A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
1
2
3
4
5
6
7
8
9
This is the author version of the paper, the definitive version is the ACA version.
A Fluorescence Anisotropy Method for Measuring Protein Concentration
in Complex Cell Culture Media.
Radu Constantin Groza, Amandine Calvet, and Alan G. Ryder.*
Nanoscale Biophotonics Laboratory, School of Chemistry, National University of Ireland, Galway,
Galway, Ireland.
10
* Corresponding author: Email: alan.ryder@nuigalway.ie Phone: +353-91-492943.
11
12
Postal address: Nanoscale BioPhotonics Laboratory, School of Chemistry, National University of
Ireland, Galway, University Road, Galway, Ireland.
13
14
Running Title: Protein quantification in media using multi-dimensional fluorescence anisotropy.
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
Abstract:
The rapid, quantitative analysis of the complex cell culture media used in biopharmaceutical
manufacturing is of critical importance. Requirements for cell culture media composition profiling, or
changes in specific analyte concentrations (e.g. amino acids in the media or product protein in the
bioprocess broth) often necessitate the use of complicated analytical methods and extensive sample
handling. Rapid spectroscopic methods like multi-dimensional fluorescence (MDF) spectroscopy
have been successfully applied for the routine determination of compositional changes in cell culture
media and bioprocess broths. Quantifying macromolecules in cell culture media is a specific
challenge as there is a need to implement measurements rapidly on the prepared media. However, the
use of standard fluorescence spectroscopy is complicated by the emission overlap from many media
components. Here we demonstrate how combining anisotropy measurements with standard total
synchronous fluorescence spectroscopy (TSFS) provides a rapid, accurate quantitation method for cell
culture media. Anisotropy provides emission resolution between large and small fluorophores while
TSFS provides a robust measurement space. Model cell culture media were prepared using yeastolate
(2.5 mg mL–1) spiked with bovine serum albumin (0 to 5 mg mL–1). Using this method protein
emission is clearly discriminated from background yeastolate emission, allowing for accurate bovine
serum albumin (BSA) quantification over a 0.1 to 4.0 mg mL–1range with a limit of detection (LOD)
of 13.8 µg mL–1.
Keywords: Fluorescence, Anisotropy, Total Synchronous Fluorescence Spectroscopy (TSFS),
Bioprocess, Protein.
Author Version of paper:
Page 1 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
1. Introduction
Macromolecule production using cell culture requires the use of complex nutrient media to
maintain optimal physicochemical conditions for productive growth. Cell culture media have very
diverse and complex chemical compositions. Despite the recent trend towards the use of chemically
defined media, there are many systems where components like complex biogenic hydrolysates [1] or
proteins [2] are key constituents added to improve cell growth. Most media for mammalian cell
culture are complex mixtures of components with component concentrations ranging from ppm to 510% w/w, making comprehensive analysis challenging. However, the analysis and control of cell
culture media variance is critical in industrial biotechnology. Conventional methods for media
monitoring and specific component analysis consist of different hyphenated chromatographic
techniques that make routine analysis time consuming and expensive [3, 4]. For measuring protein
concentration in media solutions (without purification) the only viable option is to utilise one of the
specific colorimetric based assays (e.g. Lowry, Bradford, and BCA assays) [5]. Unfortunately these
methods require various degrees of sample manipulation and reagent addition, neither of which are
desirable.
Vibrational spectroscopy has been used for monitoring compositional changes in cell culture
media, but problems can arise with spectral resolution and weak analytical signals in solution [6-10].
This is particularly the case for the macromolecular constituents of media like insulin and serum
albumin where concentrations are low, typically 1–10 µg mL–1and 1–5 mg mL–1 [11] respectively.
Fluorescence spectroscopy is an attractive alternative for near real time on-line/at-line and nondestructive monitoring of biogenic process and raw materials quality control [12]. This is because
some significant constituents of culture media are intrinsically fluorescent, like riboflavin, pyridoxine,
and the amino acids phenylalanine (Phe), tryptophan (Trp), and tyrosine (Tyr), and they can be
detected at low concentrations (mg–µg/ml) [12-15]. However, the compositional complexity of cell
culture media causes problems with fluorophore band overlap. For example, the intrinsic fluorescence
a protein like bovine serum albumin (BSA) is dominated by emission from Trp with lesser
contributions from Tyr and Phe [16].
This overlap of media amino acid and protein emission makes protein quantification difficult
using standard emission spectroscopy. The situation can sometimes be resolved by the use of multidimensional fluorescence (MDF) spectroscopy techniques like Excitation Emission Matrix (EEM) or
Total Synchronous Fluorescence Spectroscopy (TSFS) where one interrogates a much wider
fluorescence emission space. In EEM measurements, a series of emission spectra are collected, each
at a different excitation wavelength. The excitation wavelength changes incrementally so that a multidimensional plot of emission spectra versus excitation wavelength is obtained, the EEM (λex× λem×
Iλex,λem) [17]. In TSFS (for example constant-wavelength mode), both excitation and emission
monochromators are scanned simultaneously at the same rate (Δλ). By sequentially collecting spectra
over a range of Δλ increments > ~10 nm one can generate a multi-dimensional TSFS spectrum (λem×
Δλ × Iλem,Δλ) without any large Rayleigh scatter artefacts [18]. The key advantage of TSFS over EEM
is the ability to easily avoid Rayleigh scatter artefacts for example when characterising very complex
multi-fluorophore mixtures such as crude petroleum oils where light scatter can be a big problem [19].
One disadvantage with MDF-based methods is that data analysis of the complex data generally
requires the use chemometric techniques such as multi-way PCA [20] or factor based methods like
PARAFAC [21] or MCR [22].
Combining TSFS/EEM data with chemometrics provides a rapid method for the
characterization of individual fluorescent components in complex mixtures such as cell culture media
and bioprocess broths [12, 15, 23, 24]. In the context of rapid quality control/assessment of protein
Author Version of paper:
Page 2 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
containing media formulation one needs to measure protein concentration rapidly and accurately
during media preparation while simultaneously verifying correct gross composition. Gross
composition can easily be assessed using EEM [12, 23] and sometimes it is feasible to quantify
specific fluorophores using a modified standard addition method [15]. However, this approach only
works for a narrow analyte concentration range because of sensitivity to varying inner-filter/matrix
effects as composition changes. Therefore for the protein concentration measurements we use another
approach and exploit the intrinsic size (strictly speaking, the rotational correlation time) related
information in the fluorescence emission.
Fluorescence anisotropy characterizes the polarization state of the emitted light, and this in
turn can be related to the physical size and/or restricted movement of a fluorophore [25, 26].
Polarization measurements can introduce a new dimension to regular MDF spectroscopy of complex
materials that offers a new insight on the origin of the emitting fluorophore [27, 28]. Components in
complex mixtures can be differentiated by their hydrodynamic volume and rotational speed [29].
Fluorescence anisotropy as a technique has been applied to the study of protein interactions, protein
structure, reaction kinetics that occur with changes in molecule rotational time, changes of local
viscosities and protein distribution in cellular membranes [16, 25, 30-34]. However, the method has
only rarely been applied to complex biogenic mixture analysis [35-38] and the use of MDF with
anisotropy is not common. Polarized EEM fluorescence spectroscopy has been used to study multicomponent fluorescent dyes mixtures and spin-coated polymeric thin films [39-41], but we are
unaware of any application to biogenic systems.
This study showed how anisotropy incorporated into MDF measurements can accurately
characterise and quantify protein in cell culture media. Fluorescence anisotropy discriminates
between fluorescence emissions of small molecule media components and the large molecule protein
component. Yeastolate (a complex hydrolysate) spiked with BSA was used as a realistic model cell
culture media akin photophysically to those used in industrial biotechnology. BSA emission [42-44]
overlaps completely with the complex emission of yeastolate which contains appreciable amounts of
both Trp and Tyr [12, 24].
2. Materials and Methods
2.1 Materials: Bacto™ TC and DifcoTM Yeastolate (the water soluble portion of autolyzed yeast) was
purchased from BD. Bovine serum albumin (>99%), Rhodamine B (Rh-B), glycerol (99.5%),
phosphate buffer saline (PBS) tablets, and Bradford reagent were purchased from Sigma-Aldrich. All
materials were used without further purification. Stock yeastolate solution (5 mg mL–1) was prepared
by first dissolving 1.25 g yeastolate in sterilized high purity water (250 mL) and then membrane
filtering (0.22 µm) before aliquotting under sterile conditions. A sterile, filtered BSA stock solution
(10 mg mL–1) was also prepared. 80 different mixture samples were prepared by spiking 2 mL of the
yeastolate solution with BSA aliquots varying from 0 to 2 mL. Samples were brought to a final
volume of 4 mL by PBS buffer addition. The sample set, had a BSA concentration range of 0 to 5 mg
mL–1 with a constant yeastolate concentration of 2.5 mg mL–1. All samples were prepared in triplicate
and reproducibility was determined at 0.5 and 4 mg mL–1 BSA by preparing each sample 10 times. A
second sample set was prepared in the same manner using PBS buffer in place of the yeastolate
solution. All samples were stored at –70ºC to prevent compositional changes in yeastolate and left to
defrost overnight prior to analysis. These were then pipetted into 1cm pathlength quartz cuvettes
(Lightpath Optical, UK) and sealed under sterile conditions. The same samples were also analysed
using the Bradford assay with an excess of 30 parts reagent to one part sample, according to the
product technical bulletin (Sigma, Bradford Reagent Technical Bulletin, Product Number B 6916).
Author Version of paper:
Page 3 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
Samples were analysed almost immediately after addition of Bradford reagent directly into the quartz
cuvettes.
2.2 Instrumentation and Data Collection: TSFS spectra were collected at 298 K (unless otherwise
stated) using an Eclipse Fluorescence Spectrometer (Agilent) fitted with a manual polarizer, and a
temperature-regulated Single Cell Cuvette Holder CUV-TLC-50F (Quantum Northwest). In order to
produce a fluorescence anisotropy plot, four differently polarized TSFS spectra were obtained for each
sample with the excitation and emission polarizer positions set for vertical-vertical (VV), verticalhorizontal (VH), horizontal-vertical (HV) and horizontal-horizontal (HH). TSFS data were obtained
for an excitation range of 280–370 nm with a data interval of 1 nm (scan speed of 600 nm/min) and a
Δλ interval of 20–150 nm for both monochromators with a Δλ increment of 2 nm. 10 nm
excitation/emission slits were used. Each individual TSFS plot was a 66×41 data matrix (total of 2706
spectral points) and the four TSFS measurements took 15 minutes each to collect. Data collection was
performed in random order over 2 months. A Shimadzu UV1601 UV-visible spectrometer was used
for Bradford assay absorbance measurements at 595 nm (1 cm pathlength quartz cuvettes).
2.3 Data pre-processing and analysis: Data pre-processing and anisotropy calculations were
performed using MatLab software (ver. 7.0.1) and in-house written codes. To avoid the use of noisy
data in the anisotropy calculations, a minimum threshold of 10% of the HH polarized maximum
fluorescence intensity was set. All spectral data points below the threshold were set as missing values.
This was done because the relative fluorescence signals in the HV, and VH data were often much
lower, and thus noise from the weaker data-points had an adverse effect on the calculated r values. In
short, for these weakly fluorescent regions, the anisotropy values were unreliable using the data
collection parameters employed. In practice the 10% threshold was found to be effective and enabled
the calculation of accurate and reproducible the r-values. A second consideration for implementing
the threshold limit was that it enabled a better visual interpretation of the anisotropy TSFS spectra (see
supplemental information, SI). The anisotropy value for each selected matrix data point was
calculated using the standard anisotropy formula (SI) and then the values used to produce anisotropy
based TSFS or EEM matrices. Another practical consequence of the low intensity in the HV data was
that for aniso-TSFS measurements wider excitation/emission slits were required to increase the signalto-noise ratio. Alternatively one could have increased the integration time, but this made data
collection times longer and reduced sample throughput for the instrumentation used here.
3. Results and Discussion
3.1 Validation of anisotropy measurement method: The first experiments undertaken were to validate
the accuracy of the anisotropy measurement across both EEM and TSFS datasets when a 10%
threshold limit was implemented. These measurements were performed using a 1 µM solution of
Rhodamine B (Rh-B) in a glycerol/water solution (95:05 v/v.) at 283 K. Anisotropy of the Rh-B
solution was constant over the excitation range λex = 530–565 nm (λem = 573 nm) with an average rvalue of 0.3916±0.0027 and a relative standard deviation (RSD) of 0.69 % (n = 141). Changes in scan
speed (150 to 3000 nm/min) did not produce any significant change in the measurement accuracy. In
addition, the anisotropy of 1µM Rh-B solution calculated over this emission range had an excellent
linear correlation (R2 = 0.9957) with increasing glycerol concentration (0-40% v/v., n = 5, T = 293 K),
see SI.
The anisotropy EEM data matrix (λex/λem = 550–590/550–640 nm) had a constant r-value
across the entire emission space, apart from the spectral regions which included Rayleigh scatter (see
SI). When the anisotropy was calculated for raw EEM data, Rh-B in 95% glycerol had an average rvalue of 0.392±0.017, RSD = 4.33% (n = 1388). The average anisotropy of the same Rh-B solution
calculated using raw TSFS data was 0.391±0.009, RSD = 2.30% (n = 1100). The results were in good
Author Version of paper:
Page 4 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
agreement with published literature values (0.3916±0.0027 for 95:5 glycerol:water at 283 K) [26] and
0.350±0.003 for 90:10 glycerol:ethanol at 293 K [45]. Elimination of Rayleigh scatter reduced
measurement error significantly because Rayleigh scatter is highly polarized and thus its inclusion will
lead to artificially high r-values [26].
Figure 1: HH polarized TSFS spectra of a) 2.5 mg/mL yeastolate (landscape), b) 2.5 mg/mL
yeastolate spiked with 2.5 mg/mL BSA (landscape), c) 2.5 mg/mL yeastolate, d) 2.5 mg/mL BSA, and
e) 2.5 mg/mL yeastolate spiked with 2.5 mg/mL BSA, and f) 2.5 mg/mL yeastolate spiked with
5mg/mL BSA.
3.2 Spectral analysis: For cell culture media analysis a relatively broad spectral range was selected to
include the emission domains of Trp, Tyr, and Riboflavin. This also led to the inclusion of Raman
scatter in the TSFS/EEM data; however, this should not be a significant issue where the samples have
a relatively strong fluorescence emission, which is the case with most cell culture media both
chemically defined and biogenic hydrolysates [12, 15, 24, 46]. For more weakly emitting media, the
TSFS data range could be adjusted to avoid collection of the Raman band arising from the O-H
stretching band of water. Rayleigh scattering was avoided in the media TSFS spectra by use of a Δλ
setting of >10 nm which both eliminated the need for light scattering removal by computational means
Author Version of paper:
Page 5 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
and improved the anisotropy calculation accuracy. A second TSFS benefit was that data collection
was approximately 1.7 times quicker than EEM (for the same spectral region).
Yeastolate spiked with BSA produced weaker fluorescent emission intensity than BSA alone
in buffer solution at concentrations above 0.5 mg mL–1 BSA due to quenching and inner filter effects
(IFE) between all the various components present the complex mixture (see SI). As to which
components affect EEM emission, one has to consider not only the fluorophores, but also the nonemitting chromophores, and the quenchers present in the sample matrix. All of these species act in
concert and it can be very difficult to separate out individual contributions from a complex sample
matrix like cell culture media. For example, the quenchers could be any of the paramagnetic ions
present in the media like Cu2+, Zn2+ and Mn2+ [17, 47]. Another important quencher present in
yeastolate is free histidine (0.5% w/w according to product technical manual) which can quench
emission from both free and bound Trp [48]. To further complicate matters, the quenching rate will be
different for every quencher-fluorophore combination. For example free and protein bound
fluorophores (e.g. Trp) have different accessibilities and thus will be affected differently by quenchers
which are charged or are relatively large [47]. For the non-emissive chromophores there are a wide
variety of molecular species present in yeastolate (too many to detail), all of which will either absorb
the excitation or emission light to varying degrees. Finally in terms of fluorophores, yeastolate
contains a variety of which Trp, Tyr, Phe, FA, and Rf are the most dominant [24]. Identifying specific
component effects is not however always required as the cell culture media formulation should have a
static composition when used in industrial applications [12]. This means that the fluorescence
emission of the small molecule contribution (and all the processes that act on the emission) should be
stable and the only variation arises from the variation in protein concentration.
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
The HH polarized TSFS spectra (Figure 1) of the model media, yeastolate, and BSA in buffer
solution clearly showed the spectral overlap between the emission of free Trp/Tyr amino acids and
Trp/Tyr residues in BSA at λex = 290–310 nm and Δλ = 20–120 nm. BSA exhibited, as expected,
stronger fluorescent emission with the VV and HH measurements compared to VH/HV, due to its
slower rotational speed. The most polarized emission occurs at λex/λem = 300/340 nm which
corresponds to Trp emission (blue shifted compared to free Trp in aqueous solution because of
incorporation into the more hydrophobic domains of the protein). For yeastolate the maximum
intensity in the HH plots occurred in two regions λex/λem = 300/360 nm (closer to free Trp emission
[15]) and at λex/λem = 335/425 nm. The latter band could be related to pyridoxine [49], NADH [50], or
other fluorophores present in yeastolate. The polarized fluorescence emission spectra displayed very
different intensity maxima as the polarizer settings were changed (see SI, Figure S-6) and were
significantly lower when compared to unpolarised TSFS data. It was also noticeable that the points of
maximum fluorescence intensity changed between the HH/VH and HH/VV plots with the HH data
have the best correlation with protein emission. As BSA concentration increased relative to
yeastolate, we observed in the HH plots that the λex/λem = 300/360 nm band blue shifted to ~ λex/λem =
300/345 nm and then became more intense relative to the λex/λem = 335/425 nm band. The position of
the λex/λem = 335/425 nm band appears to remain relatively constant, with possibly a small red-shift in
the excitation wavelength.
237
238
239
240
241
242
The aniso-TSFS plot of pure yeastolate solution (Figure 2a) showed that most of the emission
space has negligible anisotropy. For the amino acid emission region (~λex/λem = 290–310/300–430
nm) r = 0.036 ± 0.006, which was due to the small molecular size and high rotational speed of the
small molecule fluorophores. There was a small region of slightly higher anisotropy starting at λex/λem
= 320 / 360 ± 10 nm which most likely correspond to highly polarized Raman scatter [26] from the OH stretching vibration (~3400 cm-1). We also considered that this artefact could have been caused by
Author Version of paper:
Page 6 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
243
244
245
246
the presence of some relatively high molecular weight fluorophores in the Bacto grade yeastolate used.
However, when ultra-filtered yeastolate (M.W. cut-off <10 kDa, DifcoTM, BD) was tested we observed
a similar artefact pattern and concluded that the artefact was related to Raman scattering (see SI).
247
248
249
250
251
252
Figure 2: Anisotropy TSFS contour plots of: a) 2.5 mg/mL pure yeastolate solution, b) 2.5 mg/mL
yeastolate spiked with 5 mg/mL BSA; c). yeastolate spiked with 0.5 mg/mL BSA, and d). pure BSA
(5 g/L in buffer). Light grey represents the highest measured anisotropy while the white areas
represent the no data areas excluded by the 10% threshold limit.
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
The aniso-TSFS contour plot of the model cell culture media doped with BSA (Figure2b) was
very different. Unlike the homogeneous small molecule fluorophore, Rh-B (see SI) it showed
significant variations in the anisotropy values across the emission space due to the presence of
multiple Trp and Tyr fluorophores in BSA. This anisotropy pattern was also clearly visible in the
aniso-TSFS plot for pure BSA in buffer (Figure 2d). The striped pattern observed was generated by a
variety of factors the most important of which was Förster Resonance Energy Transfer (FRET) from
the Tyr to Trp residues in BSA. A second major factor was the fact that the two Trp residues (Trp-134
and Trp-212) in BSA occupied different environments. Trp-212 in BSA was located in a hydrophobic
micro-environment in HSA (sub-domain IIA), while Trp-134 was localized in sub-domain IA, which
was more exposed to solvent and thus in a more hydrophilic environment [51]. Thus Trp-212
emission was slightly blue shifted with respect to the Trp-134 emission. The high anisotropy part of
the plot related to excitation wavelengths of >300 nm which corresponds to direct excitation of Trp
(and proportionally more of the Trp-212 residue). Emission from directly excited Trp will have its
polarisation preserved more so that the Trp excited by FRET from the Tyr residues. The emission of
each Trp residue was in itself heterogeneous [52], which also contributed to the change in anisotropy
Author Version of paper:
Page 7 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
along the vertical λ axis. As the excitation wavelength decreased below ~300 nm, proportionally
more excitation of the Tyr residues occurred, and the corresponding emission comprised of both
directly excited emission from Trp and Tyr, Trp emission from hetero-FRET, and also Tyr emission
via homo-FRET. Thus, as the source of the emission becomes more complex the overall anisotropy
decreased. It was also clear that the anisotropy banding in the aniso-TSFS plots was subtly different
for the BSA in buffer and when mixed with yeastolate. In the BSA-yeastolate mixtures, the source of
emission became more complex and one had to consider the impact of all the additional small
molecule fluorophores and quenchers now present. There were a number of processes and interaction
pathways possible, each of which influenced the anisotropy pattern: 1). radiative energy transfer
between small and large fluorophores, 2). small fluorophores from yeastolate binding to BSA, and 3).
non-fluorescent small molecule ligand binding to BSA causing structural changes, which resulted in
turn to changed photophysical properties. Ligand binding to BSA and the impact on fluorescence
properties is well documented in the literature [47, 53], and outside the scope of this study.
For BSA-yeastolate mixtures we also observed a small anisotropy increase outside the amino
acid region, with r = 0.041 to r = 0.098 ±0.010 for the λex range between 330 and 370 nm. The source
of this increase was identified conclusively, and it may not be entirely possible with a complex poorly
characterised media like yeastolate. There were two likely contributing factors; first, the observation
was due to a combination of energy transfer and quenching effects that were characteristic of complex
multi-fluorophore mixtures. This was partly polarized light emission from BSA being radiatively
absorbed and re-emitted by the long wavelength emitting vitamins like pyridoxine which is known to
be present. The second explanation was that the partly-polarized emission observed at these longer
wavelengths was due to fluorophores absorbed by BSA. These absorbed fluorophores are in a
restricted environment and thus displayed a much increased rotational correlation time. For
pyridoxine the absorption process (although not the anisotropy change) has been monitored using
synchronous fluorescence [54]. Irrespective of which process dominated and the specific species
involved, the observation was not significant in the context of the protein quantification assay, but was
significant in terms of media characterisation and is the subject of on-going investigations.
3.3 Quantification of BSA in model cell culture media: Since BSA was a multi-fluorophore
macromolecule with anisotropy values that were not constant across the entire emission range a data
matrix had to be selected that contained sufficient anisotropy variation to produce a representative rvalue. BSA quantification was performed using anisotropy values obtained from the spectral region of
maximum protein fluorescence emission (λex = 295–306 nm, Δλ = 30–58 nm, 14×11 data matrix)
where the r-value was relatively constant and large. For example the anisotropy values of BSA in
buffer solution remained relatively constant (0.223 ± 0.004) over the 0.1 to 5 mg mL–1 concentration
range (for a 14×11 data matrix) and were in good agreement with a previously published value of
0.219 ± 0.002 obtained for dansyl-labeled BSA [55]. Anisotropy-TSFS measurements of the cell
culture mixtures also showed good reproducibility, with an RSD of 0.89% at 4 mg mL–1 and 5.14% at
0.5 mg mL–1 BSA concentration respectively from 10 replicate measurements on separate samples.
When data matrices varying in size from 8×5 to 20×17 centred on the same maximum emission were
used, the variation in average r-values was no more than ~2%. While small changes in matrix size did
not have major influence on r-value accuracy, it still was critical to specify a fixed region for
quantification.
Author Version of paper:
Page 8 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
312
313
314
315
316
317
318
319
320
321
322
323
324
Figure 3: Comparison between aniso-TSFS and Bradford protein assays for yeastolate spiked with
BSA. a) relationship between anisotropy/absorbance of the coloured protein complex with BSA
concentration; b) Log-linear plots of BSA concentration versus anisotropy (●) and absorbance () for
BSA-yeastolate mixtures and pure BSA in buffer (▲). c) relative standard deviation of both methods
versus protein concentration range. Calculated anisotropy (λex = 294-305nm, Δλ=30-58nm, 14×11
data matrix, average from triplicate data).
To validate the measurement method we compared the aniso-TSFS method with a common
colorimetric method for protein concentration measurements, the Bradford assay. This method was
based on the reaction between a dye (Brilliant Blue G) and basic amino acid residues in proteins and
required a dye to protein sample volumetric ratio of 30:1. This generated a coloured complex that
shifted the absorption maximum of the dye from 465 to 595 nm. For both methods the linear BSA
Author Version of paper:
Page 9 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
325
326
327
328
329
330
331
332
333
334
335
concentration ranges (Figure 3A) were small with the aniso-TSFS method being ~ 0.1–0.5 mg mL–1
and the Bradford assay 0.1–1.4 mg mL–1 (Table 1). To extend the application of both methods for
higher BSA concentrations we applied a semi-log model with a LinLog plot (Figure 3b). In both cases
we obtained excellent linear plots (r2>0.99) over the 0.1 to ~4 mg mL–1 concentration range. The
Bradford assay measurement accuracy was slightly better at BSA concentrations above 1.5 mg mL–1,
but was considerably lower at BSA concentrations below 1 mg/mL compared to aniso-TSFS (Figure
3c). For the aniso-TSFS method, a LOD of 13.8 µg mL–1 was calculated from the semi-log model.
Since 1 mg mL–1 was a typical protein concentration for cell culture media this made the aniso-TSFS
method an ideal rapid quantification tool, particularly since virtually no sample preparation was
required.
Method type
Photometrical
Examples
Direct quantification
at 280 nm [5]
Protein-dye
complexation
(Bradford, Lowry,
BCA) [5]
Standard
Fluorescence methods
[56]
SDS-PAGE [57-59]
Electrophoretical
Chromatographic
Affinity
chromatography,
HPLC-UV,
Size Exclusion
Chromatography [60,
61]
Comments
 requires pure protein samples.
 requires relatively high protein
conc.
 non-destructive
 Reagent addition required.
 Destructive.
LoD/linearity
5 - 100 μg/ml
(European
Pharmacopoeia 5.0 )
 Trace detection possible with
pure samples.
 Susceptible to interference
(other fluorophores, pH)
 Non-destructive after
purification.
 requires protein staining for
visualization.
 semi-quantitative
 requires extensive sample
preparation.
40 ng - 200 µg mL-1
(for Fluoroprofile
protein quantitation
kit)
0.1 - 1.4 mg mL-1
(Bradford assay)
250fmol BSA
(Coomassie R250
staining)
1 µg – 10 mg mL-1
(Protein A HPLC
method)
336
337
Table 1: Selection of common methods used for quantifying protein concentration.
338
339
340
341
342
343
344
345
4. Conclusions
The increased importance of albumin as a cell culture media supplement has now gained
importance [2] and thus for quality control purposes there is a need for a protein quantification method
that is fast, accurate, and requires no sample treatment. This study demonstrated the application of
MDF anisotropy measurements to the quantification of protein in complex aqueous mixtures. The
method is rapid, non-destructive, and requires no sample handling apart from pipetting the sample into
a cuvette. Operationally, the TSFS measurement approach was found to be superior to EEM because
Author Version of paper:
Page 10 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
it intrinsically eliminates the Rayleigh scattering artefacts and was somewhat faster. The aniso-TSFS
method was able to clearly discriminate fluorescence emission from the large protein molecule against
the complex background emission from small molecule fluorophores. The linear concentration range
for aniso-TSFS method is small ~ 0.1–0.5 mg mL–1, but the semi-log range was good over the 0.1–4
mg mL–1 which covers the expected concentration range for most cell culture media. In comparison to
competing analytical methods like the Bradford Assay (BA) and A280 absorption spectroscopy method,
the aniso-TSFS requires no sample preparation whereas BA requires the addition of reagents and A280
requires a significant amount of sample preparation to remove all the interfering components from the
media (Table 1). Furthermore, neither A280 nor BA methods generate any additional detailed
“collateral” information for variance analysis. The advantages of using MDF measurements rather
than single point or 1D spectral anisotropy measurements are two-fold. First, the aniso-TSFS
measurement provides data that can be used for the identification and variance analysis [12, 24] or
yield prediction [24], using chemometric analysis. Second, the method provides more in-depth
information about molecular composition in terms of fluorophore size and interactions between small
and large fluorophores. These two topics are the subject of on-going investigations in our laboratory
with a view to deliver a comprehensive MDF based analytical platform for cell culture media analysis.
Initially we had also intended to apply multi-way chemometric methods to generate a quantitative
model, but since this simpler approach delivers sufficient accuracy there was no need. However, the
TSFS data collected in the course of the measurement can be used for other tasks such as ID/variance
analysis [12, 24] or productivity prediction [23, 62] both of which have been demonstrated with the
closely related EEM measurement method.
Apart from the media analysis tasks described above, this aniso-TSFS method can and will be
applied to a variety of other protein based systems, for instance we envisage that one could use this for
the quantitative analysis of formulated protein therapeutics where intrinsically fluorescent excipients
are present. Another application of this method lies in the potential for assessing protein
conformational change in-situ without the need to purify the protein for analysis. These and several
other unique applications are currently being investigated in our laboratory.
5. Acknowledgements
Research was supported by funding from the Irish Research Council ‘EMBARK Initiative’
Postgraduate Scholarship to RCG and the loan of a Cary Eclipse fluorescence spectrometer by Agilent
Technologies.
6. Supplemental information available
Supporting information is available at http://dx.doi.org/10.1016/j.aca.2013.12.001.
References
[1]
C.F. Shen, T. Kiyota, B. Jardin, Y. Konishi, A. Kamen, Characterization of yeastolate
fractions that promote insect cell growth and recombinant protein production, Cytotechnology, 54
(2007) 25-34.
[2]
G.L. Francis, Albumin and mammalian cell culture: implications for biotechnology
applications, Cytotechnology, 62 (2010) 1-16.
[3]
V.P. Hanko, J.S. Rohrer, Determination of amino acids in cell culture and fermentation broth
media using anion-exchange chromatography with integrated pulsed amperometric detection,
Analytical biochemistry, 324 (2004) 29-38.
[4]
L.W. Dick, J.A. Kakaley, D. Mahon, D. Qiu, K.-C. Cheng, Investigation of proteins and
peptides from yeastolate and subsequent impurity testing of drug product, Biotechnol. Prog., 25 (2009)
570-577.
Author Version of paper:
Page 11 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
[5]
B.J.S.C. Olson, J. Markwell, Current Protocols in Protein Science, John Wiley & Sons, Inc.,
2001.
[6]
R.P. Kasprow, A.J. Lange, D.J. Kirwan, Correlation of Fermentation Yield with Yeast Extract
Composition as Characterized by Near-Infrared Spectroscopy, Biotechnology Progress, 14 (1998)
318-325.
[7]
A.G. Ryder, J.D. Vincentis, B. Li, P.W. Ryan, N.M.S. Sirimuthu, K.J. Leister, A stainless
steel multi-well plate (SS-MWP) for high-throughput Raman analysis of dilute solutions, Journal of
Raman Spectroscopy, 41 (2010) 1266-1275.
[8]
C. Dietzsch, O. Spadiut, C. Herwig, On-line multiple component analysis for efficient
quantitative bioprocess development, Journal of Biotechnology, 163 (2013) 362-370.
[9]
V. Vojinović, J.M.S. Cabral, L.P. Fonseca, Real-time bioprocess monitoring: Part I: In situ
sensors, Sensors and Actuators B: Chemical, 114 (2006) 1083-1091.
[10]
M. Navrátil, A. Norberg, L. Lembrén, C.-F. Mandenius, On-line multi-analyzer monitoring of
biomass, glucose and acetate for growth rate control of a< i> Vibrio cholerae</i> fed-batch
cultivation, Journal of Biotechnology, 115 (2005) 67-79.
[11]
P.M.L. Castro, A.P. Ison, P.M. Hayter, A.T. Bull, CHO cell growth and recombinant
interferon-gamma production: Effects of BSA, Pluronic and lipids, Cytotechnology, 19 (1996) 27-36.
[12]
B. Li, P.W. Ryan, M. Shanahan, K.J. Leister, A.G. Ryder, Fluorescence EEM Spectroscopy
for Rapid Identification and Quality Evaluation of Cell Culture Media Components., Appl. Spectrosc.,
65 (2011) 1240-1249.
[13]
C. Lindemann, S. Marose, H.O. Nielsen, T. Scheper, 2-Dimensional fluorescence
spectroscopy for on-line bioprocess monitoring, Sensors and Actuators B: Chemical, 51 (1998) 273277.
[14]
P.P. Mortensen, R. Bro, Real-time monitoring and chemical profiling of a cultivation process,
Chemometrics and Intelligent Laboratory Systems, 84 (2006) 106-113.
[15]
C. Calvet, B. Li, A.G. Ryder, Rapid quantification of tryptophan and tyrosine in chemically
defined cell culture media using fluorescence spectroscopy., J. Pharm. Biomed. Anal., 71 (2012) 8998.
[16]
T.L. Freeman, S.E. Cope, M.R. Stringer, J.E. CruseSawyer, D.N. Batchelder, S.B. Brown,
Raman spectroscopy for the determination of photosensitizer localization in cells, J. Raman
Spectrosc., 28 (1997) 641-643.
[17]
A. Andrade-Eiroa, M. Canle, V. Cerda, Environmental Applications of Excitation-Emission
Spectrofluorimetry: An In-Depth Review I, Applied Spectroscopy Reviews, 48 (2013) 1-49.
[18]
D. Patra, A.K. Mishra, Recent developments in multi-component synchronous fluorescence
scan analysis, Trac-Trends Anal. Chem., 21 (2002) 787-798.
[19]
A.G. Ryder, Assessing the maturity of crude petroleum oils using total synchronous
fluorescence scan spectra, Journal of Fluorescence, 14 (2004) 99-104.
[20]
A. Smilde, R. Bro, P. Geladi, Multi-way analysis with applications in the chemical sciences,
John Wiley & Sons, Inc., Chichester, 2004.
[21]
R. Bro, PARAFAC. Tutorial and applications, Chemometr. Intell. Lab. Syst., 38 (1997) 149171.
[22]
A. de Juan, R. Tauler, Multivariate curve resolution (MCR) from 2000: Progress in concepts
and applications, Crit. Rev. Anal. Chem., 36 (2006) 163-176.
[23]
P.W. Ryan, B. Li, M. Shanahan, K.J. Leister, A.G. Ryder, Prediction of Cell Culture Media
Performance Using Fluorescence Spectroscopy, Anal. Chem., 82 (2010) 1311-1317.
[24]
B. Li, N.M.S. Sirimuthu, B.H. Ray, A.G. Ryder, Using surface enhanced Raman scattering
(SERS) and fluorescence spectroscopy for screening yeast extracts, a complex component of cell
culture media., J. Raman Spectrosc., 43 (2012) 1074-1082.
[25]
D. Pappas, S.M. Burrows, R.D. Reif, Exploring biomolecular interactions by single-molecule
fluorescence, TrAC Trends in Analytical Chemistry, 26 (2007) 884-894.
[26]
M. Ameloot, M. vandeVen, A.U. Acuna, B. Valeur, Fluorescence anisotropy measurements in
solution: Methods and reference materials (IUPAC Technical Report), Pure and Applied Chemistry,
85 (2013) 589-608.
Author Version of paper:
Page 12 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
[27]
F. García Sánchez, A. Navas Díaz, M.M. López Guerrero, Simultaneous determination of
glycols based on fluorescence anisotropy, Analytica Chimica Acta, 582 (2007) 92-97.
[28]
H. Groenzin, O.C. Mullins, Molecular size and structure of asphaltenes, Petroleum Science
and Technology, 19 (2001) 219-230.
[29]
B.A. Mazzeo, D.D. Busath, From molecular dynamics to fluorescence anisotropy of
fluorophores bound to oriented structures, Journal of Computational Physics, 232 (2013) 482-497.
[30]
D.P. Millar, Time-resolved fluorescence spectroscopy, Curr Opin Struct Biol, 6 (1996) 637642.
[31]
J.C. Owicki, Fluorescence Polarization and Anisotropy in High Throughput Screening:
Perspectives and Primer, Journal of Biomolecular Screening, 5 (2000) 297-306.
[32]
S.Y. Wang, B.S. Ahn, R. Harris, S.K. Nordeen, D.J. Shapiro, Fluorescence anisotropy
microplate assay for analysis of steroid receptor-DNA interactions, Biotechniques, 37 (2004) 807-808.
[33]
C.M. Yengo, C.L. Berger, Fluorescence anisotropy and resonance energy transfer: powerful
tools for measuring real time protein dynamics in a physiological environment, Current Opinion in
Pharmacology, 10 (2010) 731-737.
[34]
J.M. Pallicer, S.D. Krämer, Evaluation of fluorescence anisotropy to assess drug–lipid
membrane partitioning, Journal of Pharmaceutical and Biomedical Analysis, 71 (2012) 219-227.
[35]
M.L. Sánchez-Martínez, M.P. Aguilar-Caballos, S.A. Eremin, A. Gómez-Hens, Longwavelength fluorescence polarization immunoassay for surfactant determination, Talanta, 72 (2007)
243-248.
[36]
M.D. Wenger, A.M. Bowman, M.V. Thorsteinsson, K.K. Little, L. Wang, J. Zhong, A.L. Lee,
P. DePhillips, An automated homogeneous method for quantifying polysorbate using fluorescence
polarization, Analytical Biochemistry, 337 (2005) 48-54.
[37]
V. Soubeyrand, V. Luparia, P. Williams, T. Doco, A. Vernhet, A. Ortiz-Julien, J.-M. Salmon,
Formation of Micella Containing Solubilized Sterols during Rehydration of Active Dry Yeasts
Improves Their Fermenting Capacity, Journal of Agricultural and Food Chemistry, 53 (2005) 80258032.
[38]
F. Baut, M. Fick, M.L. Viriot, J.C. André, J.M. Engasser, Investigation of acetone-butanolethanol fermentation by fluorescence, Appl Microbiol Biotechnol, 41 (1994) 551-555.
[39]
F.V. Bright, Fluorescence anisotropy selective technique (FAST): a new approach to
multicomponent fluorimetric analysis, Appl. Spectrosc., 42 (1988) 1245-1250.
[40]
K.A. Destrampe, G.M. Hieftje, New instrumentation for use in excitation-emission
fluorescence-polarization measurements, Appl. Spectrosc., 47 (1993) 1548-1554.
[41]
S. Nishiyama, M. Tajima, Y. Yoshida, Excitation emission matrix of spin-coated polymeric
thin films and its application to polarized fluorescence, J. Photopolym Sci. Technol., 19 (2006) 21-28.
[42]
Y. Moriyama, D. Ohta, K. Hachiya, Y. Mitsui, K. Takeda, Fluorescence behavior of
tryptophan residues of bovine and human serum albumins in ionic surfactant solutions: A comparative
study of the two and one tryptophan(s) of bovine and human albumins, J Protein Chem, 15 (1996)
265-272.
[43]
D.M. Togashi, A.G. Ryder, D. McMahon, P. Dunne, J. McManus, Progress in Biomedical
Optics and Imaging - Proceedings of SPIE, 2007, p. 66281K.
[44]
N. Tayeh, T. Rungassamy, J.R. Albani, Fluorescence spectral resolution of tryptophan
residues in bovine and human serum albumins, Journal of Pharmaceutical and Biomedical Analysis,
50 (2009) 107-116.
[45]
P. Bojarski, A. Jankowicz, Excitation energy transport between the ionic forms of rhodamine
B in viscous solutions, Journal of Luminescence, 81 (1999) 21-31.
[46]
A. Calvet, B. Li, A.G. Ryder, A rapid fluorescence based method for the quantitative analysis
of cell culture media photo-degradation, Anal. Chim. Acta, 807 (2014) 111-119.
[47]
J.R. Lakowicz, Principles of Fluorescence Spectroscopy, Springer, New York, 2006.
[48]
Y. Chen, M.D. Barkley, Toward understanding tryptophan fluorescence in proteins,
Biochemistry, 37 (1998) 9976-9982.
[49]
R.F. Chen, Fluorescence quantum yield measurements: vitamin B6 compounds, Science, 150
(1965) 1593-1595.
Author Version of paper:
Page 13 of 14
A Fluorescence Anisotropy Method for Protein Concentration Monitoring in Complex Cell Culture
Media. R.C. Groza, A. Calvet, and A.G. Ryder. Analytica Chimica Acta, 821, 54-61, (2014).
DOI: 10.1016/j.aca.2014.03.007
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
[50]
S. Marose, C. Lindemann, T. Scheper, Two-dimensional fluorescence spectroscopy: A new
tool for on-line bioprocess monitoring, Biotechnol. Prog., 14 (1998) 63-74.
[51]
D.M. Togashi, A.G. Ryder, A fluorescence analysis of ANS bound to bovine serum albumin:
Binding properties revisited by using energy transfer, Journal of Fluorescence, 18 (2008) 519-526.
[52]
B. Valeur, G. Weber, RESOLUTION OF FLUORESCENCE EXCITATION SPECTRUM OF
INDOLE INTO 1LA AND 1LH EXCITATION BANDS, Photochem. Photobiol., 25 (1977) 441-444.
[53]
S. Naveenraj, S. Anandan, Binding of serum albumins with bioactive substances Nanoparticles to drugs, J. Photochem. Photobiol. C-Photochem. Rev., 14 (2013) 53-71.
[54]
H.X. Zhang, X. Huang, M. Zhang, Spectral diagnostics of the interaction between pyridoxine
hydrochloride and bovine serum albumin in vitro, Mol. Biol. Rep., 35 (2008) 699-705.
[55]
F.L.G. Flecha, V. Levi, Determination of the molecular size of BSA by fluorescence
anisotropy, Biochemistry and Molecular Biology Education, 31 (2003) 319-322.
[56]
J.A. Mackintosh, D.A. Veal, P. Karuso, Fluoroprofile, a fluorescence-based assay for rapid
and sensitive quantitation of proteins in solution, PROTEOMICS, 5 (2005) 4673-4677.
[57]
C.Y. Liau, C.-S. Lin, A modified coomassie brilliant blue G 250 staining method for the
detection of chitinase activity and molecular weight after polyacrylamide gel electrophoresis, Journal
of Bioscience and Bioengineering, 106 (2008) 111-113.
[58]
C. Winkler, K. Denker, S. Wortelkamp, A. Sickmann, Silver- and Coomassie-staining
protocols: Detection limits and compatibility with ESI MS, ELECTROPHORESIS, 28 (2007) 20952099.
[59]
R.C. Krieg, Y. Dong, K. Schwamborn, R. Knuechel, Protein quantification and its tolerance
for different interfering reagents using the BCA-method with regard to 2D SDS PAGE, Journal of
Biochemical and Biophysical Methods, 65 (2005) 13-19.
[60]
J. Horak, A. Ronacher, W. Lindner, Quantification of immunoglobulin G and characterization
of process related impurities using coupled Protein A and size exclusion high performance liquid
chromatography, Journal of Chromatography A, 1217 (2010) 5092-5102.
[61]
E. Schleicher, O.H. Wieland, Specific quantitation by HPLC of protein (lysine) bound glucose
in human serum albumin and other glycosylated proteins, J Clin Chem Clin Biochem, 19 (1981) 8187.
[62]
B. Li, M. Shanahan, A. Calvet, K.J. Leister, A.G. Ryder, Comprehensive, quantitative
bioprocess productivity monitoring using fluorescence EEM spectroscopy and chemometrics, Analyst,
139 (2014) 1661-1671.
532
533
Author Version of paper:
Page 14 of 14