributed raw data from
these assays to NIBSC for statistical analysis and calculation of potencies. While both intra- and inter-laboratory variation of
potency estimates was evident, overall validity of the study as a whole was clearly shown by comparison of two pairs of internal
coded duplicates, which gave the expected relative potency of 1 and the lowest inter-laboratory variability of potency estimates
in all assay types. The CIS containing Chinese hamster ovary (CHO) cell- or human fibroblast-derived, glycosylated, IFN-h
gave similar low inter-laboratory variation in potency estimates one to another as the coded duplicates, which was significantly
less than to the 2nd WHO IS of IFN-h, human fibroblast-derived, Gb23-902-531. One of these CIS, designated 00/572,
containing CHO cell-derived IFN-h and formulated with both bovine casein and human serum albumin, could be assigned a
potency, consistent for all assay types, of 40,000 international units (IU) per ampoule relative to the IU of the 2nd IS of IFN-h,

Abbreviations: AP, antiproliferative; AV, antiviral; CHO, Chinese hamster ovary; CIS, candidate international standard; ECBS, Expert
Committee on Biological Standardization; EMCV, encephalomyocarditis virus; HSA, human serum albumin; IFN, interferon; IL-6, interleukin-
6; IS, international standard; ISRE, interferon stimulated response element; JapNS, Japanese National Standard; NIAID, National Institute of
Allergy and Infectious Diseases; NIBSC, National Institute for Biological Standards and Control; NIH, National Institutes of Health; RG,
reporter gene; SEAP, secreted alkaline phosphatase; SINV, Sindbis virus; VSV, vesicular stomatitis virus; WHO, World Health Organization.
* Corresponding author. Tel.: +44 1707 641273; fax: +44 1707 650223.
E-mail address: ameager@nibsc.ac.uk (A. Meager).

0022-1759/$ - see front matter D 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.jim.2005.08.007
2 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15

Gb23-902-531. Other CIS containing glycosylated IFN-h, either CHO cell- or human-fibroblast-derived, could also be assigned
potency values that were continuous with the IU of Gb23-902-531 and 00/572. However, greater inter-laboratory variations in
estimates were evident from comparisons of Gb23-902-531 or 00/572 with either the 1st IS for E. coli-derived, non-
glycosylated, IFN-h with serine substitution at position 17 (IFN-h Ser 17 mutein), Gxb02-901-535, or with a CIS (00/574)
containing IFN-h Ser 17 mutein. Indeed, variations in potency estimates for preparations containing IFN-h Ser 17 mutein were
sufficiently large to indicate that assays could distinguish preparations of IFN-h Ser 17 mutein from preparations of
glycosylated IFN-h. Thus, neither the 2nd IS of IFN-h, Gb23-902-531, containing fibroblast-derived IFN-h, nor CIS, 00/
572, containing CHO cell-derived IFN-h, was appropriate for standardisation of preparations of IFN-h Ser 17 mutein.
Conversely, neither the IS of IFN-h Ser 17 mutein, Gxb02-901-535, or a CIS of IFN-h Ser 17 mutein, 00/574, was appropriate
for the standardisation of preparations of glycosylated IFN-h.
CIS 00/572, containing CHO cell-derived, glycosylated IFN-h, was clearly shown to be suitable to serve as a primary
standard for glycosylated forms of IFN-h, especially clinical grade IFN-h-1a products. It was further shown to exhibit high
thermal and long-term stability. Since the CHO cell-derived IFN-h used for preparation of 00/572 was of a greater purity than
the IFN-h used for the 2nd IS of IFN-h, Gb23-902-531, it was recommended by the WHO Informal Consultation on the
Standardisation of Cytokines, Growth Factors and Other Endocrinological Substances, which met in October 2003, that 00/572
should replace Gb23-902-531 as the IS for glycosylated IFN-h. This recommendation was accepted by the WHO Expert
Committee on Biological Standardization (ECBS) at its annual meeting in November 2003 and 00/572 was established as the
3rd IS for human glycosylated IFN-h with an assigned potency of 40,000 IU. As this study identified no advantage to replacing
the existing 1st IS for IFN-h Ser 17 mutein, Gxb02-901-535, WHO ECBS accepted that this should continue to serve as the IS
for this material.
D 2005 Elsevier B.V. All rights reserved.

1. Introduction (fibroblast-derived), Gb23-902-531, serves as pri-

Human IFN-h is a 166-amino acid protein with an
a-helical bundle structure and ~30% sequence
homology to the consensus sequence of human
IFN-a (Taniguchi et al., 1980; Karpusas et al.,
1997; Meager, 1998). Unlike IFN-a, IFN-h contains
a single site, asparagine 80, through which N-linked
glycosylation can occur and which has been shown
to vary according to the cells in which IFN-h is
synthesized (Kagawa et al., 1988; Utsumi et al.,
1989). It is thought that glycosylation plays impor-
tant roles in facilitating IFN-h secretion from cells, in
increasing its solubility and stability, and in its meta-
bolism in vivo (Edy et al., 1977; Knight and Fahey,
1982; Watanabe and Kawade, 1983; McCullagh,
1983; Satoh et al., 1984; Utsumi et al., 1989). How-
ever, non-glycosylated IFN-h, e.g., produced by
genetically engineered E. coli, although more hydro-
phobic than glycosylated IFN-h produced by either
human fibroblasts or genetically modified mamma-
lian cells, retains biological activity although with
lower specific activity (Utsumi et al., 1989; Runkel
et al., 1998).
The WHO ECBS had previously established two
IS for IFN-h. The 2nd IS for glycosylated IFN-h
mary calibrant for preparations of glycosylated
IFN-h. The 1st IS for IFN-h Ser 17 mutein, a stable
recombinant IFN-h molecule expressed in E. coli
following site-specific mutagenesis of the IFN-h
gene (Mark et al., 1984), serves as primary calibrant
for the standardization of preparations of IFN-h Ser
17 mutein (Standardization of interferons, 1987;
Pestka and Meager, 1997). As these IS were estab-
lished prior to the development of recombinant IFN-
h products, particularly CHO cell-derived, glycosy-
lated IFN-h (IFN-h-1a), licensed for the treatment of
relapsing and remitting multiple sclerosis, their suit-
ability to serve as IS for the marketed IFN-h pro-
ducts required evaluation. The WHO 2nd IS for
glycosylated IFN-h, Gb23-902-531, contains human
fibroblast-derived IFN-h that was partially purified
by control pore glass affinity adsorption (De Ley et
al., 1980). The IFN-h content was approximately 1%
of total protein, and the preparation has subsequently
been shown to contain cytokines other than IFN-h,
most notably interleukin-6 (IL-6), which are also
likely to be adsorbed to control pore glass (Content
et al., 1982; Van Damme and Billiau, 1982; Le and
Vilcek, 1989; A. Meager, unpublished results). The
cytokine contaminants present in this IS might mod-
 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15
ulate the activity of IFN-h in human cell-based
bioassays. Some investigators have reported that
inconsistent potency estimates of test IFN-h prepara-
tions may result when this IS is used to calibrate
bioassays and the relative potencies of in-house
dworkingT standards, particularly when different
bioassays are compared. The Informal WHO Con-
sultative Group on Cytokine Standardization consid-
ered this issue at its Third meeting, held at the
National Institute for Biological Standards and Con-
trol (NIBSC), South Mimms, Hertfordshire, UK, 15–
16th May 1997. It recommended that an international
collaborative study be organised to address this issue
and that both existing WHO IS and new CIS pre-
pared at NIBSC should be examined in the study.
NIBSC was requested and agreed to undertake the
organisation of the study, which was endorsed by the
WHO ECBS.
The relatively high hydrophobicity of IFN-h has in
the past often proved problematic for its isolation,
purification and biological standardisation (Knight,
1981; Tan, 1981). IFN-h is known to adsorb rapidly
to untreated borosilicate glass and plastic surfaces,
leading to variable losses of IFN-h from solutions in
contact with such surfaces (De Ley et al., 1980; Van
Damme and Billiau, 1982). Human serum albumin
(HSA), which is often used as an excipient in the
formulation of cytokine preparations and lyophilised
reference reagents and standards, was found to be
relatively ineffective in preventing adsorption of
IFN-h to glass and in preserving its activity in solu-
tion (A. Meager, unpublished results). However,
investigations at NIBSC have shown that addition of
bovine casein, itself a highly hydrophobic mixture of
proteins, can prevent IFN-h adsorption to glass and
significantly improve its stability (A. Meager, unpub-
lished results). Therefore, 4 new CIS for IFN-h were
prepared using mixtures of HSA and bovine casein in
their formulations prior to filling and lyophilisation in
glass ampoules at NIBSC and assessed in a WHO
international collaborative study (2000–2003) to-
gether with the WHO IS for IFN-h, the Japanese
National Standard and one other preparation of IFN-
h formulated with HSA alone. Sixteen expert labora-
tories (Appendix A) from 8 countries participated in
the study and sent back raw data to NIBSC for
analysis. The results, outcome and final conclusions
of the study and details of a replacement IS for
3
glycosylated IFN-h, that was recently established by
the WHO ECBS, are summarized.
2. Materials and methods
2.1. Materials used for the study
Generous gifts of preparations of IFN-h were con-
tributed by Rentschler (Germany), Toray Industries
(Japan), Schering AG (USA) and Biogen Inc.
(USA). Candidate IS were prepared from these
donated materials according to the procedures used
for International Biological Standards (Annex 4, 40th
ECBS Report, 1990).
Briefly, CIS of IFN-h (00/554, 00/572, 00/574
and 00/576) prepared at NIBSC from 1999 were
prepared by mixing the required amount of concen-
trated IFN-h in 4 l of pyrogen-free six-salt phosphate
buffered saline pH 7.0 containing 0.6% human
serum albumin (HSA)1 and 0.3% bovine casein {Cal-
biochem, USA}.2 Each formulated solution was dis-
tributed in 1 ml aliquots into approximately 3700
ampoules. The ampouled solution was lyophilised
and, after secondary desiccation, the ampoules were
sealed under dry nitrogen by heat fusion of the glass
and stored at
20 8C in the dark. Residual moistures
were calculated using the Karl–Fischer method (Eur-
opean Pharmacopoeia, 1992) and were b0.2%. Filling
variation was expressed as the coefficient of variation
by weight of forty ampoules taken at random prior to
freeze-drying and was b0.2%. Preparations of IFN-h
filled at NIBSC from 1995 to 1997 used the same
formulation but without bovine casein and were filled
in 1.0-ml per ampoule. Materials (Gb23-902-531 and
Gxb02-901-535) prepared for the US National Insti-
tutes of Health (NIH) were diluted in 0.1 M sodium
phosphate buffer containing 1 mg/ml HSA {Travenol,
USA} and 5 mg/ml gelatin. These two preparations
were freeze dried in 1.0-ml aliquots with approxi-
mately 3% residual moisture. (Reference reagent
notes 35, 36, and 37, Research Resources Section,
NIAID, NIH, Bethesda, USA; Standardization of
1
Alpha, UK.
2
Candidate IS 00/572 contained a final concentration of 1.875%
HSA as the source material was supplied as an IFN-h preparation
formulated in 2.5% HSA.
4 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15

interferons, 1987). The Japanese National Standard of
IFN-h, J Ref 03, was formulated in a solution contain-
ing 5 mg/ml HSA and 9 mg/ml NaCl, pH 5 and
freeze-dried in 2.0 ml aliquots. Details of the materials
included in the study together with study codes are
given in Table 1.
2.2. Study design and protocol
Each participant (Appendix A) received two sets of
coded ampoules (Table 1). Based on the results of
antiviral assays calibrated with IS Gb23-902-531,
prior estimates of the ampouled contents of these
IFN-h preparations ranged from approximately 2000
to 90,000 IU/ampoule.
Participants were instructed to reconstitute
ampoule contents with 1 ml of sterile distilled water
(except preparation I, 2 ml) and either carry out ti-
trations on the same day or store aliquots of the
solutions at
70 or
20 8C until required. They
were asked to assay the resulting solutions in their
own bioassays. Additionally, participants were asked
to include their din-houseT IFN-h standard in assays of
the contents of coded ampoules. Participants were to
include all of the preparations in each individual
assay, if possible.
For each assay method used, participants were
asked to carry out at least two independent assays
each including as far as possible all of the prepara-
tions to be tested. For this study, assays were con-
sidered independent if they were carried out on
independent occasions and if the dilutions of the
various materials were either freshly made from
reconstituted ampoule contents on the day of assay,
or from the reconstituted solution that had been
stored frozen.
Participants were requested to include no less
than four dilutions of each preparation in the linear
portion of the dose–response curve. Each prepara-
tion must have been included at least in duplicate in
the individual assays. Participants must have shown,
as clearly as possible, how the assay was carried
out. Required details included how the stock solu-
tions were diluted, what dilutions were entered into
the assay and at what positions, if microtitre plates
were used. All raw data (such as scintillation coun-
ter readouts, microtitre plate readouts, etc.) were to
be supplied to the study organisers at NIBSC for
analysis and calculation of relative potencies.
2.3. Assays contributed to the study
Results were contributed from 16 laboratories in 8
countries. These comprised 15 antiproliferative (AP)
assays (4 assay systems), 7 reporter gene (RG) assays
(3 assay systems), and 44 antiviral (AV) assays (14
assay systems). The antiviral assays used a range of
human cell lines (A549, WISH, 2D9, FL, FS4) with
encephalomyocarditis virus (EMCV), vesicular sto-
matitis virus (VSV) or Sindbis virus (SINV) as the
challenge virus. The Daudi cell line was used exclu-
sively for antiproliferative assays. Three different

Table 1
Ampouled materials and their characteristics
IFN-h preparation Ampoule code/(study code) Status Nominal content/ assigned unitage Calculated unitage
Fibroblast-derived IFN-h 00/576 (D & G) CIS 115 nga 33,000 IU*
CHO cell-derived IFN-h 00/572 (B & J) CIS 200 ngb 40,000 IU
CHO cell-derived IFN-h 00/554 (A) CIS 233 nga 55,000 IU*
Fibroblast-derived IFN-h Gb23-902-531 ( F) 2nd IS 15,000 IU (15,000 IU)
Fibroblast-derived IFN-h J Ref 03 (I) JapNS 6300 IU 5400 IU*
CHO cell-derived IFN-h 95 / 786 (H) None 214 nga 30,000 IU*
E. coli-derived IFN-h ser 17 00/574 (E) CIS 1125 ngc 64,000 IUy
E. coli-derived IFN-h ser17 Gxb02-901-535 (C) 1st IS 6000 IU (6000 IU)
CIS, candidate international standard; IS, international standard; JapNS, Japanese National Standard. Protein determinations on the donated
IFN-h preparations were done by a the Lowry method with HSA as reference protein, b by A 290 measurement with extinction coefficient of
1.71 mg/ml/cm, and c by a reverse phase-high performance liquid chromatography (RP-HPLC) method using IFN-h ser 17 as reference protein
(measured by A 280 with extinction coefficient of 1.7 mg/ml/cm). The nominal IFN-h content in the formulated CIS preparations stated in the
table were calculated from the given protein concentrations of the donated materials. Calculated unitage marked * based of an assigned potency
of 40,000 IU for 00/572 and marked y based on the assigned potency of 6000 IU for Gxb02-901-535.
 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15 5

transfected cell lines with different reporter genes cells (but not WISH cells) in laboratory 09, and AV
were used for RG assays. Assays contributed to the assays in laboratory 10, for which responses were
study are summarised in Table 2. visually assessed into one of four categories. For
An assay system is taken as a broadly defined assays with quantitative responses, dose–response
method (AV, AP or RG assay) and a set of specified lines were plotted and examined both graphically
conditions which remain the same (laboratory, cell and using analysis of variance to determine any appar-
line, quality of reagents and glassware, etc.). For ent anomalies and to assess consistency of dose–
example, an AV assay in one laboratory using the response relationships. Less than 0.1% of the total
same cell line but a different virus would give two raw responses were omitted as contributing exces-
assay systems. However, changes such as a new sively to the between-replicate variability, or as giving
batch of the same reagent would not give a different a marked discontinuity of the dose–response line.
system. The designs of assays carried out on microtitre
plates varied. In some assays, different preparations
2.4. Statistical analysis were examined on different microtitre plates. In other
assays replicates of a reference preparation and one or
The majority of assay responses were reported as more test samples were compared with replicates of
quantitative values. Exceptions were AV assays in the reference preparation being included on each
laboratories 04, 15, 16, the AV assays using A549 microtitre plate. In a number of instances, preparations
were replicated on several plates within an assay in a
Table 2 format which effectively gave replicate assays. The
Assays used in the study purpose of this study was to make comparisons
Antiviral assays among the various preparations, and no one compar-
Virus Cell type Number of Laboratories
ison was considered more dimportantT than another.
laboratory code As far as possible, responses have therefore been
EMCV A549 4 08,09,11,16
combined for analysis over all plates within an
EMCV FS4 1 10 assay. Thus, a difference in response between plates
EMCV FL 1 07 has been incorporated into the residual or error var-
EMCV 2D9 1 01 iance of the assay. However, where assays were effec-
VSV FL 1 05 tively replicated using the same stock solutions but
VSV WISH 4 04,09,14,15
SINV FL 2 06,07
different plates or operators, the replicated assays have
been separately analysed to provide an indication of
Other assays within-assay variability.
Assay type Cell type Number of Laboratories For the majority of assays (including the visually
laboratory code assessed AV assays), the dose–response lines could be
Antiproliferative Daudi 4 02,03,13*,15 satisfactorily described using a four-parameter logis-
Reporter gene 293/ISRE-SEAP 1 01 tic function. The responses, transformed to logits
Reporter gene A549/Mx-Luciferase 1 12 using the fitted asymptotes, have been analysed as
Reporter gene Vero/Mx-Luciferase 1 09 parallel line assays using weighted regression and an
Viruses: VSV, vesicular stomatitis virus; EMCV, encephalomyocar- in house program (WRANL; Gaines Das and Tyde-
ditis virus; SINV, Sindbis virus.
man, 1982) to give an assessment of linearity and
Cell types: A549, human lung carcinoma cell line; WISH, human
amniotic cell line; FL, human amniotic cell line; 2D9, human parallelism and estimates of the relative potency of
glioblastoma cell line; FS4, human foreskin cell line; Daudi, the various preparations.
human Burkitts lymphoma cell line; Vero, African green monkey For the AV assays for which only the dlast positive
cell line. wellT or the dwell giving 50% responseT was reported,
Reporter gene: ISRE, interferon stimulated response element;
the geometric mean concentration giving that
SEAP, secreted alkaline phosphatase.
*In this laboratory two processing methods were used to complete response has been calculated for each preparation,
assays, but this was not considered to give two assay methods. and the potencies determined as the ratios of concen-
Therefore, results were combined for analysis. trations giving equivalent responses.
6 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15
2.5. Assessment of assay variability and combination
of estimates
It has been our experience in many assay systems
that the within-assay replication is not representative
of the total between-assay variation. The study design
thus included two pairs of ampoules which were
identical except for code, namely, ampoules of 00/
572 (B or J), and ampoules of 00/576 (D or G). The
extent to which the potency of J relative to B (or of G
relative to D) differs from the expected potency of 1,
or 100%, provides an independent measure of the
within-assay variability which incorporates the entire
assay procedure as applied to the ampouled materials.
Additionally, any consistent difference of these esti-
mates from 100% in a series of assays would suggest
the possibility of bias in the assay system.
Estimates of relative potency have been combined
as geometric means (GM), and comparisons among
the estimates of potency have been made using ana-
lysis of variance of the logs of the estimates. Fiducial
intervals about mean estimates have been based on the
variance of the logs of the estimates combined. The
geometric coefficient of variation (GCV, determined
as exp(s)
1, where s is the standard deviation of the
log potency estimates, multiplied by 100 to give
percent) has been used to provide a summary measure
of the precision of estimates for rapid comparison.
However, the ratios of between-laboratory variances
are given in the text where the variabilities of different
preparations are directly compared.
3. Results
3.1. Dose–response lines
For the many individual assays, the log dose-logit
transformed response lines did not deviate consis-
tently and significantly from linearity and parallelism,
except that consistent statistically significant non-par-
allelism was observed in the AV-FL assays of one
laboratory and the RG assays of two laboratories. The
slopes within each assay method and laboratory were
further examined using analysis of variance to assess
whether there were any consistent trends across
assays. For the AP- and AV-WISH assays there did
not appear to be any consistent overall slope differ-
ences. In the RG assays and the AV-FL assays, there
was a significant tendency for Gxb02-901-535 (C)
and 00/574 (E), both IFN-h Ser 17 mutein prepara-
tions, and in several cases Gb23-902-531 ( F, fibro-
blast-derived IFN-h), to have flatter slopes than the
other preparations. In the AV-2D9 assays, there was a
tendency for J Ref 03 (I, fibroblast-derived IFN-h) to
have a steeper slope than the other preparations. In the
AV-A549 assays there were no overall consistent
difference among slopes, except that Gxb02-901-535
(C, IFN-h Ser 17 mutein) appeared to have flat slopes
in the assays of one laboratory.
3.2. Assessment of assay variability using coded
duplicate preparations
In this study, two pairs of coded duplicates were
included: the preparation coded J was identical except
for code to the preparation coded B (CHO cell-derived
IFN-h, 00/572), and the preparation coded G was
identical to the preparation coded D (fibroblast-
derived IFN-h, 00/576). Individual estimates of the
potency of the coded duplicate samples of 00/572
(sample J as percent of sample B) and of the coded
duplicate samples of 00/576 (sample G as percent of
sample D) are shown in Fig. 1. Estimates of the
potency for the coded duplicates broadly agreed
with their expected value of 100%. Estimates for
individual assays, or sections of assays, ranged from
about 50% to about 200%. The within-laboratory
variability also differed between laboratories, with
the between-assay and/or between-sections of assay
geometric coefficients of variation ranging from less
than 10% in some laboratories to 40% in others.
However, combination over all assays within a labora-
tory gave geometric mean (GM) estimates, for either
duplicate separately, or for the duplicates combined,
which generally differ by less than 10% from the
expected value of 100%. The overall GM of all indi-
vidual estimates of both duplicates by antiviral assays
was 100% (95% confidence limits 98% to 103%) and
the GVC (for laboratory means) was 5%.
The variability (pooled over all assays and labora-
tories) between-assays (within assay systems) was
similar for estimates for J (as % of B) and G (as
% of D). The between-laboratory variability was
significantly larger than the pooled within-laboratory
variability. This, together with the data from some
 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15 7

Fig. 1. The relative potency of the coded duplicate samples of CIS 00/572 (B & J) and CIS 00/576 (D & G). The potency estimates of the
preparation J are expressed as a percentage of the potency estimates of preparation B, and those of G as a percentage of the potency estimates of
D. Unshaded boxes, AV assays; diagonally hatched boxes, AP assays; cross-hatched boxes, RG assays (as illustrated below x-axis). The
numbers refer to the laboratory code. The letters following the laboratory code in the AV assay boxes indicate the cell–virus system, e.g., AE,
A549-EMCV.

assays, suggested the possibility of bias. For exam-
ple, the AV-FL assays of laboratory 05 or the AP
assays of laboratory 13 are instances where estimates
for duplicates were consistently less than or greater
than 100%. The overall agreement of the estimates
for the duplicates with their expected value suggests
that there was no consistent bias across all labora-
tories. Nevertheless, bias may significantly affect a
combined estimate based on a limited number of
laboratories.
3.3. Calibration of coded ampouled preparations of
IFN-b in terms of the 2nd IS of IFN-b, human fibro-
blast-derived, Gb23-902-531 ( F)
Estimates for the coded ampouled preparations of
IFN-h expressed as IU of the 2nd IS for IFN-, fibro-
blast-derived, Gb23-902-531 ( F) per ampoule are
given in Table 3. For each comparison estimates
differed significantly among the AV assays. Notably,
estimates using A549 or WISH cells tended to be
smaller than estimates by virtually any other AV
assay for each comparison.
Comparisons with Gb23-902-531 ( F, fibroblast-
derived IFN-h) for preparations of CHO cell-derived
IFN-h (00/554, A; 00/572, B & J) and fibroblast-
derived IFN-h (00/576, D & G) formulated with
bovine casein and fibroblast-derived IFN-h (J Ref
03, I) each have similar between-laboratory variabil-
ity, some 5- to 6-fold larger than that for the coded
duplicate samples, or some 3- to 5-fold larger if the
low estimates from the A549 assays of laboratories 08
and 09 are omitted (Table 3). Comparison with Gb23-
902-531 ( F) of a preparation of CHO cell-derived
IFN-h formulated without casein (95/786, H) gave
between-laboratory variability some 8-fold larger than
that for the coded duplicate samples. Comparison of
Gb23-902-531 ( F) with J Ref 03 (I, fibroblast-derived
IFN-h) gave between-laboratory variability some 15-
fold larger than for the coded duplicate samples. The
geometric mean potency for J Ref 03 (I) was found to
be 5652 IU in this study, broadly consistent with its
assigned value of 6300 IU.
In contrast, comparisons of the IFN-h Ser17
mutein containing preparations Gxb02-901-535 (C,
1st IS for IFN-h Ser17 mutein) and 00/574 (E) with
Gb23-902-531 ( F, 2nd IS for IFN-h, fibroblast-
derived) in AV assays gave estimates with between-
laboratory variability over 20-fold larger than the
between-laboratory variation for comparisons of
8 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15

Table 3
Laboratory geometric mean potencya of coded ampouled preparations expressed as IU per ampoule of Gb23-902-531 ( F), 2nd IS for IFN-â,
human fibroblast-derived
Antiviral assays
Preparation Ampoule code Geometric mean Geometric mean and GCV(%): 95% confidence limits
and GCV(%): all excluding outliers*
laboratories, n = 14
00/554 A 58,449 (72) 55,008 (n = 11) 21 48,502–62,386
00/572 B/J 42,113 (58) 41,325 (n = 11) 17 37,252–45,842
Gxb02-901-535 C 6029 (77) 5532 (n = 13) 63 4,118–7,430
00/576 D/G 34,671 (55) 33,920 (n = 11) 15 30,855–37,290
00/574 E 65,343 (118) 55,326 (n = 13) 63 41,150–74,386
95/786 H 28,982 (24) 30,161 (n = 11) 24 26,045–34,927
J Ref 03 I 5414 (21) 5652 (n = 11) 18 5070–6300

Antiproliferative assays
Preparation Ampoule code Geometric mean Geometric mean and GCV(%): 95% confidence limits
and GCV(%): all excluding outlier,y n = 3
laboratories, n = 4
00/554 A 42,468 (153) 66,808 (27) 37,034–120,520
00/572 B/J 17,862 (564) 45,970 (11) 35,580–59,393
Gxb02-901-535 C 5512 (167) 8455 (81) 1932–36,994
00/576 D/G 14,580 (577) 37,777 (24) 21,979–64,929
00/574 E 60,992 (314) 109,557 (168) 9492–1,264,561
95/786 H 17,981 (213) 31,562 (25) 18,052–55,180
J Ref 03 I 2248 (992) 7368 (40) 3214–16,890
Reporter gene assays
Preparation Ampoule code Geometric mean and GCV(%): 95% confidence limits
all laboratories, n = 3
00/554 A 62,086 (9) 50,128–76,897
00/572 B/J 40,989 (12) 31,087–54,044
Gxb02-901-535 C 7069 (56) 2327–21,479
00/576 D/G 31,824 (3) 29,673–34,131
00/574 E 88,723 (24) 52,008–151,356
95/786 H 28,794 (30) 15,027–55,174
J Ref 03 I 5264 (15) 3740–7408
*For AV assays: lab 15 only excluded (n = 13); lab 15 and A549 assays from labs 08 and 09 excluded (n = 11). y For AP assays: lab 15 only
excluded (n = 3).
a
Estimates that contributed excessively to variation were excluded in the calculation of geometric mean values.

coded duplicate samples. Nevertheless the overall
estimate (5532 IU; Table 3) for Gxb02-901-535 (C)
in terms of Gb23-902-531 ( F) was broadly in line
with its assigned potency of 6000 IU (Standardization
of interferons, 1987; Pestka and Meager, 1997).
For all of these comparisons, estimates by AP
assays and RG assays, while derived from far fewer
numbers of assays, were consistent with the estimates
derived from AV assays. For AP assays there was a
trend to higher estimates overall (Table 3). Notably,
comparisons of the IFN-h Ser17 mutein containing
preparations Gxb02-901-535 (C, 1st IS for IFN-h
Ser17 mutein) and 00/574 (E) with Gb23-902-531
( F, 2nd IS for IFN-h, fibroblast-derived) in AP assays
gave estimates with between-laboratory variation over
20-fold larger than the between-laboratory variation
for comparisons of coded duplicate samples.
3.4. Comparison of 00/572 (B & J) and 00/576 (D & G)
Estimates of the potency of 00/572 (B & J)
expressed as ampoules of 00/576 (D & G) having
 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15 9

Fig. 2. Variation of geometric mean potency estimates of the CIS 00/572 (B & J) expressed as the number of ampoules of 00/576 (D & G)
equivalent to one ampoule of 00/572 (B & J). Boxes for AV assays, AP assays and RG assays are as illustrated below x-axis. The numbers refer
to the laboratory code. The letters following the laboratory code in the AV assay boxes indicate the cell–virus system, e.g., AE, A549-EMCV.

equivalent activity to one ampoule of 00/572 are
shown as individual estimates in Fig. 2, and as labora-
tory mean estimates expressed as IU of the IS Gb23-
902-531 ( F, 2nd IS for IFN-h, fibroblast-derived) in
Table 3. Analysis of variance showed significant dif-
ferences among the estimates by the various AV
assays, and among assays by all methods. However,
the between-laboratory variability was not signifi-
cantly larger than that obtained for the coded dupli-
cates. For example, the ratio of largest to smallest
estimate for J relative to B was 1.68 (124/74) and
for G relative to D was 1.58 (123/78), but the ratio of
the largest to smallest estimate for 00/572 in terms of
00/576 was 1.25 (1.40/1.12). Thus, the between-
laboratory differences for this comparison are similar
to those resulting from the inherent variability of the
assay systems. It is nevertheless noted that four of the
five smallest estimates by AV assays were given by
the four assay systems using FL cells (Fig. 2). The
two largest estimates are given by the assay systems
using 2D9 and FS4 cells, but these cell lines were
each used in one laboratory only in this study. Esti-
mates by AP assays and RG assays were consistent
with estimates from AV assays (Fig. 2).
These data suggest that within the limits of varia-
bility of the assay systems in this study there is a
consistent relationship between these two prepara-
tions. However, it is possible that certain systems
may see a different relationship, which is not verifi-
able in this study due to the relatively small number of
assay systems used.
3.5. Comparison of 00/574 (E) with 1st IS for recom-
binant IFN-b Ser 17 mutein, Gxb02-901-535 (C)
Comparison by AV assay of the IFN-h Ser17
mutein 00/574 (E) with 1st IS for recombinant
IFN-h Ser 17 mutein, Gxb02-901-535 (C) showed
significant differences in estimates from the different
laboratories. Omitting estimates from one laboratory,
which contributed excessively to the between-
laboratory variability in several comparisons, the
between-laboratory variation for the comparison of
00/574 (E) with Gxb02-901-535 (C) was about 10-
fold larger than the between-laboratory variation for
comparisons of coded duplicate samples. Estimates
by either AP or RG assays for each of these com-
parisons showed significant differences among esti-
mates from the different laboratories, and in many
instances there was at least one estimate outside the
range of estimates obtained using AV assays (data
not shown).
10 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15

3.6. Comparison of the activity of the coded prepara- bility some 6-fold larger than that for the coded
tions 00/554 (A), 00/576 (D/G) 00/574 (E) and 95/786 duplicates (see under Section 3.3). As described
(H) with that of the candidate international standard above in Section 3.4, estimates by antiviral assays
00/572, CHO cell-derived IFN-b, coded B/J for 00/576 (D/G) in terms of 00/572 (B/J) showed
between-laboratory variability which was not larger
The CIS 00/572 (B/J, CHO cell-derived IFN-h) than that given by the coded duplicate preparations,
was used as calibrant, with an assumed potency of whereas estimates for 00/576 (D/G) in terms of Gb23-
40,000 IU per ampoule, for the estimation of relative 902-531 ( F, 2nd IS for IFN-h, fibroblast-derived)
potencies of preparations 00/554 (A), 00/576 (D/G), have between-laboratory variability some 5-fold lar-
00/574 (E) and 95/786 (H). Laboratory geometric ger than that for the coded duplicates (Table 3).
mean estimates for these calibrations are shown in Estimates from AV assays for 00/574 (E) in terms
Table 4. of 00/572 (B/J) can be subdivided into three broad
Estimates for 00/554 (A) in terms of 00/572 were groups of significantly different estimates: labora-
similar in the different AV assays. When estimates that tories 08-A549 cells, 04 and 14-WISH cells, and 01-
contributed excessively to the between-laboratory 2D9 cells which gave large estimates; laboratories 10-
variability are excluded, the between-laboratory varia- FS4 cells, 05 and 07-FL cells, and 09 and 16-A549
bility for the remaining estimates was similar to that cells which gave a middle range of estimates; and
for the coded duplicates, and its ratio to the pooled laboratories 06 and 07-FL cells, 11-A549 cells and
within-laboratory variability was only marginally sig- 09-WISH cells which gave small estimates. For 00/
nificant ( p ~ 0.06). In contrast, estimates for 00/554 574 (E) in terms of 00/572 (B/J), the ranges of
(A) in terms of Gb23-902-531 ( F, 2nd IS for IFN-h, estimates in the three groups were 2.5–3.0, 1.2–1.4
fibroblast-derived) showed between-laboratory varia- and 0.85–1.0. These groups do not clearly reflect cell
or virus. In comparison, estimates from AV assays for
00/574 (E) in terms of Gb23-902-531 ( F, 2nd IS for
Table 4
IFN-h, fibroblast-derived) are slightly more variable
Laboratory geometric mean potencya of coded ampouled prepara-
tions expressed as IU per ampoule of the CIS 00/572 (B & J), between laboratories than estimates in terms of 00/572
assuming the CIS to have potency of 40,000 IU per ampoule (B/J; GCV 63% compared to GCV of 53%; Tables 3
Preparation Ampoule Geometric 95% confidence and 4). These estimates do not clearly separate into
code mean GCV(%) limits significantly different groups, although estimates from
(excluding laboratories 14 and 04-WISH cells are significantly
outlier,a n = 13) larger than other estimates.
Antiviral assays Estimates for 95/786 (H) in terms of 00/572 (B/J)
00/554 A 54,523 (19) 44,061–60,590 differed significantly between laboratories (Table 4),
00/576 D/G 33,021 (6) 31,968–34,109
00/574 E 58,262 (53) 45,075–75,306
with estimates from assay systems using WISH cells
95/786 H 30,568 (40) 24,953–37,447 being consistently small, and with between-labora-
tory variability some 15-fold larger than that for
Antiproliferative assays coded duplicate samples. Estimates for 95/786 (H)
00/554 A 58,131 (16) 40,573–83,290 in terms of Gb23-902-531 ( F, 2nd IS for IFN-h,
00/576 D/G 32,870 (13) 24,275–44,509
00/574 E 95,329 (143) 10,552–861,223
fibroblast-derived) have between-laboratory variabil-
95/786 H 27,463 (13) 20,291–37,169 ity some 8-fold larger than that for coded duplicates,
with estimates from assays using FL cells tending to
Reporter gene assays be large.
00/554 A 60,588 (15) 43,222–84,931 Estimates for 00/574 (E) in terms of 00/572 (B/J)
00/576 D/G 31,056 (9) 25,238–38,214
00/574 E 86,582 (34) 41,845–179,146
are described above. As noted, estimates of the
95/786 H 28,099 (18) 18,520–42,635 potency of IFN-h Ser17 mutein preparations in
a
Estimates from the AV and AP assays of lab 15 contributed
terms CHO cell- or fibroblast-derived IFN-h are not
excessively to variation and thus were excluded in the calculation of consistent between laboratories and assay systems.
geometric mean values. The between-laboratory variability for estimates of
 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15
00/574 (E) in terms of 00/572 (B/J) is some 20-fold
larger, and that for estimates in terms of Gb23-902-
531 ( F, 2nd IS for IFN-h, fibroblast-derived) is some
25-fold larger, than that for the coded duplicates.
Estimates by the AP assays for these comparisons
were generally similar between laboratories in this
study. Estimates by the RG assays in laboratory 09
were smaller than estimates by the other two labora-
tories using reporter gene assays. However, estimates
by both methods are within the range of estimates
obtained using AV assays (Table 4).
3.7. Thermal stability
Ampoules of CIS 00/554, 00/572, 00/574, and 00/
576 which had been stored for between 24 and 33
months at elevated temperatures, +4, + 20, +37, + 45,
and + 56 8C, were compared with ampoules of the
respective preparations stored continuously at
150
8C using AV (2D9-EMCV) and RG (293- ISRE/
SEAP) assays. In all four preparations the activity of
the material in ampoules stored at + 4, +20, and + 37
8C relative to that stored at
150 8C, or to that of the
liquid formulation stored at
150 8C, did not differ
significantly from 100%. Activity loss at + 45 8C
varied from insignificant for 00/574 to approximately
10% for 00/554, 00/572 and 00/576. At + 56 8C,
activity losses were 24%, 22%, 8% and 44% for 00/
554, 00/572, 00/574 and 00/576, respectively. There
are as yet too few data points to predict accurately
rates of degradation at temperatures below 0 8C. Data
from thermal degradation studies of similar cytokine
materials together with these early data suggest that
yearly annual activity losses at for example
20 and

70 8C will be much less than 0.01%. An approx-
imate calculation from Arrhenius plot analysis of
assay data applying to 00/572 suggests loss of 0.1%
total activity at
20 8C would require 300 years. All
of the four preparations tested for thermal stability on
long-term storage thus are predicted to be adequately
stable as reference standards. No evidence was found
that the recommended aliquoting and storage of
reconstituted preparations in small volumes at
70
or
20 8C (see Section 2.2) resulted in significant
loss of activity following thawing of aliquots. Thermal
stability testing of IFN-h preparations formulated with
HSA and bovine casein has shown them to be very
stable (A. Meager, unpublished results).
11
4. Discussion
Progress in the development and manufacture of
clinical grade recombinant IFN-h products has
prompted a re-evaluation of biological standards for
IFN-h, particularly since the 2nd IS of IFN-h (Gb23-
902-531) containing low purity, fibroblast-derived
IFN-h, was prepared prior to the currently available
highly-purified products. The present international
collaborative study, which involved parallel titration
of IS and new CIS in chiefly AV assays, but also in AP
and RG assays, has provided high quality data. These
data have permitted comparative analysis of the cri-
teria for biological similarity of the IS and CIS
included in the study. These criteria include paralle-
lism of the dose–response lines, relative magnitudes
of between- and within-laboratory variability and
comparison of estimates of relative potency in differ-
ent systems.
Similarity of the biological activity of a reference
preparation to the material under test is one of the
most fundamental scientific considerations that need
to be taken into account when deciding on the appro-
priateness of any material to serve as a standard.
Similarity of the bioactivity of any two substances
cannot be judged by any other method except by
comparison in biological systems. Even identical
amino acid sequences are not a guarantee that two
preparations will have the same biological activity. It
is also possible that a material is contaminated with
other biologically active substances and that these
will have a significant variable impact upon the
potency of the material according to the assay sys-
tems used.
Parallelism: The available data indicated that some
assay systems may distinguish some preparations. In
particular, reporter gene assays appeared to distin-
guish between preparations of IFN-h Ser17 mutein
(C & E) and glycosylated IFN-h (A, B/J, D/G).
Duplicates: If two preparations are identical, they
will have the same dose–response curve except for
variation introduced by the assay procedure. Simi-
larly, the relative potency of one of the duplicates in
terms of the other has an expected value of 1.0.
Deviations from this value are the result of dassay
variabilityT, where this term now includes all sources
of assay variability, and not just variability of replicate
responses. Estimates for the coded duplicate prepara-
12 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15
tions thus provide an independent measure of both the
accuracy and precision of the assay system.
The two pairs of coded duplicates included in this
study provided evidence of overall accuracy of esti-
mates. The overall mean for the duplicate relative to
the material identical except for code was in good
agreement with the expected value of 1 for each of the
two duplicated samples. Although there was evidence
of some degree of bias in individual assay systems,
this is not unexpected for assays of this type (Gaines
Das and Meager, 1995).
4.1. Comparisons by antiviral assays of the relative
activity of the coded preparations
The basis for assessment of the between-laboratory
variability of all comparisons is the between-labora-
tory variability of the coded duplicate samples. These
samples are known to be identical, and their between-
laboratory variability is taken to reflect the variability
inherent in these assay systems. The pooled within-
laboratory variability for all comparisons was broadly
similar (less than 2-fold difference for the majority of
pairs of possible comparisons).
The between-laboratory variability for comparisons
of 00/572 (B/J) with 00/576 (D/G) (for the combined
estimates of the duplicates or for each of the four pos-
sible comparisons of individual duplicates) was similar
to the between-laboratory variability of the identical
samples (differences of less than 1.5 fold). The data did
not indicate any consistent significant difference in the
dose–response relations for these CIS. These data sug-
gested that these two preparations have similar ratios of
activity in each of the assay systems in this study and
that 00/572, with an assigned potency of 40,000 IU
relative to the current WHO 2nd IS of IFN-h, human
fibroblast-derived (Gb23-902-531, F), would be suita-
ble as a primary standard to which 00/576 (D/G) could
be accurately calibrated (Table 1). Thus, a primary
standard containing highly purified CHO cell-derived
IFN-h was found appropriate for accurate assessment
of potency of test preparations containing highly pur-
ified, human fibroblast-derived, IFN-h. Since the uni-
tage of activity was highly consistent for preparations
of both highly purified, CHO cell- and human fibro-
blast-derived IFN-h, assignments of 33,000, 55,000
and 30,000 IU can be made to 00/576 (D/G), 00/554
(A), and 95/786 (H), respectively (Table 1).
It is of interest that the CIS 00/554, 00/572, 00/574
and 00/576 were all formulated with bovine casein
and human serum albumin as excipients. The addition
of bovine casein, which is unique to these prepara-
tions, has appeared therefore to confirm a direct rela-
tionship among their activities (except for 00/574),
which was not so apparent in any previous studies
in which similar preparations formulated without
casein were compared. Potency estimates for 95/786
(H), which contained CHO cell-derived IFN-h in a
formulation containing human serum albumin but no
bovine casein, in relation to 00/572 were found to be
more variable suggesting the casein to act beneficially
regarding stability and recovery of IFN-h activity
from ampoules. Comparison of J Ref 03 (I)—which
also contains no bovine casein—with 00/572 or the
2nd IS for IFN-h, Gb23-902-531 ( F), was also more
variable than 00/572 or 00/576. Thus, it should be
concluded that addition of bovine casein to the for-
mulations used for preparing lyophilised standards of
IFN-h significantly improves their stability/recovery
of activity and therefore their suitability to serve as IS.
In contrast to the candidate IS of glycosylated
CHO cell- and human fibroblast-derived IFN-h, all
other comparisons showed significantly larger
between-laboratory variability than that obtained for
the duplicate samples.
The largest between-laboratory variability was
obtained for the comparisons of the IFN-h Ser17
mutein preparations 00/574 (E) and Gxb02-901-535
(C) with the CHO-cell or fibroblast-derived IFN-h
preparations (25-fold or greater between-laboratory
variability compared to between-laboratory variability
for the coded duplicates). This finding indicates that
the assays used in the study appear to distinguish the
activity of IFN-h Ser 17 mutein from that of both
CHO-cell and fibroblast-derived IFN-h. The reason(s)
for this difference in activity however remains unclear.
Possibly, interaction of IFN-h Ser 17 mutein with type
I IFN receptors is altered, but studies to investigate
such a possibility are as yet not reported.
The implication of the activity difference between
IFN-h Ser 17 mutein and glycosylated, CHO cell- or
fibroblast-derived, IFN-h for the suitability of biolo-
gical standards is clear. Standards containing IFN-h
Ser 17 mutein would not be suitable for the calibration
of bioassays of glycosylated, CHO cell- or fibroblast-
derived IFN-h, and vice versa. A similar conclusion
 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15
was reached in a previous international collaborative
study when it was recommended that a homologous
IS separate from the IS for human fibroblast-derived
IFN-h was needed for recombinant E. coli-derived
IFN-h ser17 (Standardization of interferons, 1987).
This recommendation led to the establishment of
Gxb02-901-535 as the 1st IS for recombinant E.
coli-derived IFN-h ser17. This distinction and con-
tinued need for separate standards for preparations of
glycosylated CHO cell- or fibroblast-derived IFN-h
and E. coli-derived IFN-h ser17 mutein is supported
by the findings of this study.
Although Gxb02-901-535 (C) and 00/574 (E) both
contained E. coli-derived IFN-h Ser17 mutein,
between-laboratory variability of estimates in this
study of E in terms of C was larger than that found
for the coded duplicates (B/J, D/G), but significantly
less than that of estimates of E in terms of Gb23-902-
531 ( F) or 00/572 (B/J), which contain glycosylated
fibroblast-derived and CHO cell-derived IFN-h,
respectively. A preliminary international collaborative
study also showed that Gxb02-901-535 would be
more suitable as primary IS for E. coli-derived IFN-
h Ser17 mutein and other E. coli-derived, non-glyco-
sylated IFN-h preparations than Gb23-902-531 (A.
Meager and R. Gaines Das, unpublished results).
However, it is unclear why the activities of Gxb02-
901-535 (C) and 00/574 (E) are not so similar as for
blike to likeQ comparisons among glycosylated IFN-h
preparations in this study. Although the between-
laboratory variability is especially marked for AP
and RG assays, for AV assays alone it is similar to
that for comparisons of 00/576 or 00/572 with Gb23-
902-531 if a single outlying estimate is omitted. Rela-
tive to Gxb02-901-535, 00/574 is therefore assigned a
potency of 64,000 IU (Table 1).
5. Conclusions
Comparison of all preparations with the CIS 00/
572 gave smaller between-laboratory variability than
comparisons of the same samples with Gb23-902-531
with only one exception; that for comparison with 95/
786. The between-laboratory variability of potency
estimates of 00/572, 00/576 and 00/554 was shown
to be of the same order as that for the coded dupli-
cates. Therefore, 00/572, which has high stability, is
13
suitable as a primary standard for purified CHO-cell
and fibroblast derived IFN-h, and an assignment of
40,000 IU per ampoule, relative to the current WHO
2nd IS of IFN-h, Gb23-902-531, is proposed. With
the agreement of participants and following discus-
sion of the study data at the 7th WHO Informal
Consultation on Standards for Cytokines, Growth
Factors and Endocrinological Substances (NIBSC,
October 2003), it was recommended that CIS 00/572
should be proposed to WHO ECBS as an immediate
replacement of the 2nd IS of IFN-h, Gb23-902-531,
which contains approximately 1% fibroblast-derived
IFN-h and other cytokines, most notably IL-6. Some
investigators have reported that inconsistent potency
estimates of test IFN-h preparations may result when
this IS is used to calibrate bioassays and the relative
potencies of in-house, dworkingT standards, particu-
larly when different bioassays are compared. Use of
00/572, which contains highly purified CHO cell-
derived IFN-h, is expected to overcome such incon-
sistencies. Following consideration of the report and
recommendations, the WHO ECBS in November 2003
at its 53rd Annual Meeting established 00/572 as the
3rd IS of IFN-h, with an assigned potency of 40,000
IU, specifically for the calibration of bioassays to
determine the potency of preparations of glycosylated
human IFN-h.
In contrast, 00/572 is shown not suitable as stan-
dard for IFN-h Ser17 mutein preparations. The activ-
ities of IFN-h Ser17 mutein and glycosylated, CHO
cell- or human fibroblast-derived IFN-h are distinct
on the basis of the results in this and previous studies.
Currently therefore only the WHO 1st IS of IFN-h
Ser17 mutein, Gxb02-901-535 is suitable for the cali-
bration of bioassays of test preparations of IFN-h
Ser17 mutein. Once stocks of this IS become
exhausted, CIS 00/574 containing IFN-h Ser17
mutein, with an assigned potency of 64,000 IU rela-
tive to Gxb02-901-535 on the basis of antiviral assays,
should be considered as a suitable replacement.
Acknowledgements
We are indebted to all the participants for their time
and motivation in completing their assays and con-
tributing their raw data, without which this study and
its successful outcome would not have been possible.
14 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15
We gratefully acknowledge the valuable contributions
of members of the WHO Consultative Group on
Cytokine Standardisation. We sincerely thank all of
the pharmaceutical manufacturers who generously
donated materials for the preparation of lyophilised
CIS. We also thank the staff of the NIBSC’s Standards
Division for processing materials and distributing
ampoules to participants.
Appendix A. Participants in the collaborative
study
Dr. V. Jethwa (Biogen Inc., 14 Cambridge Center,
Cambridge, MA 02142, USA).
Dr. J. Weaver (Chiron Corpn., 4560 Horton St,
Emeryville, CA 94608, USA).
Dr. J. Files (Berlex Biosciences, 15049 San Pablo
Ave, Richmond, CA 94804-0099, USA).
Dr. M. Kohase (National Institute of Health, 4-7-1
Gakuen, Musashimurayama-shi, Tokyo 208,
Japan).
Dr. A. Walser (dr Rentschler Biotechnologie
GmbH, Erwin-Rentschler-Strasse 21, D-88471
Laupheim, Germany).
Dr. A. E. Sterin Prync (BioSidus, Constitucion
4234, 1254 Buenos Aires, Argentina).
Dr. A. Meager (Division of Immunology and
Endocrinology, NIBSC, South Mimms, Herts.,
EN6 3QG, UK).
Dr. L. W. Whitehouse (Drug Research, Life
Sciences Division, Rm1 West, Sir F G Banting
Research Center, Tunney’s Pasture, Ottawa,
Ontario, K1A 0L2, Canada).
Dr. M. Wadhwa/Mrs. P. Dilger (Division of Immu-
nology and Endocrinology, NIBSC, South Mimms,
Herts., EN6 3QG, UK).
Dr. F. Antonetti (Istituto di Ricerca Cesare Serono
S.P.A., Via Valle Cais 22, 00040 Ardea, Rome,
Italy).
Dr. J. Bekisz (Center for Biologics Evaluation and
Research, FDA, 1401 Rockville Pike, Rockville,
MD 20852, USA).
Dr. S. Sakurai (Toray Industries Inc., Mishima
Plant, 4845 Mishima Shozuoka, 411 Japan).
Dr. K. Sakai (Mochida Pharmaceutical Co Ltd.,
Shizuoka Factory, 342 Gensuke Fujieda, Shizuoka
426, Japan).
Dr. S. E. Grossberg (Dept of Microbiology, Med-
ical College of Wisconsin, 8701 Watertown Plank
Road, Milwaukee, WI 53226, USA).
Dr. M. Poulis (Molecular Biology Section, TGAL,
PO Box 100, Wooden, Act 2606, Australia).
Dr. L. Browning (GeneTrol Biotherapeutics, 2220
Livingston Street, Suite 201, Oakland, CA 94606,
USA).
References
Content, J., De Wit, L., Pierard, D., Derynck, R., DeClercq, E.,
Fiers, W., 1982. Secretory proteins induced in human fibroblasts
under conditions used for the production of interferon beta.
Proc. Natl. Acad. Sci. U. S. A. 79, 2768.
De Ley, M., Van Damme, J., Claeys, H., Weaning, H., Heine, J.W.,
Billiau, A., Vemylen, C., De Somer, P., 1980. Interferon induced
in human leukocytes by mitogens: production, partial purifica-
tion and characterization. Eur. J. Immunol. 10, 877.
Edy, V.G., Desmyter, J., Billiau, A., De Somer, P., 1977. Stable and
unstable forms of human fibroblast interferon. Infect. Immun.
16, 445.
European Pharmacopoeia, 2nd edn., 1992. Determination of Water.
Fasicule 16, V. 3.5.6.
Gaines Das, R.E., Meager, A., 1995. Evaluation of assay designs for
assays using microtitre plates: results of a study of in vitro
bioassays and immunoassays for tumour necrosis factor
(TNF). Biologicals 23, 285.
Gaines Das, R.E., Tydeman, M.S., 1982. Iterative weighted regres-
sion analysis of logit responses: a computer program for analy-
sis of bioassays and immunoassays. Comput. Programs Biomed.
15, 13.
Kagawa, Y., Takasaki, S., Utsumi, J., Hosoi, K., Shimizu, H.,
Kochibe, N., Kobata, A., 1988. Comparative study of the aspar-
agine-linked sugar chains of natural human interferon-h1 and
recombinant human interferon-h1 produced by three different
mammalian cells. J. Biol. Chem. 263, 17508.
Karpusas, M., Nolte, M., Benton, C.B., Meier, W., Lipscomb, W.N.,
Goelz, S., 1997. The crystal structure of human interferon h at
2.2—a resolution. Proc. Natl. Acad. Sci. U. S. A. 94, 11,813.
Knight Jr., E., 1981. Purification of human fibroblast interferon
prepared in the absence of serum. In: Pestka, S. (Ed.), The
Interferons, Part A, Methods in Enzymology, vol. 78. Academic
Press, New York. 417 pp.
Knight Jr., E., Fahey, D., 1982. Human interferon-beta: effects of
deglycosylation. J. Interferon Res. 2, 421.
Le, J., Vilcek, J., 1989. Interleukin-6: a multifunctional cytokine
regulating immune reactions and the acute phase protein
response. Lab. Invest. 61, 588.
Mark, D.F., Lu, S.D., Creasey, A.A., Yamamoto, R., Lin, S., 1984.
Site-specific mutagenesis of the human fibroblast interferon
gene. Proc. Natl. Acad. Sci. U. S. A. 81, 5662.
McCullagh, K.G., 1983. Biological properties of human inter-
feron beta 1 synthesised in recombinant bacteria. J. Interferon
Res. 3, 97.
 A. Meager, R.G. Das / Journal of Immunological Methods 306 (2005) 1–15
Meager, A., 1998. Interferons alpha, beta, and omega. In: Mire-
Sluis, A., Thorpe, R. (Eds.), Cytokines. Academic Press, San
Diego, p. 361.
Pestka, S., Meager, A., 1997. Interferon standardisation and desig-
nations. J. Interferon Cytokine Res. 17 (Suppl. 1), S9 – S14.
Runkel, L., Meier, M., Pepinsky, R.B., Karpusas, M., Whitty, A.,
Kimball, K., Brickelmaier, M., Muldowney, C., Jones, W.,
Goelz, S., 1998. Structural and functional differences between
glycosylated and non-glycosylated forms of human interferon-
beta (IFN-beta). Pharm. Res. 15, 641.
Satoh, Y.I., Kasama, K., Kajita, A., Shimizu, H., Ida, N., 1984.
Different pharmacokinetics between natural and recombinant
human interferon beta in rabbits. J. Interferon Res. 4, 411.
Standardization of interferons, 1987. Report of a WHO Informal
Consultation. World Health Organization Technical Report Ser-
ies No. 771.
Tan, Y.H., 1981. Purification of human fibroblast interferon pre-
pared from modified fibroblasts. In: Pestka, S. (Ed.), The Inter-
15
ferons, Part A, Methods in Enzymology, vol. 78. Academic
Press, New York, p. 422.
Taniguchi, T., Mantei, N., Schwarzstein, M., Nagata, S., Mura-
matsu, M., Weissmann, C., 1980. Human leukocyte and fibro-
blast interferons are structurally related. Nature 285, 547.
Utsumi, J., Mizuno, Y., Hosoi, K., Okano, K., Sawada, R., Kajitani,
M., Sakai, I., Naruto, M., Shimizu, H., 1989. Characterisation of
four different mammalian-cell-derived recombinant human
interferon-h1s. Eur. J. Biochem. 181, 545.
Van Damme, J., Billiau, A., 1982. Large scale production of human
fibroblast interferon. In: Pestka, S. (Ed.), The Interferons, Part
A, Methods in Enzymology, vol. 78. Academic Press, New
York, p. 101.
Watanabe, Y., Kawade, Y., 1983. Properties of non-glycosylated
human interferon-h from MG63 cells. J. Gen. Virol. 64, 1391.