15.2.1 Circular Dichroism Spectroscopy

If we use polarised light in a spectroscopy experiment, we can enhance the information available for systems that interact differently with different polarisations. The most commonly used polarisation type is circularly polarised light whose electric and magnetic fields trace helices in space and time. For solutions where the molecules are randomly oriented, the only kind of molecule that interacts differently with left and with right circularly polarised light are chiral molecules – that is, molecules whose mirror images are not superposable on each other, e.g. [ 3–6]. The difference in absorbance of the two polarisations gives rise to the spectroscopy called CD, which is the difference in absorption of left and right circularly polarised light:

15.4

One can use CD quite effectively without understanding what it is, but it is helpful to have at least a pictorial understanding of the origin of the signal before attempting to interpret data. The key feature that gives rise to a non‐zero CD spectrum is that the electron redistribution that happens during the transition is helical. This is mathematically expressed by two vectors, one of which describes the linear direction of the electron density change, μ , which arises from coupling with the electric field of the light, and one of which describes the circling of charge about that direction, which we denote m , which arises from coupling with the magnetic field as discussed above. Simultaneous linear and circular rearrangements of the electrons gives us a helix. The Rosenfeld equation for CD magnitude is [ 3,7,8]

15.5

where Im denotes ‘imaginary part of’, since the magnetic dipole operation is imaginary. The sign of the CD depends on the handedness of the helix of electron redistribution – which has a 50% chance of being the same as the helix of the molecular bonding framework. The challenging task is to relate the helical motions of the electrons to the arrangement of the atoms and bonds in space.

CD is now a routine tool in many laboratories. The most common applications include proving that a chiral molecule has indeed been synthesised or resolved into pure enantiomers and probing the structure of biological macromolecules, in particular determining the secondary structure content of proteins. It is also useful for probing the binding of molecules to a chiral molecule, as the CD spectrum is perturbed by the interaction. In particular, if an achiral molecule binds to a chiral molecule, an induced CD signal will appear in the absorption bands of the achiral molecule. Most CD experiments involve randomly oriented samples. However, if the samples are oriented, even unintentionally, great care must be taken to ensure that one is not measuring LD rather than oriented CD due to instrumentation imperfections. If a sample is oriented with the unique axis along the light path, its CD will be independent of its rotation about that axis. However, any LD signal will invert when the sample is rotated 90°.

Most chiral molecules can be divided up into smaller subunits in which electronic transitions are largely located, called chromophores. For nucleic acids and proteins, the chromophores are approximately achiral. Thus, we can often understand the CD of a molecule as arising from the coupling of electronic transitions in different parts of the molecule. If we consider two monomers, the dimer system has two new transitions corresponding to collective charge displacements, described as the sum or difference of the two monomeric electric dipole transition moments to give two new ones [3]:

15.6

where the factor of ensures that the total absorption intensity is conserved. However, as a result of the + and − signs in Eq. 15.6, the absorption intensities of the two transitions may differ from each other. For example, if the monomeric transition moments are parallel, the plus combination will be a transition that contains all the absorption intensity of the two monomers while the minus combination will have zero absorption (the two anti‐parallel monomeric dipoles exactly cancelling each other). In the spectrum, this extreme case will look like a shift of the monomer absorption band to either shorter (H‐aggregate) or longer (J‐aggregate) wavelengths, depending on the changes to the transition energies. The coupling is called exciton coupling and the direction of shift depends on whether the geometry of the system means that dipole combination is, respectively, repulsive (destabilising) or attractive (stabilising) when one puts + and − signs on each end of the dipole.

If the monomer transitions in Eq. 15.6 are oriented parallel, anti‐parallel or orthogonal to each other, then they make an achiral system as the mirror image is the same as the original and we get no CD. However, any other angle gives a system where there is a net circulation of charge between the ground and excited states, which interacts with both the electric and magnetic fields of the light (Eq. 15.5). The mirror image system of the one we start with has the opposite handed helices and so for each transition will have a CD spectrum that is −1 times the original, as illustrated in Figure 15.3. If the two monomers are identical so their transitions are degenerate, the energies of the transitions of the coupled system is

15.7

where

15.8

for the vector from the origins of monomer A to that of B and we get an obvious +/− couplet in the CD spectrum due to overlap and cancellation. If the monomers are different the transition energies are very close to the uncoupled ones and we do not see the overlapping bands, but the CD signals at the two energies are opposite in sign. A real molecule has more than one transition and so what we observe is always a combination of many different interactions.

Figure 15.4 shows the absorbance (lower dashed line) and CD (upper solid line) spectrum for a chiral molecule where two anthracenes are held in a skewed arrangement. In the 254 nm region of the spectrum we see essentially a single exciton couplet from that intense transition. The CD is zero at the absorbance maximum where there is cancellation of the positive and negative components. In the 380 nm region of the spectrum we see evidence of the exciton coupling the vibronic components, but the multiple couplings of neighbouring components result in a less symmetric spectrum.

15.2.2 Linear Dichroism

When the electric field of the light oscillates in a plane, the result is linearly polarised light. The polarisation of the light is parallel to its electric field. From Eq. 15.2 we can see that oriented systems may interact differently with different polarisations of linearly polarised light and so may give an LD spectrum: the difference in absorption of light linearly polarised parallel (//) and perpendicular (⊥) to an orientation axis is, e.g. see [ 3 10–12],

15.9

Figure 15.5 illustrates many of the possibilities of LD spectroscopy. In this case anthracene has been deposited on a stretched oxidised polyethylene film (a high quality plastic bag held in a film stretcher). When the sample was prepared to ensure monomers dominated (0.1 mg/ml stock solution, black solid lines), we simply see a spectrum that tells us the polarisation of the transitions relative to the stretch direction. The 254 nm region is parallel to the stretch direction in accord with its known polarisation being along the long axis of anthracene. It follows the shape of the normal absorption spectrum. The 380 nm region is negative in accord with it being short‐axis polarised and, thus, perpendicular to the 254 nm region. The vibronic components follow the negative of the absorbance spectrum. At first sight the positive 325 nm signals are perplexing since they are part of the 380 nm vibronic progression. However, the closer the transitions get to the large 254 nm band the more it couples into the transition contributing long‐axis polarised intensity, so the high energy part of the progression has the same sign as the 254 nm band.

When the sample is deposited from an 8 mg/ml solution (pink dashed lines), it is no longer monomeric and we see exciton shifts of bands and the appearance of exciton couplets in the spectrum. Since the two new transitions from any exciton coupling are always polarised perpendicular to each other, in the simple film orientation system used for anthracene in Figure 15.5 they will almost certainly be of opposite sign. Anthracene forms a range of different oligomeric structures and different experiments can yield quite different spectra depending on how the sample dries on to the film [13].

In more complex orienting systems, such as dyes binding to flow‐oriented DNA, more complex options are possible. Quantitative analysis using the ratio of LD and absorbance, the reduced LD, is often helpful:

15.10

where α is the angle between the orientation direction and the transition dipole moment, S is the orientation parameter and the iso subscript is to remind us that this is strictly the absorbance of an isotropic sample, not an oriented sample. S is usually the ‘Achilles’ heel' of quantitative LD analysis as it typically varies from experiment to experiment and unless there is an internal standard (such as a DNA base signal when we are doing a ligand binding experiment, as discussed below) it can be challenging to estimate.

15.3.1 The Sample: Concentration, Pathlength and Matrix

Any absorption spectroscopy technique has a dynamic range limited by the need to have enough photons reaching the detector for us to be able to count and a signal above the background noise level. What this means varies from instrument to instrument and as a function of wavelength. It also depends not only on the sample but also its matrix.

For biomolecules one cannot design a spectroscopy experiment without considering the buffer and any associated salts. For example, for a backbone protein CD experiment on a bench top (i.e. not synchrotron) instrument we usually aim to collect data from 260 nm to at least 190 nm and ideally a bit lower in wavelength. Many buffer components, including chloride ions, sugars and amino acids as well as the more obvious buffer components such as the sulphite of HEPES and MOPS, absorb light in this region, thus reducing the effective lamp intensity for the sample. Likewise, sugars and amino acids are chiral themselves, so also contribute to the CD spectrum. Chloride ions in ‘standard’ PBS (phosphate buffered saline) vary from laboratory to laboratory, but in all cases are more than one wants to use in a CD experiment as most of the photons will be absorbed by chloride by 215 nm. So step (i) for any spectroscopy experiment is to measure an absorbance spectrum of the buffer, solvent or other matrix. This can usually be done in the CD instrument if that option is checked in the parameter menu.

Let us assume that we have found a largely invisible buffer (phosphate and TRIS without extra chloride are a good start for backbone protein CD), we then must consider how much sample is needed. A rough rule‐of‐thumb is that 0.1 mg/ml in a 1 mm pathlength cuvette for backbone protein spectroscopy is ideal. Typical spectra for predominantly helical myoglobin (solid black line), β‐sheet rich concanavalin (dashed pink line) and a random coil polymer (N‐formyl aspartic acid, green dotted line) measured at 0.1 mg/ml in water are illustrated in Figure 15.6.

So what is the significance of 0.1 mg/ml and a 1 mm pathlength? The average amino acid weight in a protein is between 105 and 110 Da or g/mol, so 0.1 mg/ml is approximately 1 mM residue concentration. Below 250 nm, most of the absorbance is due to the amino bonds of the protein backbone, with only a little coming from any side chains. Since any protein is a mixture of amino acids, it follows that the concentration of the amides is the key. The protein molar concentration is not relevant as different sized proteins will have different numbers of amide bonds. Following Eq. 15.3, a 1 mM residue concentration in a 1 mm pathlength gives an absorbance of approximately 1, which is optimal for instrument performance. If one wishes to use a higher concentration then, following Eq. 15.3, one must reduce the pathlength such that the total absorbance of the sample does not exceed approximately 2. In practice, the extent to which one can do this depends on the operator's ability to assemble demountable cells reproducibly below 0.1 mm (the smallest available fixed pathlength quartz cuvette). It should also be remembered that any non‐protein species that absorbs reduces the photons available to the protein – hence the discussion about buffers above. If buffer components are chiral, they not only absorb photons but also give a CD signal and so baseline correction must be very carefully done.

It should be noted that we tend not to use the backbone region of the protein absorbance spectrum to determine concentration. This is partly because of the potential contribution from buffer components and light scattering (LS) but also because there is no maximum just gradually increasing signal as one decreases the wavelength below about 250 nm. However, in some cases the protein backbone absorbance spectrum can be useful for estimating protein concentration and typical extinction coefficients are available in reference [6]. By way of contrast, the so‐called aromatic region, which is dominated by tyrosines and tryptophans along with a small contribution from disulphide bonds, can be used to estimate protein concentration (typically at 280 nm). The CD signal in the aromatic region can also provide a fingerprint to compare proteins. However, in this case care must be taken to choose an appropriate combination of protein concentration and pathlength based on the protein being analysed. Unless the percentage of aromatic residues and disulphide bonds is known, then it is really just a sophisticated guess to say a 1 mg/ml sample in a 1 cm cuvette has an absorbance of 1. For proteins with known sequences, the ExPASy ProtParam website provides a useful tool to estimate the 280 nm extinction coefficient from approximate extinction coefficients for residues. These 280 nm protein extinction coefficients typically range from approximately 10 000–200 000 mol⁻¹ dm³ cm⁻¹. Strictly speaking, unfolded proteins should be used to determine the protein concentration using these extinction coefficients.

Similar to the situation with the backbone region of proteins, we can identify ideal concentrations for DNA spectroscopy since DNAs are composed of the same four components. For DNA we usually expressed concentration in terms of the concentration of bases (or base pairs). A random sequence genomic DNA has an extinction coefficient of about 6600 mol⁻¹ dm³ cm⁻¹ per DNA base at 258 nm so an ideal concentration for a UV absorbance or CD experiment is approximately: 150 μM in a 1 cm path length cuvette for DNA spectroscopy as this gives a 258 nm absorbance of about one, which optimises instrument performance. While this could be expressed as 0.05 mg/ml, for DNA sample preparation it is better to avoid assuming one's sample only contains DNA and to use the Beer–Lambert law to determine the concentration. The choice of a 1 cm pathlength for DNA samples is due to the empirical observation that the Beer–Lambert law (Eq. 15.3) tends to break down when the base concentration goes above 200 μM. The DNA CD spectrum for 100% alternating G‐C DNA with increasing concentrations of spermine is given in Figure 15.7. As the concentration of the highly cationic spermine is increased, the DNA converts from standard B‐form to Z‐form with approximate inversion of the CD spectrum. The inversion is only approximate as, although the helix changes handedness, the final structure is not the mirror image of the original.

For other samples, some idea of the extinction coefficient is needed to determine the optimal concentration and pathlength combination for an experiment. As a general rule, UV–visible spectroscopy extinction coefficients range up to 20 000 mol⁻¹ dm³ cm⁻¹. By way of contrast, IR samples are seldom greater than 100 mol⁻¹ dm³ cm⁻¹. Unfortunately for IR spectroscopy of proteins and DNAs, water has a significant absorbance in the same region and omitting it from the sample cannot be assumed to have no effect on the biomacromolecule structure. We return to this issue below.

15.3.2 The Instrument

15.3.2.3 Linear Dichroism

LD experiments require data to be collected with two perpendicular linear polarisations. Although one could simply rotate a polariser or rotate the sample with respect to a fixed polariser, adapting a CD instrument to produce linearly polarised light invariably produces better‐quality spectra. The simplest way to do this is to insert a quarter‐wave plate to convert alternating circularly to alternating linearly polarised light. However, most electronic CD instruments now come with a software option to increase the voltage on the photoelastic modulator and the user needs to do nothing except remember to change the software setting.

Another essential feature of an LD experiment is sample orientation. Although there are many ways to achieve this, the most common way to orientate small molecules is to absorb them on to, or into, a polymer film either before or after the film is stretched. For biomacromolecular systems of high aspect ratio, Couette flow is more commonly used to orientate the long axis of the sample with the direction of flow. A film stretcher and a microvolume Couette flow cell are illustrated in Figure 15.8. A recent innovation has simplified LD for polar small molecules. Previously the best data were collected with polyvinyl alcohol films, which required the analyte to be included in a polymerising film and the sample to be left for a few days to dry before stretching under gentle heat [3]. By way of contrast, polyethylene (in the form of commercial plastic bags) could be used for hydrophobic analytes. We recently discovered that placing polyethylene in an oxidising plasma asher for a few seconds made the surface sufficiently hydrophilic for charged and polar molecules to bind and align with their long axes along the polymer stretch direction, but had no effect on the polymer absorption spectrum in the UV–visible region. Most modern LD instrument define // (Eq. 15.9) to be horizontal.

15.3.3 Structure Fitting of Proteins from Spectral Data

Sometimes one uses spectroscopy simply to determine whether two samples are structurally the same. This can be done by visually comparing the shape of the two spectra. However, for many purposes it is helpful to reduce the complexity of visual comparison of spectra by extracting an alternate simpler information set. Using protein backbone CD spectra (including data from 250 to 190 nm or lower wavelength) is a well‐established method to estimate the average secondary structure of a protein, based on the CD spectra of a dataset of proteins with a known secondary structure content. However, as the CD signal is proportional to protein concentration, the accuracy of these secondary structure fitting methods is most affected by the accuracy of the knowledge of the concentration of the sample. If the methods [16–23] that are collated on, for example, the Dichroweb platform [24] give similar answers then one can believe them. However, if they differ and particularly if Selcon [17] refuses to give an answer, then one needs to question the data quality that is being used. We developed a neural network approach, SSNN, that with good data gives similar answers to the other methods with the same reference set, has an accompanying scaling option that can be used to estimate the real concentration of the sample and then give a secondary structure estimate [22, 23,25]. Of course, each stage of data manipulation or approximation introduces the potential for error.

Baseline‐corrected protein IR transmission spectra also depend on the secondary structure of proteins. The protein amide I band occurring between 1700 and 1600 cm⁻¹ is fairly independent of the ions in the buffer and is widely accepted to reflect the secondary structure [ 15,26]. Extracting secondary structure content of proteins from IR data is typically done by band fitting with the relative areas of fitted bands at 1645–1660 cm⁻¹ attributed to α‐helix, at 1620–1640 and 1670–1695 cm⁻¹ to the β‐sheet, at 1620–1640 and 1650–1695 cm⁻¹ to turns and at 1640–1657 and 1660–1670 cm⁻¹ to other structures [27]. In our hands, this approach can work reasonably well if the spectrum has a good signal‐to‐noise ratio, the baseline subtraction is perfect and the spectrum has no unusual features. However, our fits do vary noticeably. We have therefore been working to apply our SSNN [ 22, 23] neural network approach for CD to IR data and have generalised the method, now calling it SOMSpec (self‐organising map spectroscopy). It generally seems to be more effective than the band‐fitting approach, though it depends on the quality of the reference set used. The spectral NRMSD (normalised root mean squared deviation), defined as

15.11

where x _i is the value at each wavenumber, N is the number of data points, M is the largest intensity and m is the smallest, so (M–m) is the range that gives a numerical measure of the goodness of spectral fit that is equally weighted across the spectrum. A more limited range focusing on the peak maximum region may be more useful. SOMSpec is available from reference [28].

15.4.1 Comparison of Different α‐Lactalbumin Protein Structures

Historically, it was assumed that the amino acid sequence of a protein defines a single, unique native structure, with at most small local rearrangements to accompany function. However, we now understand that the process of protein folding is far more complex – many proteins exist in a dynamic ensemble of different structures and/or adopt intermediate folding states or form different structures under different environmental conditions. Hence, both the sequence and structure of a protein are important considerations in comparing the biosimilarity of different protein preparations for applications such as biopharmaceuticals. In contrast to traditional protein structure determination methods, such as X‐ray crystallography, spectroscopic techniques provide several distinct advantages for characterising and comparing different protein structures, including the ability to work in different solvent environments and look at multiple different structures simultaneously and in real time. Different spectroscopic methods can also be useful in providing complementary structural information.

An illustrative example of the application of CD and related spectroscopy to look at protein structure is α‐lactalbumin (α‐LA). α‐LA is a relatively small protein from milk that has been extensively studied as a model of protein folding. α‐LA has a single strong Ca²⁺ binding site. The crystal structures of the Ca²⁺ bound holo α‐LA reveals a compact globular structure consisting of an α‐helical domain and a small β‐sheet domain that flank the Ca²⁺ binding cleft (Figure 15.9a) [29]. Several other α‐LA structures have also been identified. These include: the Ca²⁺ minus apo‐α‐LA form [30], molten globules formed at low pH (pH 3–5) or elevated temperature, and partially folded intermediates formed in moderate guanidine hydrochloride (GuHCl) concentrations [31]. The therapeutic potential of a partially unfolded α‐LA form stabilised by integration of an oleic acid co‐factor (also known as a HAMLET complex) is also currently being explored as a tumour suppressor [32].

As discussed above, CD in the protein backbone region (260–190 nm and lower) is commonly used to determine the secondary structure content of proteins. For example, the CD spectra of holo bovine α‐LA (in 50 mM sodium phosphate pH 7 + 2 mM CaCl₂, solid black line, Figure 15.8b) shows the negative bands at 222 and 208 nm and the positive band at 193 nm, characteristic of primarily α‐helical proteins.

In this case, the CD spectra were also collected in a series of buffer conditions designed to favour alternate structures of α‐LA – including the apo (50 mM sodium phosphate pH 7 + 1 mM ethylenediaminetetraacetic acid [EDTA]), molten globule pH 5 (50 mM sodium phosphate pH 5 + 2 mM CaCl₂), molten globule pH 3 (50 mM sodium phosphate pH 3 + 2 mM CaCl₂), partially folded (50 mM sodium phosphate pH 7 + 1 M GuHCl) and unfolded (50 mM sodium phosphate pH 7 + 4 M GuHCl) states. Each sample was prepared at ∼0.1 mg/ml and the CD spectra were measured in a 1 mm pathlength cuvette. To account for any error in concentration, the absorbance of each sample was measured at 280 nm and the concentration was estimated using the Beer–Lambert Law (Eq. 15.3) and an extinction coefficient of 28 460 M⁻¹ cm⁻¹, obtained by providing amino acid sequence of the bovine α‐LA to the ProtParam tool available on EXPASY [33]. These concentrations (c) were then used to convert the CD in units of mdeg (as output from the instrument) to molar ellipticity (θ):

15.12

where MW is the molecular weight (14 154 g/mol for α‐LA), ℓ is the pathlength in cm (here 0.1 cm), C is the molar concentration and r is the number of amino acids (123 for α‐LA), to enable direct visual comparison of the different spectra (Figure 15.9b). Molar ellipticity, which is widely used in the literature, differs from the Beer–Lambert extinction coefficient Δε (Eq. 15.4) by a factor of 3298. The unit of θ is deg cm²/dmol, which is a historical legacy. Using the above‐mentioned units in Eq. 15.12, a factor of 10 is needed in the denominator for the correct conversion.

Comparison of the α‐LA protein backbone CD spectra reveals some clear differences in the secondary structure of the different α‐LA structures. The spectra of the holo and apo structures almost completely overlay, while the molten globule pH 5 and partially folded spectra follow a very similar shape, suggesting that the native secondary structure is also maintained in these states. The shape of the molten globule pH 3 spectra is significantly different from that of the apo spectra, with a decrease in the CD signal in the region of 222 nm and a shift of the 208 nm negative maximum towards 200 nm, corresponds to a loss of secondary structure, while the unfolded spectrum (4 M GuHCl) more closely resembles that of a random coil presumably with a negative maximum at 200 nm. Qualitative comparison of the spectral shapes and amplitudes can point out the fact that there are structural differences between α‐LA samples in different buffers. However, the real question, what the structural changes are and their extent, can be answered reliably only by analysing the spectra for the secondary structure contents with the methods presented above.

In this case, a 50 mM sodium phosphate buffer was chosen to buffer the samples as it is largely invisible in CD. However, to form the different α‐LA structures, inclusion of some highly absorbent species was required, including GuHCl (in the case of the partially unfolded and unfolded samples), HCl (in the case of the pH 3 sample) and EDTA (in the case of the apo samples). Thus, these spectra have been cropped at lower wavelengths. Where absorbance is high (i.e. transmission is low) the high tension (HT) voltage that is applied to amplify the detector's sensitivity is also high. Since different wavelengths have different light energies, the HT accommodates for fluctuations in the light levels by changing the gain. When the HT increases above a threshold of ∼600 V there are not enough photons being sampled to measure a reliable or valid CD signal and these data points should not be used. While this typically negates performing secondary structure fitting for such samples, useful structural information can still be obtained, especially when comparing the CD spectra of different protein sample as in this example.

CD in the aromatic region can also provide useful information about protein structure. Figure 15.9c shows the CD spectra for the same α‐LA samples as in Figure 15.9b, but this time measured at concentrations of ∼0.4 mg/ml in a 1 cm pathlength cuvette. Again, the holo and apo spectra overlay. The loss of the CD signal around 280 nm in the partially unfolded and, to a greater extent, in the molten globule pH 5 spectra are consistent with a loss of tertiary structure, which removes the well‐defined chiral environment from around the aromatic residues. In accord with this explanation, the pH 3 spectra in the aromatic region more closely resembles that of the unfolded state than of a folded one. However, from the backbone CD we know the secondary structure of this sample is not unfolded. Hence, taken together, the backbone and aromatic CD of the pH 7 holo sample and pH 5 and pH 3 samples are consistent with a classic molten globule transition in which a native‐like secondary structure is maintained but a defined tertiary structure is progressively lost as the pH is lowered.

One may then ask: how can we distinguish between the holo and the apo α‐LA structures as there is no discernible difference in either the protein backbone or aromatic region CD for these two samples? Absorbance in the aromatic region also results in fluorescence emission predominantly derived from the tryptophan (Trp) residues (for further discussion, see [34]). The wavelength of maximum emission is highly sensitive to the solvation and rotational restriction of the Trp residues and can thus be used as a site‐specific probe of protein folding. In folded proteins, where the Trp residues are buried in the protein structure, fluorescence emission occurs at a lower wavelength (e.g. the α‐LA holo sample, 336 nm), while for unfolded proteins, in which the Trp residues are maximally solvated, maximum fluorescence emission occurs at longer wavelengths (e.g. the α‐LA unfolded state, 349 nm, see Figure 15.9d). The maximum emission of the partially unfolded and molten globule α‐LA structures are all intermediate between those of the folded and unfolded states. Of particular note, the maximum fluorescence emission of the apo state (342 nm) is significantly higher than that of the holo state. This is the result of a change in the local environment of two Trp residues (Trp 26 and Trp 104) located in an aromatic cluster in the binding cleft of bovine α‐LA, as absence of the stabilising Ca²⁺ ion in the apo form allows the binding cleft to adopt a more ‘open’ conformation [30]. It should be noted that we have chosen to plot the fluorescence data normalised to 1 to emphasise the wavelength shifts. Fluorescence magnitudes do vary with environment, but the influences are much more complex and can be misleading.

The ability to monitor protein structural changes in real time, for example during thermal unfolding, can also be particularly informative when multiple spectroscopic methods are used in combination. Figure 15.10 shows the melting curves obtained for holo α‐LA using three different spectroscopic methods: (i) the change in protein backbone CD signal monitored at 222 nm; (ii) the change in aromatic CD signal monitored at 280 nm; and (iii) the shift in maximum Trp fluorescence emission wavelength (plotted as the ratio of emission at 350 nm to that at 330 nm). Initial destabilisation of the α‐LA binding site cleft at slightly elevated temperatures (20–45 °C) does not result in significant change to the overall secondary or tertiary structure and is therefore not discernible in the CD spectra. However, this initial destabilisation is visible in the fluorescence melting curve. At higher temperatures (above approximately 55 °C), the 280 nm CD melting curve mirrors that of the fluorescence curve due to loss of tertiary structure as α‐LA transitions through the molten globule state, while loss of secondary structure with increasing temperature is far more gradual, as can be seen in the 222 nm CD melting curve.

Fourier transform infra‐red (FTIR) can also be a useful spectroscopic method to characterise and compare different protein structures. Original spectra and baseline corrected spectra are shown in Figure 15.11. The Amide I band is centred at about 1650 cm⁻¹ and is generally deemed to correlate with secondary structure. The 1550 cm⁻¹ Amide II band seems to be influenced by the salts in the solution. As discussed above, a significant drawback of FTIR is that the protein signal overlaps with that of water (for example, see Figure 15.11a, where the black protein in water spectrum overlays almost exactly the dotted red buffer spectrum). While subtraction of appropriate baseline spectra can be used to eliminate the water FTIR signal, at the protein concentration chosen for these experiments (which is lower than what is usually used) this can make these data hard to interpret. We can see the effect of water vapour most obviously in the Apo spectrum of Figure 15.11 (blue solid line). The FTIR‐ATR data are consistent with little secondary structure change between the three samples considered (the wavenumber maxima of the Amide I bands are similar). However, data quality is poor at such a low concentrations where water dominates the original spectra.

High protein concentrations (50–200 mg/ml) are typically used for protein FTIR spectra. However, these concentrations are above the solubility limit of a lot of proteins and thus not always a feasible option. One possible solution is to work in D₂O, which removes the issue of the water background. However, biopharmaceutical products are not formulated in D₂O. An alternative approach would be to measure the FTIR‐ATR of solid protein samples. This is an attractive solution for comparing protein therapeutic preparations, which are often produced as lyophilised proteins. However, methods to extrapolate secondary structure or compare similarities between different protein structures have not yet been developed. We are currently working on establishing methods to use FTIR‐ATR spectroscopy on aqueous or solid protein samples to generate secondary structure estimates accounting for refractive index and light intensity variations with wavenumber. This requires first transforming ATR data into the equivalent transmission form [15] and using a self‐organising structure fitting approach. [22].

15.4.2 Linear Dichroism Examples

15.4.2.1 Flow Oriented DNA and Ligand Binding

When long pieces of DNA are flow oriented in a Couette flow cell there is a preferential alignment of the sample. At 1000 rpm in a microvolume cell, the orientation parameter S = 0.1. Figure 15.12 illustrates the type of spectrum one expects: zero LD until about 300 nm, where the DNA bases start absorbing, and then a negative LD signal that mirrors the shape of the DNA absorbance spectrum. Its negative sign follows from the fact that the transitions are all in the plane of the aromatic bases of DNA and these are oriented approximately perpendicular to the helix axis at an average angle of about 86°. When the groove binder Hoechst 33258 is added to the solution, the spectrum gains additional signals from the Hoechst molecules that bind in the minor groove of the DNA. The 370 nm transition of Hoechst is polarised along its long axis [35], which is therefore oriented at about 45° from the helix axis, so less than 54.7° makes the LD (Eqs. 15.9 and 15.10) positive. Groove binding ligands do not significantly stiffen or lengthen DNA, so the fact that the 250 nm LD in the presence of Hoechst is more negative than the DNA signal indicates that the Hoechst transition in this region is approximately short‐axis polarised so is more than 54.7°.

15.4.2.2 M‐13 Bacteriophage

Bacteriophage are well‐defined assemblies of proteins and nucleic acids. In the case of M13, it is a filamentous structure with DNA in the middle and a very limited number of proteins forming the capsid structure. M13 bacteriophage are about 800 nm in length and have a very a large persistence length (>1 μM) and so orient extremely well, giving an enormous protein LD signal in the backbone region below 250 nm [36–38]. The spectra shown in Figure 15.13 have a negative signal at 280 nm, which faintly shows the three bands of tryptophans sometimes seen in the aromatic region of protein CD spectra. The 250 nm region is due to the DNA whose bases are oriented perpendicular to the long axis of the phage. The backbone region of the spectrum has a 188 nm negative signal and 205 nm positive signal, which are dominated by the two components of the α‐helix π–π* transition. The 222 nm n–π* region has a similar but much less intense couplet (positive maximum and a minimum that would be negative were it not overlaid on the tail of the large positive 205 nm signal), which we speculate arises from the intensity that the electronically forbidden transition ‘borrows’ from the intense allowed neighbouring π–π* transition.

References

1. 1 Zimmermann, T., Rietdorf, J., and Pepperkok, R. (2003). Spectral imaging and its applications in live cell microscopy. FEBS Lett. 546 (1): 87–92.
2. 2 Lasch, P. and Naumann, D. (2006). Spatial resolution in infrared microspectroscopic imaging of tissues. Biochim. Biophys. Acta Biomembr. 1758 (7): 814–829.
3. 3 Nordén, B., Rodger, A., and Dafforn, T.R. (2010). Linear Dichroism and Circular Dichroism: A Textbook on Polarized Spectroscopy, 304. Cambridge: Royal Society of Chemistry.
4. 4 Berova, N., Nakanishi, K., and Woody, R.W. (eds.) (2000). Circular Dichroism Principles and Applications. New York: Wiley‐VCH.
5. 5 Woody, R.W. (2009). Circular dichroism spectrum of peptides in the poly(Pro)II conformation. J. Am. Chem. Soc. 131: 8234–8245.
6. 6 Kelly, S.M., Jess, T.J., and Price, N.C. (2005). How to study proteins by circular dichroism. Biochim. Biophys. Acta 1751: 119–139.
7. 7 Rosenfeld, L. (1928). Quantenmechanische Theorie der natürlichen optischen Aktivität von Flüssigkeiten und Gasen. Z. Phys. 52: 161–174.
8. 8 Schipper, P.E. and Rodger, A. (1983). Symmetry rules for the determination of the intercalation geometry of host/guest systems using circular dichroism: a symmetry adapted coupled‐oscillator model. J. Am. Chem. Soc. 105: 4541–4550.
9. 9 Harada, N., arada, N., Takuma, Y., and Uda, H. (1976). The absolute stereochemistries of 6,15‐dihydro‐6,15‐ethanonaphtho[2.3‐c]pentaphene and related homologs as determined by both exciton chirality and X‐ray Bijvoet methods. J. Am. Chem. Soc. 98: 5408–5409.
10. 10 Nordén, B. (1977). General aspects on linear dichroism spectroscopy and its application. Spectrosc. Lett. 10: 381–400.
11. 11 Halsall, D.J., Dafforn, T.R., Marrington, R. et al. (2004). A linear dichroism technique for quantitative PCR applications. IVD Technol. 6: 51–60.
12. 12 Marrington, R., Dafforn, T.R., Halsall, D.J. et al. (2005). Validation of new microvolume Couette flow linear dichroism cells. Analyst 130: 1608–1616.
13. 13 Razmkhah, K., Gibson, M.I., Chmel, N.P., and Rodger, A. (2014). Oxidized polyethylene films for orienting polar molecules for linear dichroism spectroscopy. Analyst 139: 1372–1382.
14. 14 Rodger, A., Sanders, K.J., Hannon, M.J. et al. (2000). DNA structure control by polycationic species: polyamines, cobalt ammines, and di‐metallo transition metal chelates. Chirality 12: 221–236.
15. 15 Pinto‐Corujo, M., Sklepari, M., Ang, D. et al. (2018). Infra‐red absorbance spectroscopy of aqueous proteins: comparison of transmission and ATR data collection and analysis for secondary structure fitting. Chirality .
16. 16 Woody, R.W. (1994). Circular dichroism of peptides and proteins. In: Circular Dichroism Principles and Applications (ed. K. Nakanishi, N. Berova and R.W. Woody). New York: VCH.
17. 17 Sreerama, N. and Woody, R.W. (2000). Estimation of protein secondary structure from circular dichroism spectra: comparison of CONTIN, SELCON, and CDSSTR methods with an expanded reference set. Anal. Biochem. 287: 252–260.
18. 18 Sreerama, N. and Woody, R.W. (1993). A self‐consistent method for the analysis of protein secondary structure from circular dichroism. Anal. Biochem. 209: 32–44.
19. 19 Nakanishi, K., Berova, N., and Woody, R.W. (1994). Circular Dichroism: Principles and Applications. New York: VCH.
20. 20 Johnson, W.C.J. (1988). Secondary structure of proteins through circular dichroism spectroscopy. Ann. Rev. Biophys. Biophys. Chem. 17: 145–166.
21. 21 Johnson, W.C. (1999). Analyzing protein circular dichroism spectra for accurate secondary structures. Proteins Struct. Funct. Genet. 35: 307–312.
22. 22 Hall V, Sklepari M, Rodger A (2014) Protein secondary structure prediction from circular dichroism spectra using a self‐organizing map with concentration correction. Chirality 26: 471–482. doi:10.1002/chir.22338.
23. 23 Hall, V., Nash, A., and Rodger, A. (2014). SSNN, a method for neural network protein secondary structure fitting using circular dichroism data. Anal. Methods 6 (17): 6721–6726.
24. 24 Whitmore, L. and Wallace, B.A. (2004). DICHROWEB: an online server for protein secondary structure analyses from circular dichroism specroscopic data. Nucleic Acids Res. 32: W668–W673.
25. 25 Hall, V., Nash, A., Hines, E., and Rodger, A. (2013). Elucidating protein secondary structure with circular dichroism and a neural network. J. Comput. Chem. 34: 2774–2786.
26. 26 Corujo MP, Praveen A, Steel MJ, Ang D, Chmel N, Rodger A. Attenuated Total Reflectance Infra Red Absorbance Spectroscopy for Proteins in H₂O: Electromagnetic Fields at Boundaries. Submitted 2018.
27. 27 Haris, P.I. (2013). Infrared spectroscopy of protein structure. In: Encyclopedia of Biophysics: European Biophysical Societies' Association (ed. G.K. Roberts).
28. 28 Ang D, Dukor R, Pinto‐Corujo M, Reason A, Rodger A. SOMSpec: a general purpose neural network‐based tool for rapid protein secondary structure prediction from circular dichroism and infra‐red absorbance spectra. In preparation 2019.
29. 29 Acharya, K.R., Stuart, D.I., Walker, N.P.C. et al. (1989). Refined structure of baboon α‐lactalbumin at 1.7 Å resolution: comparison with C‐type lysozyme. J. Mol. Biol. 208 (1): 99–127.
30. 30 Chrysina, E.D., Brew, K., and Acharya, K.R. (2000). Crystal structures of apo‐ and holo‐bovine α‐lactalbumin at 2.2‐Å resolution reveal an effect of calcium on inter‐lobe interactions. J. Biol. Chem. 275 (47): 37021–37029.
31. 31 Dolgikh, D.A., Gilmanshin, R.I., Brazhnikov, E.V. et al. (1981). α‐Lactalbumin: compact state with fluctuating tertiary structure? FEBS Lett. 136 (2): 311–315.
32. 32 Ho, J.C.S., Nadeem, A., and Svanborg, C. (2017). HAMLET – a protein‐lipid complex with broad tumoricidal activity. Biochem. Biophys. Res. Commun. 482 (3): 454–458.
33. 33 Gasteiger, E., Hoogland, C., Gattiker, A. et al. (2005). Protein identification and analysis tools on the ExPASy server. In: The Proteomics Protocols Handbook (ed. J.M. Walker). Humana Press.
34. 34 Lakowicz, J.R. (2006). Principles of Fluorescence Spectroscopy. Springer‐Verlag.
35. 35 Moon, J.‐H., Kim, S.K., Sehlstedt, U. et al. (1996). DNA structural features responsible for sequence dependent binding geometries of Hoechst 33258. Biopolymers 38: 593–606.
36. 36 Clack, B.A. and Gray, D.M. (1992). Flow linear dichroism spectra of four filamentous bacteriophages: DNA and coat protein contributions. Biopolymers 32: 795–810.
37. 37 Pacheco‐Gomez, R., Roper, D.I., Dafforn, T.R., and Rodger, A. (2011). The pH dependence of polymerization and bundling by the essential bacterial cytoskeltal protein FtsZ. PLoS One 6 (6): e19369. https://doi.org/10.1371/journal.pone.0019369.
38. 38 Pacheco‐Gomez, R., Kraemer, J., Stokoe, S. et al. (2012). Detection of pathogenic bacteria using a homogeneous immunoassay based on shear alignment of virus particles and linear dichroism. Anal. Chem. 84 (1): 91–97.

Further Reading

1. Berova, N., Nakanishi, K., and Woody, R.W. (eds.) (2000). Circular Dichroism Principles and Applications. New York: Wiley‐VCH.
2. Gasteiger, E., Hoogland, C., Gattiker, A. et al. (2005). Protein identification and analysis tools on the ExPASy server. In: The Proteomics Protocols Handbook (ed. J.M. Walker). Humana Press.
3. Lakowicz, J.R. (2006). Principles of Fluorescence Spectroscopy. Springer‐Verlag.
4. Nakanishi, K., Berova, N., and Woody, R. (eds.) (1994). Circular Dichroism: Principles and Applications. New York: VCH.
5. Nordén, B., Rodger, A., and Dafforn, T.R. (2010). Linear Dichroism and Circular Dichroism: A Textbook on Polarized Spectroscopy. Cambridge: Royal Society of Chemistry.
6. Wallace, B.A. and Janes, R. (eds.) (2009). Modern Techniques for Circular Dichroism Spectroscopy. Amsterdam: IOS Press.