Note: Descriptions are shown in the official language in which they were submitted.
WO92/02815 PCT/US~1/055~
2 ~ 6 ~3
QUANTITATIVE ANALYSIS AND MONITORING
OF PROTEIN STRUCTURE BY
SUBTRACTIVE CHROMATOGRAPHY
Backqround o~ the Invention
The present invention qenerally relates to a method
for performing quantitative and structural analyses on
solutions containing multiple solutes. In particular,
the invention relates to methods for determining the
presence and concentration and for characterizing the
structural profile of an analyte in a solution
utilizing subractive frontal breakthrough analysis.
It promotes efficiency in the production of
therapeutic substances to monitor each step of the
synthesis process to insure that ~uality and quantity
~0 of a desired intermediate or product is within
specification. In addition to there being.extensive
federal regulation designed to insure the integrity of
such processes, economics dictate that proper
safeguards be taken to maintain precision with regard
to the production of therapeutics. Indeed, in any
manufacturing stream which produces a drug, pesticide,
food additive, dye, or other high value substance in a
multistep process, monitoring product concentration
rapidly and accurately at various production or
purification stages can be a key to maintaining
e~ficient operation.
While analytical methods exist for precisely
quantitatively determining a particular solute such as
a protein or organic compound in a mixed solution, due
E3$TIT~T~
WOg2/0~815 PCT/US9~/05~
2~883~0 2-
to the time reqllired to carry out these analyses, and
to their cost and complexity, they are impractical for
routine use. It is not feasible to use these methods
in conjunction with continuous monitoring of a
manufacturing process because they do not offer "real
time" information. ~y the time data from a particular
sample can be processed, the state of the system may
have changed dramatically.
Further, qualitative or structural analysis of
structural variants of proteins, e.g., recombinantly
produced therapeutic proteins, is not presently
possible except in those instances where the variation
in the amino acid sequence or tertiary structure
results in an altered activity profile or the separate
species can be resolved on a gel. Determining the
presence of structural variants is particularly
important in the field of biosynthetic protein
production, e.g., in recombinant DNA technology as
applied to the manufacture of therapeutic proteins.
Recombinantly produced proteins have a higher rate of
expression error, and can include a number of "minor"
structural variants having, for example, diminished
activity, harmful side effects, or undesirable
antigenicity. Structural variants also may be produced
during post translational protein modification by, for
example, variation in glycosylation pattern, disulfide
bonding, or protein folding.
In the biotechnology industry, it is o~ten
necessary to monitor the concentration of a product in
a mixture which may contain hundreds of contaminating
species including cell debris, various solutes,
nutrient components, DNA, lipids, polysaccharides,
protein species having similar physiochemical
~3UIE3ST~T~JT1~
. , , : ~.
..
. . .
W~92/0281~ PCTIU~91/055~
208~3~
properties, and structural variants of a single
recombinantly produced protein. While the
concentration of the target product in the harvest
liquor is usually ol the order of 100 mg/l, it is
sometimes as low as 1 mg/l. To complicate
identificatio~ further, due to the fragility of many
target solutes, they must be treated with relatively
low fluid shear, and preferably with only minimal and
short duration contact with potentially denaturing
surfaces.
One known method of identifying solutes in a
solution is affinity chromatography, which involves
passing a feed mixture over a matrix such as a packed
bed of selectively sorptive particles to bind one or a
subset of the solu~es in the mix. Subsequent passage
of solutions that modify the chemical environment at
the sorbent surface results in elut.ion of sorbed
species. Solutes flow throuyh these systems
convectively in the interstitial space among the
particles and diffusively within the particles. The
media used for liquid chromatography typically
comprises soft, highly porous particles having a high
surface area to volume ratio. As a result, a liquid
chromatography process cannot be run at pressures
exceeding about 50 psi, and attempts to increase the
fluid velocity are counterproductive to separation.
High Performance Liquid Chromatography (HPLC) employs
as a matrix rigid porous beads made typically of an
3Q inorganic material such as silica or a rigid polymer
such as styrene divinylbenzene. HPLC allows somewhat
faster and higher resolution separations.
Typically, chromatographic procedures employed for
purification involve four steps: loading; washing;
eluting, and re-equilibrating. The rate limiting step
~;VB8T9TaJTE SHEET
W~92/02~l5 PCT/US91/05~
20883~0
in each stage is the ~ransport of molecules between the
mobile fluid and the static matrix surface. Optimum
ef~iciency is promoted by rapid, preferably
instantaneous mass transfer and high fluid turnover.
During sorbent loading, fewer moleculares are sorbed as
the velocity of the mobile phase in the bed increases.
At some mobile phase velocity, some target solute is
lost in the effluent as "breakthrough", i.e., passes
through the matrix without binding and appears in the
effluent. If the breakthrough concentration is limited
to, for example, 5% of the inlet concentration, that
limit sets the maximum bed velocity which the bed will
tolerate. Further increases in bed velocity thereafter
are wasteful and can only serve to decrease loading per
unit surface area.
Recently the art has developed novel matrix designs
for increasing the speed of separation of liquid
chromatography. For example, by using as a matrix a
non-porous microparticulate material, one can avoid the
rate limiting diffusion step and greatly increase the
speed of separation. That is, solution will flow
! convectively through the interstitial spaces between
particles forming the matrix, and solutes will interact
at the particle surface exposed at the walls of the
interstices. In this manner, a separation can be
carried out much more quickly than in "diffusion bound"
chromatography systems, but at the cost of greatly
diminished capacity.
One can increase chromatographic throughput by
using a matrix comprising small porous part ~les haviny
a relatively large pore diameter, so that convective
flow can be induced through, as well IS around, the
particles. This type of chromatography is referred to
SVB~STITUTE SHEE~T
.; ~ ; . , . -:
,~
.. . .
.
WO9~/02815 PCTtVS91/05~
5. 2~8~3~
as Perfusive Chromatography and is described in
copending application serial number 376,885, filed
July 6, 1989, the disclosure of which is incorporated
herein by reference. Perfusive chromatographic
techniques permit high speed, high capacit:y, high
resolution separation~ Perfusive matrices may be
purchased from PerSeptive Biosystems, Inc. of
Cambridge, Mass.
Even the fastest of the above discussed known
chromatographic techni~ues are too slow to afford a
method for real-time analysis of solute in a product -
stream. Due to the complexity of the solutions being
monitored, and the multiple steps required, these known
analytical techniques cannot provide data quickly
enough to allow meaningful adjustments to be made to a
production process, nor are they sensitive enough to
detect structural variants of a proteinaceous solute.
It is an object of the invention to provide methods
which exploit the benefits of high speed
chromatographic techniques to allow real time
monitoring of solute concentration in a process liquor.
Another object is to provide such methods that can
rapidly, accurately, and precisely monitor the
concentration of a therapeutic substance or other
solute at any stage of a production or purification
process. A further object is to provide methods for
producing a profi}e of the mixture representative of
the nature and relative concentration of structured
variants. Another object is to provide such methods
that have a self-checking capability.
SlJlE3~3TIT~TI~ EI~T
WO92/02~15 PCTtUS91/OS~
._..
2~8836~ ~-
Summary of the Invention
The present invention provides a method and
apparatus for rapid assay and characterization of
therapeutic and other substances based on what is
described herein as subtractive chromatography. In
accordance with the invention~ a solution containing
multiple solutes is passed through a matrix ha~ing
binding sites specific for one or more target solutes.
As used herein, "target solute" is broadly defined and
encompasses any water soluble analyte but typically is
a protein such as a recombinantly produced protein. ~y
analyzing the effluent flowing ~rom the column, the
presence and concentration or the profile of the
structural variants of the target solute can be
determined.
In accordance with a first aspect of the invention,
a feed solution containing at least one target solute,
for example, a biologically active molecule such as a
polypeptide, protein, polysaccharide, or the like, in
admixture with other solutes, is passed through a
matrix comprising binding sites specific to the ~arget
solute. As the feed solution passes through the
matri.x, the target solute will adsorb at the binding
sites, thereby virtually eliminating any concentration
of the target solute in the effluent. During this
process, a limited amount of non-target solute may also
non-specifically adsorb to the matrix. The effluent is
monitored to determine its solute concentration.
While, in a preferred embodiment of the invention, this
will entail monitoring the ultra-violet absorption of
the effluent, which is proportional to concentration,
it should be understood that any number of alternative
TUT1~ 9~
,: '
WO92/02815 PCT/IJ~91/055~
7- 2~8~36~
methods can be used to the same effect. Any method
which produces data related to the solute concentration
of the effluent is suitable.
As the effluent begins to flow from the column, the
concentration f contaminating or "non-target"
solute~s) in the effluent will increase until the
concentration of non-target solute in the effluent
reaches an equilibrium level equal to the concentration
of non-target solute in the feed. When graphed as the
relationship between, for example, ultra-violet
absorption and time, this stage of the assay procedure
will result in an upturned slope or vertical line,
depending on the nature of the matrix, which develops
into a flat, horizontal line as solute concentration in
the effluent maximizes.
The equilibrium concentration of solutes in the
effluent will remain substantially constant as the feed
solution is passed through the matrix as long as
binding sites remain available. Eventually, however,
the binding sites of the matrix become saturated, and
the target solute will flow directly through the matrix
without net interaction. This is referred to as
breakthrough. Thus, the emergence of the target solute
from the matrix will result in a detected increase in
ultra-violet absorption of the effluent. Thus, when
solute concentration reaches a plateau indicating that
the feed is simply flowing through the column without
net solute interaction with the matrix surface, the
solute concentration in the effluent equals the
concentration in the feed.
When a non-diffusively bound chromatography matrix
is used, or when liquid flow rates are slow relative to
diffusion times, the above discussed phenomena result
~IB~;TITlJ~E ~;HE~
, :
,
WO92/02815 PCT/US9l/055~
~8836~
in a graph with two well defined steps. That is, when
an equilibrium concentration representative of the
concentration of non-target solutes in the effluent is
reached, a first well-defined plateau will result.
~his will be followed by a transition period indicated
by a vertical line, or a line with a slope approaching
the vertical, and a second plateau representative of
the concentration of the target and non~target solutes
together.
The difference between these equilibrium
concentrations may be used to calculate the
concentration of target solute in the sample as the
difference between equilibrium concentration is
directly proportional to the concentration of target
solute in the sample. Furthermore, since the second
plateau is indicative of the additive concentration of
all solutes in the feed, that value can be obtained by
monitoring the sample prior to the time it enters the
matrix. Thus, all information necessary to calculate
the target solute concentration is available as soon as
a plateau in the breakthrough of non-target or
contaminating solute is reached. The device is
calibrated by passing through the solute detector known
concentrations of pure target solute so that
concentration units can be correlated directly with,
e.g., absorbance units. The product of the difference
between the sensed plateaus and the correlation factor
equals the concentration of the target solute.
In an alternative embodiment, the method of the
invention is used for detecting difference~ in the
structural profile of a protein in separate samples.
I The phrase "structural profile", as used herein, refers
to the particular mix of molecular species in a protein
!3UB5TITUTIE Sl-IEET
... .
;, ,~,
,
, .. . .. .. . .
.
... . .
..
. . .
WO92/02815 PCT/~S91/05~
~. 2~8~360
solution which can vary from batch to batch or over
time due to expression errors, differences in DNA
sequence among the clones in a culture, truncation by
proteases, or differences in post translational
modification resulting in variations in conformation or
derivatization. In this embodiment, the method
comprises the steps of passing the samples through a
matrix comprising immobilized binding sites which vary
with respect to their binding properties to structural
variants in the sample. For example, polyclonal
antibodies may be used, cloned variants of which are
specific for a particular epitope on a particular
variant of the protein. Alternatively, a single type
of binding site may be used which varies in binding
affinity or specificity with variants of the protein to
be analyzed. This procedure can produce a breakthrough
function characteristic of the structural profile of
the protein in the sample as the concentration of
protein exiting the matrix is measured after at least
sGme of the binding sites have been saturated with the
protein. Comparing the characteristic functions of
different samples permits indirect comparison of their
structural makeup.
Since molecular subspecies in the protein mix have
separate and distinct structural features, each
subspecies has at least some unique epitopes. Each
fraction of the binding protein in the matrix therefore
will be capable of discriminating, (i.e., selectively
binding) particular molecular subspecies, or of binding
a molecular subspecies preferentially. Thus, as the
protein sample is passed through the matrix, various of
its subspecies reach equilibrium saturation, and
thereafter break through into the ef~luent. If the
protein concentration of effluent is monitored over
$VE38T3TI~T~
- .
, . ., .
: - . ` . :
~:
,
..
. . , ' ~ : '
WO~2/02815 PCT/US91/055~
2~3~
time, there is an interval over wllich protein
concentration in the effluent increases from a baseline
value, typically zero, to a value substantially
identical to protein concentration in the feed. During
the interval the protein concentration increases
progressively in a way that is indicative of the
particular structural profile of the protein sample.
When this function is compared for separate protein
samples, one can determine whether those samples have
uniform structure. This method can be used, for
example, to monitor a product stream periodically as a
means of assuring that the product remains within a
predetermined specification.
15 One may display the function of the sample as a
plot of a parameter indicative of concentration (e.g.,
U.V. absorbance) against a parameter indicative of
volume of sample exiting the matrix (e.g., time if flow
rate is uniform). One may also elute the protein from
the matrix, wash the matrix, and repeat and passing the
measuring steps. The method is highly effective when
the protein sample is an aqueous protein solution which
has been at least partially purified. The method may
involve passing the protein through the matrix by a
means of a pressure gradient or a charge gradient.
The matrix preferahly is a rigid, substantially
non-microporous, particulate material having a
hydrophilic surface, and preferably is a perfusive
chromatography matrix. The matrix also may be defined
by the interior surface of a capillary. Where the
matrix comprises surface regions comprising immobilized
protein A, protein G, and the binding protein is
immuno~lobulin, one can remove the binding sites from
the matrix after each run, and reload the matrix with
~;IIJE3 5T~T~JTE~
- ..
.- :' ~, . ': :
. ~': . , ' ' " ~ .
. .
WO92/02815 PCT/US91/~55~
1 1 ~08~36~
fresh binding sites. Immunoglobulin and other types of
protein binding sites also may be non-specifically
adsorbed on a hydrophobic polymer matrix surface and
removed with mixed organic/ionic stripping solutions.
It is nec_ sary to the proper exploitation of the
various embodiments of the present inventlon that a
chromatographic or electrophoretic technique be used
that results in a well-defined breakthrough. ~his can
be achieved readily using essentially any matrix
geometry provided the flow rate through the matrix is
slow. At slow flow rates, the time required for
solutes to diffuse into and out of the pores of the
conventional HPLC or other chromatography medium is
insufficient to destroy the development of a distinct
concentration plateau in the effluent. However, at
higher flow rates using conventional media, the
concentration plateaus in the effluent typically are
not discernible. This essentially means that, for
desired high speed operation, non-diffusion bound
chromatographic matrices should be used.
Also, the matrix should be as small as possible.
The volume of sample that can be present in the matrix,
coupled with the flow rate, dictate the time interval
between introduction of the sample and breakthrough.
Higher flow rates and small volume columns promote high
speed analysis. This approach can resolute in assays
being performed in periods of time substantially less
than one minute and easily less than 10 seconds. For
all practical purposes, these short time frames can be
considered ~real time" measurements.
The quantitative analysis technique is independent
of flow rate, and does not require the target solute
and the matrix to reach equilibrium. Thus, the sample
$1JBSTITUTE SHE~E~
.
... . -
.:
W~92/02815 PCT/US9~/055~
~ 2
` 2~883~
may be impelled through the matrix by any convenient
method The sample may be impelled through the matrix
manually, e.g., using a syringe, by an electrically
driven pump, or by a charge gradient as in
electrophoresis.
It is a particular advantage of the invention that
assays can be performed repeatedly without comprising
the accuracy of the process. While an unknown subset
of binding sites of the matrix may be degraded with
repeated sequences of binding, elution, and
reequilibration, the method of the invention generates
information based on concen~ration differences of the
target and non-target solutes. Thus, the availability
of fewer binding sites will translate to earlier target
solute breakthrough but will not give inaccurate
indications of concentration.
The invention also affords a self checking
capability. If detected concentration differs between
the feed and the final effluent plateau, the system may
be operating improperly. Self checking also can be
implemented by washing the matrix after the final
effluent plateau has been reached and then eluting the
target solute. Integration of the detected pulse in
the eluate will give an indication of the amount of
bound target solute, which should correlate with the
previous datum.
Another advantage of the invention is that it is
very flexible. Consider, for example, a situation in
which a sample having high concentration o' target
solute is passed through a matrix. This may result in
almost immediate saturation of the binding sites of the
matrix, and therefore, almost immediately breakthrough.
$1J 113$T~I~ITE ~;HIEE~
; .
,~ ' ' , ,
.
, ~ ' .
:. . .
WO92~2815 PCT/VS91/055~
li 3 20~$36~
On a graph like that discussed above, the output will
appear as a single vertical line followed by a
horizontal plateau, giving no information about the
concentration of target or non-target solute. To
remedy this situation, the sample need only be diluted
with buffer solution or the like. By diluting the
sample breakthrough is delayed, thereby affording a
clear distinction between the equilibrium concentration
of the non-target solute in the effluent and the
equilibrium concentration of the target arld non-target
solute together. If the amount of diluent is known,
dilution does not adversely affect the precision or
accuracy of the results. Assay of very dilute samples
can also be conducted routinely. The only potentially
negative effect on the system is that the time reguired
to saturate the binding sites increases. This, of
course, is a liability only for the self-checking
aspect of the process, as the plateau reached after
breakthrough of the target solute can be determined
directly from the sample.
Another feature of the invention is that the
binding sites on the matrix, e.g., monoclonal or
polyclonal antibodies or other binding proteins, can be
interchanged readily depending upon the identity of the
target solute, using known techniques. This feature
permits construction of a single matrix and assay
device which can be customized for any target solute.
A further advantage of the inven~ion is that it can
be utilized on an extremely small scale. Even
microliter sized samples can be analyzed. Moreover,
rater than filling a traditional chromatography column
with high surface area particles to serve as a matrix,
on~ can coat binding protein on the inner surface of a
3 3UB~;T~U~ lEE~T
.; , , : .
.
WO92/02815 PCT/~IS91/OSS~
~ 4
2088360
capillary tube. Passing a solution through the
capillary tube can achieve the same results as those
discussed above. It is contemplated that an assay
device embodying the invention, including sample,
eluant, and buffer ports, matrix channel ready to be
activated with bindiny protein, detector for solute
concentration in the effluent, and circuitry to convert
the output of the detector to a meaningful form, all
could be housed in a single module. Alternatively, an
inexpensive matrix module can be produced for placement
in a device comprising the other necessary components.
The module can be discarded after a short useful life
and then replaced. Alternatively, sets of modules each
of which bind a different target solute will permit
rapid adaptation of an assay apparatus for given target
solute.
These and other advantages, objects, and features
of the present invention will be more fully understood
with reference to the following detailed description in
conjunction with the attached drawing in which like
reference characters indicate corresponding parts and
in which:
Brief Description of the Drawinq
FIG. 1 is a rPpresentative chromatogram generated
in conjunction with diffusion bound chromatography at
high throughput;
FIGS. 2 through 4 are various chromatograms
illustrating the principles of the present invention;
FIG. 5 and 6 are schematic representations of two
embodiments of apparatus embodying the invention, in
8l.JlE38TlTUTE SHEE~T
~'' ' ' ' . . ' '- .
.
W092/028~5 PCT/US91/055~
1 5 2~83~'~
which like reference characters indicate corresponding
parts;
FIGS. 7 and 8 are chromatograms generated using the
analysis technique of the present invention for
measuring the concentration of Immunoglobulin in
solution using a Protein A column where BSA is a
contaminant;
FIG. 9 is a chromatogram showing the results of an
experiment involving the tertiary structural profile of
mouse gamma globulin and demonstrating the feasibility
of an embodiment of the invention; and
FIG. lO is a representative calibration curve of
milli absorbance units (mAu) versus protein (IgG)
concentration in mg/ml.
Detailed Description of Preferred Embodiments
In its broadest aspects, the invention provides a
method for monitoring the production of a solute based
on subtractive frontal breakthrough analysis.
' '
The concept is to exploit an affinity
chromatography matrix to remove selectively at least
one solute of interest from a solution, and to measure
the equilibrium concentration of contaminating solutes
in the effluent exiting the matrix. In a first
embodiment, the next step involves determining the
concentration of the target solute in the analyte
sample from the difference between the sensed
- concentration of all solutes in the sample and the
sensed concentration of the contaminants. In a second
embodiment, the solute of interest is a protein,
.
:'
SU13!STIT~TE ~HEET
.. . .
. .
'
, .
W092tO2815 PCT/US91/0~5~
2~8836~
particularly a purified recombinantly produced protein,
comprising an unknown number of structural variants
each of which vary at least subtly in their affinity
for a particular binding protein, and have at least
some unique epitopes. The protein sample, preferably
substantially free of contaminants, is passed through a
matrix comprising a single binding protein or
immobili7ed polyclonal antibodies to the protein of
interest. Variant protein molecules in the sample
saturate the various clonal species of the polyclonal
antibodies or compete for sites of attachment to a
single type of bindinq site and then break through.
Output monitoring produces a step-like plot of the
breakthrough fronts characteristic of that particular
sample.
Apparatus
Figures S and 6 schematically illustrate apparatus
designed for implementing ~he process of the invention.
Referring to Figure 5, a valve 10 directs through its
output 12 either a sample from sample input 14, a
buffer solution from reservoir 16 for washing and
reequilibrating a chromatography matrix, or an eluent
from reservoir 18 capable of inducing release of sorbed
species from binding sites in a chromatography matrix.
The output of valve 10 ultimately directs a selected
solution through a chromatography matrix 20 of a nature
hereinafter described in more detail, which comprises
binding sites disposed about a surface and capable of
selectively adsorbing an analyte or target solute
sought to be determined. Optionally, inte- osed
be~ween valve 10 and matrix 20, as indicated in phantom
at 22, is a solute concentration detector capable of
providing a signal through line 24 representative of
~!3UB~TlTl~T~ ~3HE~
WO92/02815 PC~ S91/055~
1 ~ 2~g3SO
the concentration of solutes in the sample.
Detector 22 may be a conventional device of the type
commonly used in chromatography equipment comprising,
for example, a U.V. light source which provides a beam
through a film of the sample and a U.V. detector which
permits measurement of absorption by solutes in the
sample. Liquid exiting matrix 20 enters detector 26
which also measures a parameter characteristic of
solute concentration, this time in the effluent, and
delivers a signal representative of that quantity
through line 28. Lines 24 from detector 22 and line 28
from detector 26 enter electronic calculator means 30,
where, for example, the difference between the sensed
absorption maxima in detectors 2~ and 26 is calculated,
and that difference is used by multiplication with a
conversion factor to determine target solute
concentration. The concentration value may be
delivered through line 32 to a display 34.
When detector 22 is omitted, the apparatus of
Figure 5 operates slightly differently. Specifically,
detector 26 detects a first plateau representative of
the concentration of non-target solutes or contaminants
exiting matrix 20, and at a later time, after
breakthrough of the target solutes, detects total
solute concentration. Data points representative of
these sensed plateaus are delivered through line 28 to
calculator means 30 and process as set forth above.
,
Figure 6 depicts another embodiment of the system
of the invention. Its operation is conceptually
identical to that a Figure 5 excepting that effluent
from matrix 20 is returned to detector 22 via line 36.
This permits a single detector to measure the total
so]ute concentration in the sample prior to its
SWE~STITUTE SHIE~ET
:.: ,- , , ~ , . .
WO92/02815 PCT/VS91105~
2~8836~ 1 ~
introduction into the matrix 20, and thereafter to
measure the level of the plateau achievecl in the
effluent prior to breakthrough of the ~arget solute.
Design of this embodiment of the system may require
inclusion of a liquid accumulator (not shown) in
line 38 disposed between detector 22 and matrix 20, or
some other means to insure that all sample has been
removed from the detector 22 prior to the time solute
concentration in the effluent reaches a plateau.
Signals representatives of the solute concentrations
sensed by detector 22 are transmitted through line 24
to calculator means 30 as disclosed above.
Calculator means 30 may be omitted if the purpose
of the device is solely to monitor protein structure.
In this case, the display 34 is adapted to display a
plot of a function representative of protein
concentration in the effluent versus a function
representative of effluent volume. The display thus
produces a curve characteristic of the structural
profile of the protein sample which can serve as a
"fingerprint" of the sample which will identify a given
sample composition and change if the structural profile
of the protein changes.
Assayi~ Solute Concentration
In the case of the each of the embodiments of
Figures 5 and 6, prior ~o beginning an analysis, the
system has been filled with a buffer 16 used to
equilibrate matrix 20 and to assure no solute residues
remain in detectors 22 or 26. To initiate an assay,
the valve lO is adjusted to permit sample 14 to be
introduced into the system impelled by a pressure
gradient created by a pump or syringe, or by means of a
SUE35TIT8JTE SHEIET
. . . ` .". . .
,
WO92/02815 PCT/US91/OS~
. . ,
1 3 2a8~
charge gradient to promote electrophoretic ~ovement
through the ma-trix 20. In the optional embodiment of
Figure 5 and in Figure 6, a data point indicative of
the total solute concentration in the sample is sensed
by detector 22. Thereafter, the sample enters the
matrix 20. '~rget solute begins binding to the binding
sites immobilized in the matrix; contaminants which do
not bind pass through the matrix and emerge in the
effluent. In the embodiment of Figure 5, the buildup
of contaminants in the effluent is sensed by
detector 26; in the embodiment of Figure 6, the buildup
is sensed by return to detector 22. In both cases,
prior to the time target solute saturates the binding
sites in matrix 20 and begins breaking through into the
effluent stream, the concentration of non-target
solute(s) or contaminant(s) in the effluent stream
reaches a plateau, and a signal indicative of the level
of the plateau is passed to calcula~ed 30.
At this point, all information needed to calculate
the concentration of the target solute is available,
and the assay is complete. However, as a check, flow
through the system can be continued until the target
solute breaks through matrix 20 and, together with the
contaminants, produces a higher plateau which should be
equal to the concentration sensed in the sample prior
to its introduction in~to the matrix.
At this point, as an additional self check, if
desired, valve 10 can be switched to direct buffer from
reservoir 16 through the system, thereby washing
detectors 22 and 26 and matrix 20 free of non
specifically adsorbed contaminants but leaving target
solute non-covalently bonded ~o the binding sites in
the matrix. After this wash step, valve 10 is again
SUIBSTITUTE 811 IEET
,
..
W092/02815 PCT/US91/055~
2~883~ 2~
switched to introduce eluent from reservoir 18 through
the system. The eluent serves to elute the target
solute from the matrix 20. The eluted target solute is
detected by detector 22 (in the embodiment of Figure 6)
or 26 (in the embodiment of Figure 5) as a pulse of
solute. Integration of the pulse curve or other
determination of the area under the curve gives an
indication of the quantity of target solute bound
during the assay which, again, can be correlated to the
concentration derived previously.
From the foregoing, it will be appreciated that
design and construction of all components of this
sys~em are well within the skill of the art. Indeed,
many other configurations suitable for the practice of
the process of the invention can be devised, and
additional features incorporated as desired. For
example, the system can be designed to have replaceable
matrix modules, individual ones of which comprise
; 20 binding sites specific for predetermined target
solutes. Since accuracy of the assay is independent of
flow rate, it matters not how one chooses to promote
flow through the system. Thus, for example, a pump may
; be placed anywhere in the fluid flow line.
Alternatively, the sample may be placed in a syringe
and simply rammed through the system.
The calculator means or processor 30 can take
various forms, and indeed, in the broader aspects of
the inventiorl, is not required. A conventional plotter
attached to detector 22 and/or 26 would permit an
operator of a production or purification s--tem to
determine visually by observing plural consecutive
plots whether concen~ration of the target solute and/or
the impurities is changing with time or is constant.
'
~3UIB5T~TE E;HEET
.~.,
.. . .
~. :
.
. ~ , . . .
.'; :
WO92/O~X15 PCT/US91/055~
~08~3
However, calculator 30 may include means for storing
signals representative of data points indicative of the
sensed solute concentration ratios, and correlation
factors, and an arithmetic calculation module which
calculates target solute concentration and/or
contaminant solute concentration. These data may be
displayed digitally in display 34 after each assay.
Alternatively, the data may be used to produce a plot
of target solute concentration over time, or other
desired indication of the state of the system, as a
record of the dynamic behavior of the system under
analysis.
~,
A chromatogram is generated by measuring and
charting a characteristic of the effluent that varies
in proportion to the concentration of detectable solute
in the effluent. In a typical application, commonly
used in commercial chromatography equipment, ultra-
violet radiation is passed through the effluent and the
degree of ultra~violet absorption is charted.
Absorption of U.V. light in such systems is
proportional to solute concentration, provided the
solute is absorptive of this wav~length. It should be
understood, however, that any characteristic of the
effluent which is representative of the concentrations
of analyte and impurities therein can be monitored for
purposes of the present invention.
The abcissa of the chromatogram of Figure l
indicates tlme, and the ordinate absorption. For
illustrative purposes the graph is divided into five
periods which are labelled A, B, C, D, and E. The
- periods define stages of solute concentration in the
effluent during a chromatographic loading cycle that
might be encountered when passing a sample through a
$aJB~3TIT~JT1~ 8HE~ET
.
.. : , . .
.
.~ , . . . .. .
WO92/02815 PCT/US91/05~44
~8~ti 22
conventional affinity chromatography matrix housed in a
column at a high rate, e.g., 1800 cm/hr. It is easy to
see that due to poor resolution the boundaries between
periods must be drawn rather arbitrarily.
Initial period A represents the condition where the
effluent consists entirely of buffer. When the
effluent begins to include impurities from the sample,
the solute concentration begins to rise as shown in
period B. Eventually, an equilibrium concentration
will be reached as depicted in period C. This will
occur when non-specific binding (if any) of impurities
to the matrix has stopped and target solute is being
retained by binding to the matrix so that the
concentration of impurities in the feed is equal to the
concentrations of impurities in the effluent. As
sample continues to flow through the matrix, the target
solute begins to saturate the binding sites of the
matrix. This results in the emergence of target solute
in a gradually increasing concentration in the
effluent, commonly referred to as "breakthrough",
illustrated in period D. When the binding sites are
completely saturated (period E), the sample merely
flows through the matrix and the concentration of
solute in the effluent is equal to the concentration of
solute in the feed.
Note that the ~plateau~ of period C is the critical
information necessary to calculate the concentration of
the target solute, but that the height of the plateau,
and its boundaries, are far from distinct. For samples
containing multiple solute of differing physical
properties, the chromatogram can be far less
informative, and the faster one passes the sample
through the matrix, generally the more the critical
plateau is marked by band spreading.
$UIBSTITUT!E ~3Ç IEE~
WO92/02815 PCT/US91/05544
~3 ~8~
There is shown in FIG. 2 a chromatogram typical of
that obtained by passing the sample very slowly through
the matrix. The single most significant distinction
between the graphs ~f FIGS. 1 and 2 is that the latter
has sharply defined breakthrough points and equilibrium
levels. Per._d A~ of FIG. 2 corresponds to period A of
FIG. 1 and is representative of the period over which
buffer alone constitutes the effluent. FIG. 2 shows
that the matrix becomes saturated with impurities due
to non-specific binding over a short interval so that
an equilibrium concentration of impurities is reached
in period C' at a very well defined point in time.
This is represented in the figure as breakthrough
point B~. The equilibrium concentration of period C~
will be maintained as analyte contained in the solution
is loaded onto the binding sites of the matrix until
those binding sites become saturated. When this
occurs, a second breakthrough point D' will be reached
wherein the concentration of the analyte in the
effluent will become equal to the concentration of the
analyte in the feed solution. The concentration of the
analyte and impurities to~ether will be directly
proportional to the equilibrium concentration of
period E' which follows the second breakthrough
step D~. The difference, therefore, between the height
of plateau E~ and the height of plateau C' can be used
to calculate the concentration of analyte in the feed
solution~
: . .
Additionally, if the capacity of the matrix is
known, by monitoring the amount of solution passed
through the matrix before the breakthrough step D'
occurs, the concentration of analyte in the solution
also can be determined. Since, however, over repeated
uses the binding capacity (the number of binding sites
!
~IJBSTIT~JT~ ~IE~
. . : . ......................... . ~
: ' - :: ' :
W~92/02815 PCTtUS91/055~
2as.s3~ ~,~
in the matrix) will decrease, it will more often be the
case that the concentration of analyte in the solution
will be determined based upon the principles discussed
above. The concentration so determined, therefore, can
be used in conjunction with the timing of the
breakthrough step D' to determine how many binding
sites remain in the matrix.
Line F' in FIG. 2 represents the point at which
solution has ceased being passed through the matrix,
and the effluent once again comprises only buffer. A
third way to determine the amount of analyte in the
solution is during desorption of the analyte from the
matrix by way of passing an eluent through the matrix
to free the analyte from the binding sites. This
process is represented in the figure by the behavior of
the chromatogram during period G'. The area under the
curve in this period is directly proportional to the
amount of analyte bound to the matrix as of the
breakthrough point D'. It is clearly possible,
therefore, to check the accuracy of-the determination
of target solute concentration made based on the height
of step D' to that determination made based upon the
area under the curve during period G~.
Figure 3 shows a chromatogram of the type which can
be produced in the apparatus of Figure 5 with optional
detector 22, or in the apparatus of Figure 6.
Period A' represents the interval when detector 22 is
measuring total solute content in the sample prior to
i~s entry into matrix 20. After leaving detector 22
the sample enters matrix 20, and during th~ time the
sample is displacing buffer in the matrix, the
concentration of effluent from the m2trix shows a
solute free state as illustrated during interval B~.
SaJ1135TlTlJTE SHI~ET
WO92/02815 PCT/US91/05~
25 2~36~ :
Impurities break through at C' and their concentration
is represented a~ interval D' in Figure 3. ~s flow
through the matrix proceeds, target solute breaks
through at E , and total solute concentration in the
eluent, indicated at F~, equals the conclentration
indicated by the interval A'. The data necessary to
know target solute concentration is in hand as soon as
the level of plateau D' is known with confidence, and
is here indicated by way of example by a vertical
dotted line. The time it takes for the level D' to be
established is dependent on flow rate and on the volume
of the column, which is proportional to length B~.
Small volume columns which can be run at high flow
rates are therefore preferred for rapid analysis.
With repeated use the chromatography matrix will
break down in the sense that its capacity will
decrease. This will not, however, affect the accuracy
of the data generated in accordance with the principles
of the present invention. All that occurs is that the
length of plateau C' in Figure 2 or D' in Figure 3
becomes shorter, as target solute breaks through
sooner. Neither the height of the breakthrough plateau
representative of the concentration of the impurities
nor the solute concentration maxima change, and
accuracy is not compromised.
It is possible, however, for a problem to develop
if, in relation to the concentration of the analyte,
there is an inadequate number of binding sites. That
is, if the concentration of analyte in the solution is
so high with respect to the number of available bindiny
sites, or fluid velocity so high, that the binding
sites become saturated almost immediately, a condition
will result wherein, rather than displaying two well
SUBSTITUTIE~ SHIE~T
. .
: . :
WO 92/02815 PCr/lJ~i91/05544
æ~83~ ~ s
defined steps, a chromatogram will show only one step.
This situation is portrayed in the chromatogram
depicted in FIG. 4, wherein a single breakthrough
point X represents the simultaneous saturation of both
analyte and impurities in the chromatography matrix.
In this situation it is obviously impossible to discern
equilibrium concentration levels. While the area under
the-curve in period G~ will still be proportional to
the amount of analyte captured by the matrix at the
point of saturation, saturation has occurred so quickly
that reliable determination of concentration based on
saturation cannot be made. To remedy this problem, the
feed solution need only be diluted with, e.g., buffer,
prior to being passed through the matrix, so that
breakthrough of the non-target solute can be
distinguished. After dilution, a chromatogram such as
that depicted in either FIG. 2 or 3 will be generated
wherein an equilibrium concentration of impurities is
established in the effluent before any analyte appears
in the effluent.
As has been mentioned throughout this description,
a preferred aspect of the present invention involves
high speed assays, e.g., less than lO seconds. The
above discussed analyses can be performed in periods
substantially shorter than one minute, often shorter
than 30 seconds, and frequently less than 10 seconds,
if one employs a small volume column containing a
matrix medium of the type described below.
The present invention also helps to increase the
speed with which assays are carried out by being able
to provide meaningful data with only a very small
sample of feed solution. For example, assays routinely
can be performed on microliter sized samples. For this
SUBSTITIJTE~ SHEET
WO92/02815 PCT/US91/05~
- 20~3
21
purpose, the chromatography matrix can be supported in
a column having suitable dimensions. For extremely
small samples, it is possible to pass the sample
through a capillary tube having an inner surface
coating of binding protein.
One currently preferred matrix i5 non-porous (or
very low porosity) affinity-based silica particles. A
particularly advantageous matrix medium is POROST M
brand column packing materials which may be obtained
commercially from PerSeptive BioSystems, Inc.
(Cambridge, MA). These materials are produced through
suspension polymerization techniques and classified to
the desired particle size range. POROS~ M columns have
been shown to have significantly reduced band spreading
in high speed assays, thereby allowing analysis
according to the practice of the present invention to
be performed in extremely short periods of time.
The invention will be understood further from the
following nonlimiting examples~
EXAMPLE l.
The subtractive frontal analysis technique is
demonstrated using a Protein A column to measure the
concentration of human Immunoglobulin in solution.
Various concentrations and purity levels of IgG were
analyzed using an HP 1090 liquid chromatograph
(available from Hewlett Packard, Waldbronn, GmbH) and a
"chem station" (available from Hewlett Packard). Pure
samples of human IgG (available from Sigma Chemical
Company, St. Louis, Missouri) are prepared in
concentrations ranging from 0.01 - 10 mg/ml. These
samples are first pumped into a detector flow cell
SUBg~TlTUTE SHEET
W092~02815 PCT/US91/05~
2088360
~ .
without a column in line, and the absorbance at
equilibrium is quantified to obtain a calibration
curve. Next 250 ~1 injections of IgG samples
contaminated with various amounts of BSA ( Sigma
S Chemical Company) are mixed in 10 mM sodi.um phosphate
and 150 m~ sodium chloride to pH 7.4 and run on a 2.1 x
30 mm POROS~ M A/M column, a commercially available
perfusive column comprising 20 ~m polystyrene/divinyl
benzene beads with a pellicular coating of protein A.
10 As an example, absorbance plateaus of the output of 0.5
mg/ml IgG mixed with 0.5 mg/ml BSA at flow rates of 1
ml/min and 0.1 ml/min are shown in Figures 7 and 8,
respectively. The profiles indicate breakthrough of
BSA and IgG are clearly distinguishable.
The accuracy of this system may be checked by
analyzing samples of known concentration. A
calibration curve relating the detector absorbance and
concentration for a single component is needed to
obtain meaningful comparisons with the known
concentration in a sample mixture. Figure 10 shows
such a calibration plot for human gamma globulin using
the 1090 diode array detector at 280 nm. Two
independent sets of experiments yielded the data and
correLation shown. Human gamma globulin is about 92.4%
~ pure, as determined by subtractive frontal analysis.
; Therefore this factor has to be included when
converting a measured absorbance to equivalent IgG
concentration. The calibration plot also reveals the
linear range to be up to about 4 mg/ml for pure IgG or
about 3000 mAu total for a mixture.
A set of experiments were conducted using mixtures
of different concentration and purit~ of IgG (using BSA
as contaminant). Comparing the expected absorbance
against the actual column absorbance and correcting for
.~
,~
SU13STlTq.lTE SHEFr
'
,
:
WO92/02815 PCT/US91/05544
- 29 20883~
the non-retained plateau value as shown in Table 1 and
described below allows an accurate estimation of the
IgG concentration~
As is evident from the data, the prototype device
can detect varying IgG concentration in the presence of
varying amounts of contaminating protein, here BSA.
Certain experiments detailed in Table 2 revealed
conditions under which the method is less accurate.
Errors result from a number of factors, including the
calibration plot, the value of the absorbance at the
plateau, and measurements outside the linear range. In
addition, in these experiments there may also be errors
in the measurement of amounts of proteins used in the
lS test mixture. Errors due to the calibration are found
in the examples in Table 1. Errors in estimating the
absorbance of the front emerging from the column (i.e.,
the non-bound contaminant) are magnified for cases
where the IgG purity is 10% or less as shown in
Table 2. For example a 5% error in the estimate of the
plateau absorbance yields a 45% error in the IgG
concentration estimate for a sample of 10% pure IgG.
The same 5% error in the plateau only leads to a 5%
error in the concentration estimate for a sample of 50%
purity. Errors caused by exceeding the linear range
are evident in Table 2. In this case, a simple
solution would be to dilute the sample appropriately.
Alternatively, the use of a slightly different
wavelength (higher or lower) than 280 nm will produce a
weaker signal response and can be used for samples of
high concentration.
These data, taken collectively, demonstrate the
utility and feasibility of the assay procedure.
Routine engineering principles may be used in the
design of a commercial device to improve accuracy and
$WBSTOTUTE SHFE~T
W092/0281~ PCT/US41/05~
208836~
ease of use. For example, computing means 30 may be
programmed to require 5 or 10 consecutive readings over
an appropriate time interval to be within some small
margin of error before a "plateau" is recognized and
recorded.
3E31JBSTITUTE~ SHE~ET
,.
. ' ' . ~'
WO92~02815 PCT/~S91/055~
3 ~i
3 1
Table 1
A B C D E F G
5 IgG BSAno col column exp conc % VAR
mg/ml mg/ml mAU mAU IgG IgG IgG
0.5 0.5530 200 570 0.49 1.61
0.5 0.055350 41 386 0.45 7.87
1 1 1110 395 1139 1.07 6.59
1 0.111725 84 772 0.96 4.45
1,5 1.51595 605 1709 1.48 1.61
1.5 0.15S1170 135 1153 1.54 2.86
~.05 0.0553 18 57 0.05 4.35
0.05 0.0056 37 3 39 0.05 1.37
0.1 0.1106 37 114 0.10 2.86
0.1 0.01178 8 77 0.10 4.35
0.2 0.2221 79 228 0.21 5.84
0.2 0.022lq4 18 15q 0.19 6.09
0.4 0.4425 15~ 456 0.40 0.50
0.4 0.044294 37 309 0.38 4.22
0.1 0.2333 163 95 169 0.10 1.37
0.2 0.4666 322 178 338 0.21 7.33
0.9 0.9333 645 348 676 0.44 10.68
0.5 2 1091 745 1189 0.52 3.16
0.3 2.71170 964 1333 0.31 2.36
0.15 1.35610 496 666 0.17 13.~9
0.075 0.675 273 226 333 0.07 6.5~
30 avg deviation 4.73
Table 2
0.05 0.45200 160 222 0.06 19.3
0.1 1.9800 718 857 0.12 22.2
1 92970 2680 4443 0.43 56.
1.5 13.53020 2900 6665 0.18 88.1
~3lJBSTITllJT~ IEI~T
.
.: . ' ' ' '' ' ' - ' ' ' ' :
::,.' ` ' . : , ' '
'; . ,., ' ' ' ' ';'
W092tO2815 PCT/US91/~55~
~08836~ ~2
Table Leqend
A: ~nown IgG concentration (CIgG) in mg/ml;
B: Known BSA concentration (CBsAJ in mg/ml;
C: Absorbance in milli absorbance units (mAU) without
the column in line (corresponding to effluent after
saturation of all binding sites in column, or height of
second plateau);
D: Absorbance in mAU with the column in line
(corresponding to absorbance of BSA alone - first
plateau caused by BSA breakthrough);
E: ~Expected~ value of absorbance without the column
in line calculated by:
(CIgG~(726 mAU) + (CBsA)(413 mAU) where 726 mAU is IgG
absorbance factor from calibrator curve of Fig. 10 and
413 is BSA absorbance from calibration curve (slope);
F: Detected IgG concentration - (C-D/726 X 0.924)
where 0.924 is the percent IgG absorbed); and
G: Detected IqG - Actual IqG
Actual IgG X 100.
While there is some band spreading in the second
IgG-containing front of Figure 7, the lOX increase in
speed between the runs represented by Fiqs. 7 and ~
significantly offsets the disadvantage of the delay of
the second plateau. In fact, as discussed above, the
second plateau is not necessary for calculating the
concentration of the sample where the original r
absorbance of the test solution is known, making the
band spreading of the second front irrelevant.
;0E~3TITUTE 8~1E3~T
. .
,. , , . : `
, . : . . ..
.~ . .
. . ~ . ..
W~92/0~815 PCT/US91/055~
3 3
EXAMPLE 2
To demonstrate the practice of the invention in
characterizing the structural profile of a protein, a
Protein A column of the type described above is loaded
with Mouse Gamma Globulin (Sigma). Mouse gamma
globulin actually contains several subclasses of IgG
which vary with respect to their bin~ing affinity for
Protein A. If a sample of 2 mg/ml protein in 10 ml PBS
plus 1% MeOH is run on a 2.1 x 30 mm column at a flow
rate of 0.2 m./mn followed by elution with 0.15 M NaCL
+ 2% Acetic Acid + 1% MeOH, a frontal chromatogram as
shown in Figure 9 is produced. The shape of the curve
is indicative of the affinity of the various IgG
species in the sample for protein A on the POROS A/M
column, and changes in the structural profile of
protein in the sample will induce variations in this
curve. Thus, the curve constitutes a "fingerprint"
uniquely identifying this particular mix and condition
of IgG species, and repetition of the procedure with
other samples will produce a curve permitting one to
compare the structural profile of the samples.
Other embodiments are within the following claims.
~3lJlE3STlTlJll-E SHE~
:. .
. .. .. . . . ..
.
.
.
