Dynapro Data Interpretation

DynaPro Data Interpretation Guide Understanding Results Obtained from the DynaPro Light Scattering Instrument Operated in Batch Mode An overview of Size Distributions, Autocorrelation Functions, and Molecular Weight Estimates Quickly and Easily determine the ‘goodness’ of results by comparing actual data to typical data.

December 15, 2002

Proterion Corporation. ©

1

Table of Contents •

Part I: Introduction – – –

•

Part II: Size Distributions – – – – – – –

•

Introduction Glossary of common terms Interpreting a Dynamic Light Scattering Measurement Size Distribution Results Monomodal size distribution Multimodal size distribution Polydispersity Size Distribution Interpretations Hydrodynamic radius: the physical interpretation of ‘size’ The physical interpretations of size distributions

Part III: Goodness of Data – – –

Good or Bad: Judging the goodness of the data Summary of Operation The Autocorrelation Functions • • • •

–

•

Summary of Goodness of Data

Part IV: Molecular Weight Estimates – – – –

•

Sample vs. Solvent Large particles, Large Fluctuations Large particles, Multimodal Size Distribution Weak Signals

Molecular Weight Estimates Molecular Weight Interpolated from Radius Interpreting the BSA Standard The many size distributions of the BSA Standard

Part V: Beyond the Single Data Point – The Experiment December 15, 2002

Proterion Corporation. ©

2

Part I: Introduction • Introduction • Glossary of common terms • Interpreting a Dynamic Light Scattering Measurement

December 15, 2002

Proterion Corporation. ©

3

Introduction • The purpose of the DynaPro Data Interpretation manual is to provide a basic understanding of the results obtained from your DynaPro Light Scattering instrument. • The manual will define the common terms associated with data interpretation, provide a framework for determining if the data are ‘good’ or ‘bad’, and explain typical results. • The manual will not cover the basic theories of light scattering, however it will provide you with references which do cover the theories in detail.

December 15, 2002

Proterion Corporation. ©

4

Glossary: Common Terms •

The DynaPro determines Size distributions of particles in solution. Size Distributions are defined by several terms. – Size: refers to the radius or diameter of the particle modeled as a sphere that moves or diffuses in the solution (in contrast to the Molecular Weight of the particle). Usually expressed as the mean value of the peak of the size distribution. –

Size Distribution: the manner in which the sizes of the particles are dispersed or spread or allocated among one or more peaks; presented in a graphical form known as a Histogram.

–

Peak: A Peak in a Size Distribution represents a distinct and resolvable species or population of analytes or particles. A Peak is comprised of several size particles, represented by bins or bars, and is defined by a mean (average) value and polydispersity.

–

Modality: refers to the number of ‘peaks’ in the size distribution. A size distribution with one peak is called Monomodal. A size distribution with more than one peak is called Multimodal (Bimodal, Trimodal are common terms for size distributions with 2 or 3 peaks).

–

Mean value: the mean value of the peak is the weighted average of the various size particles (bins or bars) in the distinct or resolvable population. The various sizes are weighted by their probability of being detected.

–

Polydispersity: the standard deviation of the histogram which refers to the width of the peak. Sometimes referred to as the percent polydispersity (polydispersity divided by the mean value), it is a measure of heterogeneity or homogeneity of the species comprising the population.

–

Bin: a discrete numerical particle size component of the Histogram or Size Distribution which is defined by an x-axis value in nanometers (size), and a y-axis value in relative amount of light scattered by the bin to the other bins. The number of bins, the value or particle size represented by the bin, and relative amount of scattered light are determined by numerical algorithms associated with the analysis of the raw data from the DynaPro. The bins do not reflect actual, physical particles.

December 15, 2002

Proterion Corporation. ©

5

Interpreting a Measurement What is a measurement? –

The DynaPro defines a measurement as a collection of acquisitions for a particular sample. An acquisition is a period of time, typically 10 seconds, during which the light scattered by the sample is averaged and correlated. We recommend a measurement last 100 seconds (10 acquisitions, 10 seconds each or 20 acquisitions, 5 seconds each). – The result of a measurement contains N number of acquisitions, which are averaged and presented in a number of ways. – Generally speaking we want to see the ‘size distribution’ of the sample: analyte in solution. Information of the distribution of the sizes of the analyte is applied to protein crystallization, protein based drug development, drug delivery nanoparticle development, nanoparticle characterization and many other areas of advanced materials characterization. The ‘tree’ lists the

measurements and Their acquisitions. December 15, 2002

Proterion Corporation. ©

The size distribution is data of interest.

6

Part II: Size Distributions • • • • • • •

Size Distribution Results Monomodal size distribution Multimodal size distribution Polydispersity Size Distribution Interpretations Hydrodynamic radius: the physical interpretation of ‘size’ The physical interpretations of size distributions

December 15, 2002

Proterion Corporation. ©

7

Size Distribution Results Results are shown in graphical as well as tabular form. The table located below the size distribution histogram describes the number of peaks and their mean value (Radius), polydispersity (Pd), % polydispersity (%Pd), molecular weight estimated from the measured radius (MWR), relative amount of light scattered by each population (%Int), and estimated relative amount of mass (concentration) of each peak or species (%Mass).

December 15, 2002

Proterion Corporation. ©

8

Monomodal Size Distribution Histogram This histogram has one peak so we call it a monomodal size distribution. The peak is defined by the mean value and polydispersity.

Y-axis Relative amount of light scattered by each bin, % Intensity (% of Total Light Scattered). Represents the probability of existence of the species.

The ‘width’ of the peak is the standard deviation of the weighted bin values, also known as the Polydispersity.

The mean value of the peak is defined by a weighted average of the number of bins comprising the histogram, in this case three. The bins by themselves do not represent real, distinct, physical particles however their mean and standard deviation do. December 15, 2002

Proterion Corporation. ©

X-axis Discrete particle sizes, in nanometers

9

Multimodal Size Distribution What causes Modality? The presence of different and resolvable species in the sample cause modes in the size distribution. To be resolved as a separate peak, a species must have a size (radius) larger than another species by a factor of two or more, and be detectable (produce sufficient scattered light for detection by the DynaPro). Roughly speaking a factor of two in radius is equivalent to a factor of eight (octamer) in MW. When the sizes of the species are below this factor, a separate peak will not be resolved for each species.

This histogram has more than one peak so we call it a multimodal size distribution. Specifically this histogram is trimodal. The DynaPro determined three distinct populations exist in this sample.

By definition, a multimodal size distribution is heterogeneous: the sample contains distinct populations of particles that are not the same size. The DynaPro can resolve up to four or five modes in a size distribution. For each mode, the DynaPro estimates the relative amount of light scattered and the relative amount of mass based upon one of several possible particle scattering properties. Often times the relative amount of mass of a peak is quite small e.g. less than .1 % and is considered to be negligible. December 15, 2002

Proterion Corporation. ©

10

Polydispersity

Each peak has a unique mean value and width or Polydispersity. It is useful to normalize Polydispersity to the mean size of the peak, also known as percent polydispersity.

Polydispersity refers to the level of homogeneity of the sizes of the particles. When the level of homogeneity is high, the particles can be considered to be virtually identical in their size, or monodisperse. The level of homogeneity is considered high when the percent polydispersity is less than 15%. When the level of homogeneity is low (percent polydispersity greater than 30%), the particle population can be considered to contain significantly different sizes, or ‘polydisperse’.

What causes Polydispersity? Heterogeneity is caused by the presence of different species that can not be resolved by the technique of dynamic light scattering (species with sizes less than a factor of two relative to other species exist in solution can not be resolved). A peak containing 100% monomer will have a smaller polydispersity than peak containing a mixture of monomer:octamer. The peaks shown here all have % Polydispersity greater than 30%. Note: refer to appendix for an alternate cause of polydispersity. December 15, 2002

Proterion Corporation. ©

11

Size Distribution Interpretations Monomodal Monodisperse

Monomodal ‘Polydisperse’

Multimodal Polydisperse

BLGA, 4 mg/ml, PBS, T = 25 C Peaks: 1 Mean Radius: 2.8 nm % Poly: 13.8 % Majority monomer

BLGA, 4 mg/ml, PBS, T = 5 C Peaks: 1 Mean Radius: 3.4 nm % Poly: 22.1 % Increasing amounts of Dimer

BSA, 2 mg/ml, PBS, T = 25 C Peaks: 2 Peak 1: • Mean Radius: 4.3 nm • % Poly: 32.1 % • Monomer, Dimer, Trimer Peak 2: • Mean Radius: 130.9 • % Poly: 34.5 % • Various non-specific aggregates

December 15, 2002

Proterion Corporation. ©

12

Hydrodynamic Radius: The Physical Interpretation of ‘Size’ The DynaPro measures the size distribution of the particles in the sample. The size, previously defined as the radius or diameter of the particle, is represented in this figure as Rh. Rh, or Hydrodynamic Radius, is a particle radius that embodies a ‘hard sphere’ particle which is in fact aspherical and typically surrounded or covered by solvent. Please refer to PSI Books or the article for a more thorough explanation of the physical interpretation of particle size as measured by the DynaPro.

December 15, 2002

Proterion Corporation. ©

13

Physical Interpretations of Size Distributions Monomodal Monodisperse Radius = R1

Monomodal Polydisperse

Radius = 1.5*R1

The samples contains two types of particles, monomer and trimer. The radius of the trimer is less than twice the radius of monomer so only one peak is resolved, the distribution is monomodal. However the population consists of two species and this increase in size heterogeneity causes an increase in measured polydispersity compared to the samples containing 100% monomer and 100% trimer. Also, the mean radius of the peak will be larger than R1 but smaller than 1.5 R1.

Radius = 5*R1

December 15, 2002

Bimodal Monodisperse

The sample contains two types of particles, the monomer and a large aggregate. The large particle is more than twice the radius of the monomer and in sufficient quantities to be measured, so two peaks are resolved by the DynaPro. The mean radius of Peak 1 will be R1 and Peak 2 will be equal to 5*R1. Both species are homogeneous so measured polydispersity is low.

Proterion Corporation. ©

Bimodal Monodisperse Polydisperse

The sample contains three types of particles monomer, trimer, and larger aggregate. In this case the DynaPro resolves only two peaks. The monomer and trimer are not resolved from each other and form only one peak, a polydisperse peak. The second peak is formed by the larger particle, which is resolvable from both monomer and trimer. The second peak is monodisperse. 14

Part III: Goodness of Data • Good or Bad: Judging the goodness of the data • Summary of Operation • The Autocorrelation Functions – – – –

Sample vs. Solvent Large particles, Large Fluctuations Large particles, Multimodal Size Distribution Weak Signals

• Summary of Goodness of Data

December 15, 2002

Proterion Corporation. ©

15

Good or Bad: Judging the Goodness of the Data •

At this point we have covered the basic interpretations of DynaPro results, describing size distributions and understanding the meanings of mean size, modality, polydispersity.

•

However, how do we determine if the results are acceptable or unacceptable, good or bad?

•

The DynaPro software, Dynamics, does provide basic data analyses that indicate if the data are in ‘acceptable’ ranges. The analyses are based upon simple numerical data filters or qualifiers. Yet these data filters do not always capture or allow for good and bad raw data.

•

We will explain what is good or bad by commenting on various examples of raw data. The name applied to the raw data is an autocorrelation function. An autocorrelation function is a collection of correlation coefficients – unitless values indicating the level of similarity among sets of data. At this time we will not concern ourselves with the underlying theory and physical meaning of the autocorrelation function. We will examine only how to interpret these functions. December 15, 2002

Proterion Corporation. ©

16

Summary of Operation Intensity

Particles moving in solution, illuminated by a laser, create time intensity fluctuations (on the order of microseconds).

Time

The rate of time intensity fluctuations are determined by Autocorrelation, resulting in an Autocorrelation Function

Numerical methods determine the rates of decay in the Autocorrelation Function, which are related to the particle sizes. The result of the analysis of the Autocorrelation Function is a Size Distribution. December 15, 2002

Proterion Corporation. ©

17

The Autocorrelation Function The DynaPro determines the size of particles in solution by exploiting the physical process of Brownian Motion: the particles are moving in solution as a function of time, and their rate of motion is related to their size. The rate of motion is measured by illuminating the particles with laser light and determining the rate at which light scattered or reflected by the particles changes with time. The technique of autocorrelation determines the rate of these time intensity fluctuations, expressed as an autocorrelation function (shown here). An autocorrelation function is an exponential function comprised of correlation coefficients (yaxis) dependent upon the ‘delay time’ (x-axis), the time-value separating the sets of data. The function can be mathematically described by one or more decays. The rate of decay is related to particle size. A faster decay indicates a smaller particle, a slower decay a larger particle. Autocorrelation functions are determined during each acquisition comprising a measurement, as described earlier. December 15, 2002

Numerical algorithms are applied to determine the rates of decay or size distributions of the exponential autocorrelation functions. The DynaPro utilizes a proprietary ‘non-negative least squares’ algorithm, a method that finds the size distribution producing the smoothest distribution with the least amount of error. The error is the difference between the measured autocorrelation function and the fitted autocorrelation function.

Proterion Corporation. ©

18

Autocorrelation functions: Sample vs. Solvent Not all samples can be measured, nor are all samples properly suited for measurement by the DynaPro, and therefore not all samples produce valid autocorrelation functions. Without a valid autocorrelation function, it is not possible to determine a valid size distribution. A valid autocorrelation function is generally smooth and continuous, exponentially decaying from a maximum value of 2 to a value of 1. To the right is shown a valid autocorrelation function. By eye we observe one decay in the function.

The function contains random values centered around 1, asymptotically reaching 1. Randomness represents the result from measuring pure solvent: solution containing zero analyte or analyte below the limits of detection. The size distribution analysis will attempt to find a result for these functions. These must be marked and removed from the analysis. It is generally a good idea to measure the solvent to confirm its purity. If one unexpectedly sees a function characteristic of solvent increase laser power, measure the sample unfiltered (to avoid potential binding to the membrane) or unspun, and/or increase the concentration of the analyte.

December 15, 2002

Proterion Corporation. ©

19

Autocorrelation Functions: Large particles, Large fluctuations If during the measurement of an autocorrelation function the total intensity scattered by the population of particles changes rapidly or spikes, the detector and/or correlator can saturate, resulting in a discontinuity as shown here. These functions must be marked and removed from the size distribution analysis. The situation can be remedied by removing bubbles, spinning or filtering the sample, or changing solvent conditions to remove large aggregates or particles. Alternatively, the particles may be larger than the size range of the instrument (several microns in radius).

The decay of this function has not been fully captured, it is prematurely terminated. This is caused by have an acquisition time too short relative to the long decay of the autocorrelation function. Generally a larger particle size requires a longer acquisition time. The size distribution analysis can be performed however there will be greater error in the results. The additional autocorrelation coefficients can be captured by extending the acquisition time of the measurement. Note: increasing the number of acquisitions will not capture additional coefficients in the longer time delays.

December 15, 2002

Proterion Corporation. ©

20

Autocorrelation Functions: Large Particles, Multimodal populations The autocorrelation function shown here contains two visually observable decays. One is faster, representing a smaller particle and the other is slower, representing a larger particle. These functions are valid and can be analyzed.

The autocorrelation function associated with larger particles has a longer decay, as shown here. Note the yvalue of the function has asymptotically reached a value of 1 yet the function has some variation at the larger time delays. The variation is referred to as ripple or noise. The noise is due to insufficient numbers of correlation coefficients being collected and calculated. The noise can be reduced by collecting additional numbers of acquisitions. With less noise, the size distribution analysis will be of higher quality.

December 15, 2002

Proterion Corporation. ©

21

Autocorrelation Functions: Weak signal Ripple or a lack of smoothness of the function in the short time delay area indicates a weaker signal from the particles. These functions can be fitted however the polydispersity may be greater due to this ‘noise’. The remedy for this situation is to either extend the acquisition time, collect more acquisitions, increase laser power, and/or increase analyte concentration. Refer to PSI Books for a discussion of the advantages and disadvantages of increasing acquisition time and the number of acquisitions.

December 15, 2002

Proterion Corporation. ©

22

Summary of Goodness of Data Proceed

December 15, 2002

Caution

Stop

Increase acquisition time

Spin or filter sample

Increase acquisition time Increase acquisitions Increase laser power Increase concentration

If not pure solvent… Increase acquisition time Increase acquisitions Increase laser power Increase concentration

Proterion Corporation. ©

23

Part IV: Molecular Weight Estimates • • • •

Molecular Weight Estimates Molecular Weight Interpolated from Radius Interpreting the BSA Standard The many size distributions of the BSA Standard

December 15, 2002

Proterion Corporation. ©

24

Molecular Weight Estimates •

The Molecular Weight of a biological molecule can be estimated from the measurement of the size or hydrodynamic radius. The estimate is based upon on an empirical curve of known proteins and measured hydrodynamic radius.

•

The error of the estimated Molecular Weight from Hydrodynamic Radius ranges from several percent to over 100%. The wide range of error is due to the nature of the estimate. Not all proteins fall on the curve. The estimated value must be used with caution.

•

When applying the Molecular Weight estimate, make sure the intensity weighted size distribution analysis is selected. The empirical curves are based upon the use of the intensity weighted calculation of the mean of the peak.

•

Also, if the peak is determined to be polydisperse by the DynaPro size distribution analysis, then the mean radius is a weighted average of more than one species. The estimated molecular weight will be a weighted estimate based upon the weighted average size.

•

The molecular weight estimate can be qualified by examining the shape factor, the relationship between the measured hydrodynamic radius and the ‘hard sphere radius’ calculated from the known molecular weight and density of the protein. Please refer to PSI books for a review of the concepts of the shape factor or axial ratio. December 15, 2002

Proterion Corporation. ©

25

Molecular Weight Interpolated from Radius MW = 14

Rh = 1.9 MW-R is the Molecular Weight Interpolated from the Measured Hydrodynamic Radius. Ideally the size distribution is monodisperse, otherwise the measured radius is a weighted average of more than one species, and the estimated MW – even for a protein or other particles that falls on the empirical curve – will be in error. Select the model that best fits the a priori knowledge of the sample. Or, match the model that best matches the known molecular weight or oligomer to obtain an estimate on the shape or conformation of the sample.

December 15, 2002

Proterion Corporation. ©

26

Interpreting the BSA Standard MW ~ 130 kDa MW = 66kDa

Rh = 3.5 nm

Rh ~ 4.5 nm

The DynaPro is provided with an ampoule containing 2 mg/ml of BSA prepared in a PBS solution. Often the sample is measured and the molecular weight results are higher than the expected value for monomeric BSA (Rh = 3. nm and MW = 66 kDa), sometimes as much as a factor of two larger. The reason for the difference is that the BSA ampoule contains monomer, dimer, trimer, and large non-specific aggregates. The majority peak of the size distribution typically comprises the specific aggregates, and the minority peak (low % mass peak) typically comprises large non-specific aggregates. Depending upon the relative amounts of the specific aggregates, the mean value of the majority peak can range Rh = 3.6 nm (virtually 100%) monomer to 4.5 nm or more (dimer and trimer), with the large amounts of polydispersity.

December 15, 2002

Proterion Corporation. ©

27

The many size distributions of the BSA standard Monomodal Monodisperse Monomer

Bimodal Peak 1: Monodisperse Monomer Peak 2: Monodisperse Aggregate December 15, 2002

Monomodal Polydisperse Monomer, Dimer, Trimer

Bimodal Peak 1: Polydisperse M:D:T Peak 2: Polydisperse Aggregate Proterion Corporation. ©

Multimodal Peak 1: Polydisperse M:D:T Peak 2: Polydisperse Aggregate 28

Part V: Beyond the Single Data Point Design, Perform, and Interpret an Experiment The power of the DynaPro.

BLGA as a function of Temperature. By determining the size distribution over a range of experimental conditions the DynaPro determines that the mean size, total intensity, and polydispersity of BLGA increase with decreasing temperature due to the formation of specific aggregates. Refer to all of our application notes to learn how to capitalize on the DynaPro. www.protein-solutions.com December 15, 2002

Proterion Corporation. ©

29