1y ago

168 Views

2 Downloads

1.29 MB

50 Pages

Transcription

UPTEC X 20016Examensarbete 30 hpJuni 2020Characterizing the pore structureof porous matrices using SEQ-NMRspectroscopyElla Strömberg

AbstractCharacterizing the pore structure of porous matricesusing SEQ-NMR spectroscopyElla StrömbergTeknisk- naturvetenskaplig orietLägerhyddsvägen 1Hus 4, Plan 0Postadress:Box 536751 21 UppsalaTelefon:018 – 471 30 03Telefax:018 – 471 30 ization of the pore structure is a crucial part in themanufacturing of porous media used for purification of biologicalpharmaceuticals. This project took place at Cytiva in Uppsala and aimedat optimizing a newly developed method in pore structurecharacterization called size-exclusion quantification NMR (SEQ-NMR). Bymeasuring with diffusion NMR on a polymer solution before and afterequilibration with a material of interest the pore structure of thematerial can be determined. This project aimed at reducing the durationof a SEQ-NMR experiment while examining the performance of the methodduring different conditions with the goal of making the methodapplicable for quality control procedures. The method was optimizedboth by simulations and by experimental diffusion NMR measurements. Itwas discovered that the performance of the method could be improved byhaving an optimal mixture of the polymer solution and duringexperiments distributing ten measurement points with linear spacing.With these parameters optimized the duration of the method could bereduced with 22 hours landing on a total duration of 8 hours. Theduration combined with the complexity of the method still makes themethod unsuitable for use in quality control of porous media. Despitethe small possibility of SEQ-NMR being a quality control method thisproject has proven the method to be both reproducible and sensitive.Handledare: Fredrik ElwingerÄmnesgranskare: Katarina EdwardsExaminator: Erik HolmqvistISSN: 1401-2138, UPTEC X 20016

Populärvetenskaplig sammanfattningKanske hade du precis som jag en klosslåda när du var liten. I de olika hålen på lådan fickklossarna plats om de hade rätt form och storlek. För ett företag som ska tillverka och sälja enklosslåda krävs väldigt precis kunskap om vilka mått, vilken struktur, hålen i lådan har. Det härprojektet fokuserar kring optimeringen av en metod för att kunna bestämma just storleken pålådans håligheter där själva lådan egentligen är porösa geler som Cytiva i Uppsala tillverkar.Klossarna som passar eller inte passar i lådan motsvarar biologiska läkemedel som renas frammed hjälp av den porösa gelen. Metoden i projektet heter size-exclusion quantification nuclearmagnetic resonance (SEQ-NMR) och bygger på mätningar av en lösnings koncentration innanoch efter den varit i kontakt med en porös gel. Om en samling av klossar i olika storlekar ochformer hälls över klosslådan så kommer vissa klossar gå ner i lådan medan andra hamnarutanför. Genom att mäta klosshögens koncentration, hur många av varje storlek och form detfinns i förhållande till hela högen, och jämföra den med koncentrationen av klossar somhamnade utanför lådan kan strukturen på lådans ihåligheter bestämmas.Koncentrationen mäts indirekt genom diffusions-NMR där molekylers rörelse mäts med hjälpav ett magnetfält och applicering av magnetiska pulser. Pulserna appliceras i par och om enmolekyl förflyttar sig, diffunderar, mellan pulserna kommer det synas som en försvagning avden signal som fås av mätningen. Ju starkare den applicerade magnetiska pulsen är desto störreblir försvagningen i signal. Den försvagade signalen följer en avtagande kurva och kanöversättas till en koefficient som karaktäriserar diffusionen av den molekylen. Då pulsstyrkanär noll är signalen direkt proportionell, lika med, koncentrationen av molekylen i lösningen.Det här kan sedan matematiskt översättas till vilken porstorlek den undersökta porösa gelenhar.Det är viktigt att strukturen på de porösa material som Cytiva producerar är karakteriserade påett korrekt sätt för att framreningen av biologiska läkemedel som sedan distribueras till patienterhåller hög kvalité. För att uppfylla efterfrågan på biologiska läkemedel krävs det också attefterfrågan på porösa material produceras effektivt där hög noggrannhet upprätthålls ochproduktionen sker på ett reproducerbart sätt. Alla geler som produceras testas därför för attkontrollera att de upprätthåller den kvalité som krävs för produktion av läkemedel. Debiologiska läkemedlen kan sorteras på olika egenskaper och för geler som separerar medavseende på storlek är det porstorleken hos gelen som specificerar produkten. Porstorlekenmotsvarar alltså hålen i klosslådan och genom att veta dess storlek kan produkten specificerasför vilka storlekar på läkemedelsprodukter den kan rena fram.Den metod Cytiva använder idag för att bestämma porstrukturen tar cirka 15 timmar vilket inteär optimalt i en process där man vill producera stora mängder gel. Lösningen som är i kontaktmed gelen vid SEQ-NMR består av dextran, en stor grenad molekyl som kan ha olika storlekar.iii

En aspekt av optimeringen var att hitta den perfekta blandningen av dextranstorlekar,fördelningen av klossarna i samlingens storlekar. Vid starten av detta projekt tog SEQ-NMR15 timmar per diffusionsmätning vilket ger en total tid på 30 timmar då mätningar ska göras pålösningen både före och efter jämvikt. När den optimala blandningen hittats genom simuleringoch mätningsmetoden optimerats hade totala experimenttiden förkortats till 8 timmar. Detmotsvarar en förbättring hos utförandet av metoden men den perfekta blandningen av storlekarpå dextran visade även från simuleringar att samma samling klossar inte fungerar för attbestämma storleken på hålen hos samtliga lådor. Cytiva producerar en mängd olika porösa gelerbara för storleksseparation och att behöva en optimal storleksfördelning av dextran för varje gelgör metoden svårare att tillämpa inom till exempel kvalitetskontroll vilket var det tänktaanvändningsområdet för metoden.iv

Table of Contents1Introduction .32Background .42.1NMR .42.2Diffusion NMR .42.2.1Pulsed field gradient stimulated echo .52.2.2Data processing .62.3Sensitivity and SNR.62.4Polydispersity and dextrans .62.5SEQ-NMR .72.5.1Selectivity curve .72.5.2Attenuation curves .82.5.3Evaluation of fitted models .82.62.6.1Convective flow .92.6.2Non-uniform gradient pulses.92.6.3Eddy currents. 102.73Porous matrices . 10Materials and methods . 123.1Simulations . 123.1.1Optimal mixture of dextrans . 133.1.2The number of gradient points and their distribution. 143.1.3The robustness of method . 143.24Experimental errors in diffusion NMR.9Experiments . 153.2.1Evaluation of diffusion NMR measurements . 153.2.2The robustness of method . 153.2.3Sample preparation and the optimal experiment . 16Results . 174.1Simulated results. 174.1.1Optimal mixture of dextrans . 174.1.2The number of gradient points and their distribution. 194.1.3The robustness of method . 214.2Experimental results . 224.2.1Evaluation of diffusion NMR measurements . 234.2.2The robustness of method . 254.2.3Optimal parameters. 28v

5Discussion . 286Conclusion . 327Acknowledgements . 32References . 33Appendix A. 35Appendix B. 42vi

��SEQ-NMRSECSNRVaccesibleVporeVtotVvoidNMR signal attenuationdiffusion coefficientpulse durationdiffusion timegradient strengthinverse size exclusion chromatographydistribution coefficientmolecular weightdistributionpulsed field gradient stimulated echocoefficient of determinationhydrodynamic radiuspore radius of resinsize exclusion quantification nuclear magnetic resonancesize exclusion chromatographysignal-to-noise ratioaccessible pore volume for a certain moleculetotal pore volumetotal volume in columnvolume between resin beads1

2

1 IntroductionChromatography is a common method for separating molecules by letting them pass through aporous material called resin. The resin can have different characteristics regarding size andchemical properties. What resin to use in chromatography depends on the characteristics of themolecule of interest. The most important characteristic of the resin when separating moleculeswith respect to size is the pore structure, which is difficult to both specify and asses. The mainaim of this project is to optimize a newly developed method for pore characterization. Thisproject is performed as a master thesis at Cytiva in Uppsala. The current method for porestructure characterization used by Cytiva is inverse size exclusion chromatography (ISEC)which uses defined polymer standards to characterize the pore structure of resins. ISEC is areversed form of size exclusion chromatography (SEC) where instead the pore structure isknown, and molecules are separated with respect to that structure. SEC is a common methodfor determining molecular weight distribution and in practice is the experimental procedure thesame for both methods. ISEC is a simple method in terms of execution but it is also timeconsuming and requires packing of a column making it less optimal for quality controlmeasurements (Guo et al. 2017). Cytiva produces a range of materials used to purify biologicalpharmaceuticals and a routine procedure for characterizing these materials in a reproducibleway is required.The method to be optimized in this project was developed in 2018 by Elwinger et al., where anovel and promising approach for pore structure characterization was presented. The method iscalled size-exclusion quantification nuclear magnetic resonance (SEQ-NMR), and the principleof the method is based on a solution of polymers with a wide size range equilibrating with theporous material to be examined. Smaller polymer fragments within the solution have access toa greater part of the total pore volume compared to larger polymers and their concentration willtherefore be reduced in the surrounding solution. The focus of this project is to optimize SEQNMR to make it more time efficient and, in the future, suitable for quality control analysis.Throughout the optimization the method will also be examined regarding its sensitivity androbustness together with other aspects of its performance.In SEQ-NMR a solution of polymers is analysed with diffusion NMR before and afterequilibration with a resin. Through a multiexponential fit to the received data the change in sizedistribution is obtained making it possible to determine the pore structure of the resin. In SEQNMR no column packing is needed, and the pore structure can be determined by using asolution with broad size distribution of polymers, i.e. no monodisperse polymers are needed.This makes SEQ-NMR more advantageous compared to the ISEC method (Elwinger et al.2018). ISEC takes 15 hours which is equivalent to one diffusion measurement of SEQ-NMR.Two measurements are needed giving SEQ-NMR a total duration of 30 hours. The aim of thismaster thesis is to reduce the duration of SEQ-NMR to less than one hour per diffusionmeasurement while making an evaluation of the method.3

2 Background2.1 NMRNMR is based on nuclei having properties as angular momentum and magnetic moment,referred to as spin (Hore 2015). The spins of the nuclei in a sample to be analysed are at firstrandomly oriented. When the sample is put in a static magnetic field the magnetic momentswill take the direction with or opposite the direction of the magnetic field. The distinctorientations relative to the magnetic field will exhibit slightly different energies and will bepopulated according to the Boltzmann distribution. The difference in energy between spindirections is what makes all NMR measurements possible (Bruice 2010). Further insight to thebasics of NMR can be found in P.J. Hore’s book Nuclear Magnetic Resonance.2.2 Diffusion NMRDiffusion NMR is being used in medical, biological and material science applications (Guo etal. 2017). By measuring the self-diffusion of a molecule, it becomes possible to study its sizeand shape. Diffusional movement is driven by thermodynamic energy and can be quantitativelydescribed by the self-diffusion coefficient D. This coefficient is a measure of with what rate amolecule moves in the unit m2s-1. From a diffusion NMR measurement information on therelative size of a molecule can be received from the relationship between D and hydrodynamicradius 𝑟𝐻 ,D kB T.6πηrH(1)This is known as the Stokes-Einstein equation where kB is the Boltzmann constant, T is theabsolute temperature and η is the solution viscosity. Hence, the diffusion coefficient is inverselyproportional to the radius of the diffusing molecule. How well Eq. (1) provides an accurateestimation depends on the shape of the studied compound, the more sphere like a molecule is,the better is the estimation. (Claridge 2009).When relating the diffusion coefficient to size via the Stokes-Einstein equation the temperatureis often known from calibration, but the solution viscosity can be harder to find or determine.At low concentrations, in the millimolar range, where interactions between the diffusing entitiescan be neglected the viscosity of the sample can be assumed to be the same as the solventviscosity. Other ways to determine the viscosity is by using a molecule with knownhydrodynamic radius in the solution and calculate the solution viscosity from the measureddiffusion coefficient or by measuring the viscosity of the mixed solution. The requirement forusing a molecule with known hydrodynamic radius is that there should be no interferencebetween the signal of the reference molecule and the other components in solution (Claridge2009).4

2.2.1 Pulsed field gradient stimulated echoThe pulse program used for diffusion NMR experiments in this project is pulsed field gradientstimulated echo (PGSTE). Here, two gradient pulses with strength g and duration δ are appliedto the sample separated by the diffusion time Δ (Claridge 2009). The applied pulses can havedifferent shapes and rectangular pulses are used throughout this project. It is the simplest pulseshape and can therefore suffer some drawbacks when using high gradients (Willis et al. 2016).The first gradient pulse gives a spatial dependence of the magnetization in the sample providinginformation on the position of the nuclei. The second pulse reverses this dependence byrefocusing the dephased magnetization. The refocusing is only perfect if the nuclei are inexactly the same physical location at the time of the first and second pulses. If diffusion occursthe refocusing will not be complete and an attenuated signal is obtained. The attenuated signalis both dependent on the length and strength of the gradient pulse and on how far the moleculesdiffuse during Δ which in turn depends on the diffusion coefficient of the molecule. To measurethe diffusion coefficient of a molecule, g can be applied in a gradient of increasing strength,higher g means more dephasing of the magnetization in the sample leading to less signal beingrefocused. With increasing g there will be more attenuation of the signal (Claridge 2009).The data analysed from PGSTE experiments is the attenuation (E) of the signal due to diffusion.By integrating the peak area as a function of g the attenuation can be described by followingthe equation,𝐸 𝑒 𝑏𝐷(2)𝛿𝑏 𝛾 2 𝑔2 𝛿 2 ( )3(3)where b in the exponent is given by,where γ is the gyromagnetic ratio of the nuclear spin (Price 1997). In diffusion measurementsit is common to use protons, 1H, since it is stable and the nucleus with the highest γ (Hore 2015).As illustrated by Eq. (1) above is the diffusion coefficient not only dependent on the size andshape of the analysed molecule but also the viscosity of the solution, the temperature of thesample as well as concentration. This means that the parameters used in a diffusion experimentoften need to be optimized for each new sample. The principal parameters δ, g and Δ can bedetermined once the physical parameters of the sample have been set. When deciding theprincipal parameters, the goal is to receive a considerable attenuation, so that the subsequentfitting to the data yields reliable results. If the attenuation is too rapid the late recorded datapoints, having high b values, will not contribute to meaningful fitting to the data and if theattenuation is too slow it will not give an accurate determination of D (Claridge 2009). Commonvalues for Δ are between milliseconds and hundreds of milliseconds making it possible formacromolecules to diffuse a distance much longer than their own radii during Δ. The study oflarger molecules having small values of D require higher values of Δ which can lead to a5

decrease in the signal-to-noise ratio (SNR). Small sample volumes can also give a reduction inSNR. Common values for δ are 1-10 milliseconds (Claridge 2009, Stilbs 2019).2.2.2 Data processingThe data obtained in a diffusion NMR experiment is a measure of the attenuated signal as eitherpeak height in the spectra or integrated peak area, both as a function of g. From this can D bederived either by plotting E against b and make an exponential fit, or by plotting E on alogarithmic scale against b where a straight line will be given with -D as the slope. This ispossible since all parameters within b are constant except g (Claridge 2009).2.3 Sensitivity and SNRThe signal strength of NMR as a technique is considered low. SNR is a well-established conceptof signal processing and in NMR is the SNR defined as the height, or as in this report, the areaof the NMR peak divided by the root mean square of the noise (Hyberts et al. 2013). The noiseis obtained by integrating ten equal areas far away from peaks in the spectrum and calculatingthe standard deviation for these integrals. The signal is deterministic, constant, and the noise israndomly fluctuating, making SNR increase by the square root of the number of scans (Hore2015).To receive a high-resolution spectrum, it is important that the magnetic field is homogeneous.To increase the homogeneity of the field the currents applied to specially designed assistingcoils can be adjusted, a process called shimming (Topgaard et al. 2004). The shimming isperformed prior to all measurements and, when using short sample heights, the process can bevery difficult. The sample must moreover be placed in the sensitive region of the signalreceiving coil which is approximately 1 cm along the sample height (Price 2009).2.4 Polydispersity and dextransThe polymers used in the SEQ-NMR measurements in this project are dextrans, which arecommon in biotechnological and pharmaceutical applications. A dextran is a branched glucanpolymer with branched chains that can differ in both length and weight. This means that adextran is polydisperse and the branches may often consist of one or two glucose molecules.The molecular weight of a polymer is therefore a mean value that is dependent on its actualmolecular weight distribution. Three variables are commonly used for describing the weight ofa dextran polymer. Mn is the number-average molecular weight, Mw is the weight-averagemolecular weight and Mp is the peak molecular weight. The molecular weights are oftenmodeled by lognormal distribution with the relationship between the weights as Mp Mw Mn(Kuz’mina et al. 2014). Throughout the report the dextran weights will be given as Mp values.An NMR diffusion measurement on a polydisperse solution will give data as the integral of thesignal from all polymers in the solution (Guo et al. 2017). Consequently, the diffusionalattenuation measured is affected by polydispersity present in the sample.6

2.5 SEQ-NMRAs described in the introduction, SEQ-NMR uses a solution of polymers which is analysedbefore and after equilibration with a resin. The polymers used in this project are dextrans asstated in Section 2.4, the solution is made by dissolving the dextrans in heavy water, D2O. Thedistribution of each dextran follows lognormal distribution and is computed as𝑃𝑖 1 2𝜋𝜎𝑖2𝑒1 𝑙𝑛(𝑀) 𝜇𝑖( 2() )𝜎𝑖,(4)where M is the molecular weight, µi is the expected value and σi is the standard deviation ofdextran i (Chang 2015). M is connected to 𝑟𝐻 through the Mark Houwink equation,𝑟𝐻 𝑎𝑀𝛼 ,(5)where a and α are molecule specific parameters. The distribution coefficient Keq and 𝑟𝐻 givesthe selecitivity curve further described in Section 2.5.1.2.5.1 Selectivity curveThe theory behind the selectivity curve as a result of SEQ-NMR experiments is shared with thetheory of ISEC experiments. The selectivity curve is the distribution coefficient Keq as afunction of molecular size expressed in 𝑟𝐻 , for a one-pore model are they connected byKeq (1-rH 2) ,𝑟p(6)where 𝑟p is the pore radius of the resin. The one-pore model assumes all pores of the resin beadto have the same size and structure. Regardless of the pore model is Keq also connected to thevolumes describing the functionality of a resin by the relationship𝐾𝑒𝑞 ���𝑜𝑟𝑒(7)Vaccessible is the accessible pore volume for a certain molecule, and Vpore is the total pore volume.In a packed column of porous beads, Vpore is given by Vtot minus Vvoid where Vtot corresponds tothe total available volume of the column and Vvoid corresponds to the volume in between resinbeads (Knox & Scott 1984). In practice, Vvoid can be determined with the help of a largemolecule that cannot access any pores in the column while Vtot is assessed with the help of asmall molecule that can access all pores (Knox & Ritchie 1987). In this project Vtot is determinedby measurements using D2O, and Vvoid by measurements with a large polyethylene oxidepolymer (PEO). These are not measured with diffusion NMR, instead the volumes arecalculated by the NMR signal being proportional to the concentration c. The dilution equation𝑐1 𝑉1 𝑐2 𝑉2 where the concentrations c1 and c2 are from the NMR signals before and after7

equilibrium and V1 is the volume added to the resin making V2 the sum of V1 and the searchedfor volume. The NMR signal is hence, turned into volume by comparing the signals from beforeand after equilibrium (Bruice 2010).The solution of polymers will after equilibrium with a resin be diluted compared to the stocksolution added to the resin. This dilution d is described by𝑑 𝑉0,𝑉0 𝑉𝑣𝑜𝑖𝑑 𝑉𝑎𝑐𝑐𝑒𝑠𝑖𝑏𝑙𝑒(8)where V0 corresponds to the volume of stock solution added to the resin before equilibration.The value of d will be specific for each polymer length in the solution (Elwinger et al. 2018).The change in concentration will give information on the pore size distribution in terms of theselectivity curve.2.5.2 Attenuation curvesWhen a solution contains multiple compounds giving rise to the attenuated signal theattenuation becomes a summation of Eq. (2) for each individual component. The attenuationbefore equilibrium is calculated by𝑁𝐸 𝑃𝑖 𝑒 𝑏𝐷𝑖 ,(9)𝑖 1where the distributions and diffusion coefficients are specific for each dextran i. The attenuationafter equilibrium is then given by combining Eq. (8) and Eq. (9) as𝑁𝐸 𝑑𝑖 𝑃𝑖 𝑒 𝑏𝐷𝑖 .(10)𝑖 12.5.3 Evaluation of fitted modelsThe model fit can be expressed in terms of R2 known as the coefficient of determination. R2 isthe sum of fit residuals squared relative to the sum of the mean square deviations from theaverage value of the data,𝑅2 𝑛𝑖 1(𝑦𝑖 𝑦̂𝑖 )2. 𝑛𝑖 1(𝑦𝑖 𝑦̅)2(11)In equation 11, yi are the observed values, ŷi the fitted curve and y̅ is the mean y-value. WhenR2 is close to one the data overlaps well with the fitted model (Smith 2015).8

2.6 Experimental errors in diffusion NMRThe most common factors giving errors in the diffusional attenuation are convection, nonuniform gradient pulses and eddy currents (Kuz’mina et al. 2014).2.6.1 Convective flowOne common reason for receiving unreliable diffusional data from a modern NMRspectrometer is convection within the sample. Convection arises from temperature gradients inthe sample caused by the temperature regulation in the NMR spectrometer. The regulation isoften performed by a flow of gas passing over the sample tube. To ensure the sample is exposedto constant temperature the gas is often heated before entering the probe. The sampletemperature is regulated via a feedback mechanism controlled by a sensor placed in the probeat the base of the sample tube. As a result, the overall temperature of the sample will be stable.However, if extensive heating is required thermal gradients may appear within the sample. Thiscauses convective flow which displaces molecules leading to a faster signal attenuation thanwhat self-diffusion would generate. Hence, the data provide larger inaccurate diffusioncoefficients (Claridge 2009). Since convection is present over the whole sample volume, everymolecule in the solution is affected by convection in the same way regardless of their size (Price2009). It is important to check for convective flow before performing diffusion measurements(Claridge 2009).One way to reduce the temperature gradients leading to convective flow is to removetemperature regulation, having no gas pass over the sample and make sure that the temperatureis equilibrated. This method is both limiting and impractical. Before all diffusion measurementsthe sample should be allowed to equilibrate for a period of approximately 30 minutes. Thetemperature gradients can also be reduced by having a high flow rate of the gas passing overthe sample. This can limit convection but may instead cause vibrations in the sample. A way totest if convective flow is present is by doing the same experiment with different values of Δ andthen compare the resulting diffusion coefficients. If no convection is present, the diffusioncoefficient will be the same for the different experiments (Claridge 2009).2.6.2 Non-uniform gradient pulsesThe attenuated signal as described by Eq. (2) will not be complet

produktionen sker på ett reproducerbart sätt. Alla geler som produceras testas därför för att kontrollera att de upprätthåller den kvalité som krävs för produktion av läkemedel. De biologiska läkemedlen kan sorteras på olika egenskaper och för geler som separerar med

Related Documents: