Analytica Chimica Acta, accepted, 07/07/2015. This is the accepted version without proofing
corrections. DOI: 10.1016/j.aca.2015.06.011
.
Page 4 of 26
studies (10–70 °C, at 10 °C increments), a 1 mg mL
–1
HSA solution was used. For HSA
chemical unfolding at 25 °C, appropriate amounts were used to prepare 1 mg mL
–1
HSA
solutions with 0–4 M GuHCl concentrations (0.5 M steps). For the room temperature
phosphorescence (RTP) validation, 5 M KI (0.18 mL) and 2 M Na 2 SO 3 (0.36 mL) were
added to 3 mL (1 mg mL
–1
) HSA solution to induce the “heavy atom” effect and chemical
deoxygenation respectively, and then aniso-TSFS measurements made at 10, 20, and 30 °C
[5].
2.2 Fluorescence instrumentation and data collection: Spectra were measured in triplicate
from solutions in 1×1 cm quartz cuvettes using an Eclipse Fluorescence Spectrophotometer
(Agilent) fitted with a manual polarizer accessory and a temperature-regulated single cell
cuvette holder (Quantum Northwest). TSFS data were collected over an excitation range of
λ ex = 270–370 nm and Δλ interval of 20–150 nm (2 nm step increments in each case) with 10
nm excitation/emission slit widths (unless otherwise stated) [41]. For each sample, four
differently polarized spectra were collected: vertical-vertical (VV), vertical-horizontal (VH),
horizontal-vertical (HV) and horizontal-horizontal (HH) which took ~30 minutes per sample.
Anisotropy (r) at each emission wavelength was calculated using the standard anisotropy
formula [1], which was then used to construct aniso-TSFS plots. Accurate calculations
required that a reasonable intensity be recorded for each of the four polarization spectra [41].
When spectral intensity dropped below the 10% threshold, anisotropy values generally
became unreliable with abnormally high or negative r values (vide infra).
Chemometric methods: Data analysis was performed using MATLAB (ver. 7.0.1),
PLS_Toolbox4.0
®
, and in-house written codes. All data were organised with Δλ (mode 1),
λ ex (mode 2), and the sample as mode 3. Datasets had dimensions of 66×51×24 (thermal) and
66×46×27 (chemical) respectively. Spectral variance, and number of factors, were first
assessed using NPFPCA (noise perturbation in functional principal component analysis)
which is a variant form of the functional PCA better suited to noisy data [44, 45]. NFPCA
makes use of systematic noise effects and the addition of synthetic noise to produce more
reliable results. It then implemented PCA, which relies just on the existing random effect in
the data to identify representative eigenvectors. Multi-way decomposition was performed
using MCR alternating least squares (ALS) where models were constructed on the augmented
matrix datasets with Δλ as the ‘spectral’ mode and the ‘λ ex ×sample’ as the ‘concentration’
mode. Mode 1 loadings represent the ‘delta’ profiles while mode 2 loadings represent
component excitation profiles/spectra.
3. Results and Discussion
3.1 Spectroscopic Analysis: ARMES first required the accumulation of four separate
polarization datasets (HH, VV, VH and HV) and the first thing of note was that the polarized
emission intensity of HSA was approximately an order of magnitude lower than standard
TSFS measurements (Supplemental information, Fig. S-2) which caused a significant
reduction in signal-to-noise ratio (SNR). HH and VV spectra were more intense compared to
HV and VH due to the more rigid fluorophore environment. HH spectra were always more
intense than the VV, because of the better transmittance of the filters in this orientation (see
Figure S-3). For HSA, strong emission from Trp and Tyr were observed over λ ex =280–320
nm and Δλ=20–140 nm ranges (corresponding to λ ex =280–320 nm, λ em =300–460 nm EEM
corrections. DOI: 10.1016/j.aca.2015.06.011
.
Page 4 of 26
studies (10–70 °C, at 10 °C increments), a 1 mg mL
–1
HSA solution was used. For HSA
chemical unfolding at 25 °C, appropriate amounts were used to prepare 1 mg mL
–1
HSA
solutions with 0–4 M GuHCl concentrations (0.5 M steps). For the room temperature
phosphorescence (RTP) validation, 5 M KI (0.18 mL) and 2 M Na 2 SO 3 (0.36 mL) were
added to 3 mL (1 mg mL
–1
) HSA solution to induce the “heavy atom” effect and chemical
deoxygenation respectively, and then aniso-TSFS measurements made at 10, 20, and 30 °C
[5].
2.2 Fluorescence instrumentation and data collection: Spectra were measured in triplicate
from solutions in 1×1 cm quartz cuvettes using an Eclipse Fluorescence Spectrophotometer
(Agilent) fitted with a manual polarizer accessory and a temperature-regulated single cell
cuvette holder (Quantum Northwest). TSFS data were collected over an excitation range of
λ ex = 270–370 nm and Δλ interval of 20–150 nm (2 nm step increments in each case) with 10
nm excitation/emission slit widths (unless otherwise stated) [41]. For each sample, four
differently polarized spectra were collected: vertical-vertical (VV), vertical-horizontal (VH),
horizontal-vertical (HV) and horizontal-horizontal (HH) which took ~30 minutes per sample.
Anisotropy (r) at each emission wavelength was calculated using the standard anisotropy
formula [1], which was then used to construct aniso-TSFS plots. Accurate calculations
required that a reasonable intensity be recorded for each of the four polarization spectra [41].
When spectral intensity dropped below the 10% threshold, anisotropy values generally
became unreliable with abnormally high or negative r values (vide infra).
Chemometric methods: Data analysis was performed using MATLAB (ver. 7.0.1),
PLS_Toolbox4.0
®
, and in-house written codes. All data were organised with Δλ (mode 1),
λ ex (mode 2), and the sample as mode 3. Datasets had dimensions of 66×51×24 (thermal) and
66×46×27 (chemical) respectively. Spectral variance, and number of factors, were first
assessed using NPFPCA (noise perturbation in functional principal component analysis)
which is a variant form of the functional PCA better suited to noisy data [44, 45]. NFPCA
makes use of systematic noise effects and the addition of synthetic noise to produce more
reliable results. It then implemented PCA, which relies just on the existing random effect in
the data to identify representative eigenvectors. Multi-way decomposition was performed
using MCR alternating least squares (ALS) where models were constructed on the augmented
matrix datasets with Δλ as the ‘spectral’ mode and the ‘λ ex ×sample’ as the ‘concentration’
mode. Mode 1 loadings represent the ‘delta’ profiles while mode 2 loadings represent
component excitation profiles/spectra.
3. Results and Discussion
3.1 Spectroscopic Analysis: ARMES first required the accumulation of four separate
polarization datasets (HH, VV, VH and HV) and the first thing of note was that the polarized
emission intensity of HSA was approximately an order of magnitude lower than standard
TSFS measurements (Supplemental information, Fig. S-2) which caused a significant
reduction in signal-to-noise ratio (SNR). HH and VV spectra were more intense compared to
HV and VH due to the more rigid fluorophore environment. HH spectra were always more
intense than the VV, because of the better transmittance of the filters in this orientation (see
Figure S-3). For HSA, strong emission from Trp and Tyr were observed over λ ex =280–320
nm and Δλ=20–140 nm ranges (corresponding to λ ex =280–320 nm, λ em =300–460 nm EEM