Biomedical Engineering - Biomedical Physical Chemistry

Completed notes of the course

Complete course

1 LECTURE 1 INTRODUCTION TO THERAPEUTIC PROTEINS BIOLOGICAL MEDICINAL PRODUCTS (BIO – THERAPEUTICS/BIOLOGICS) We talk about whenever there is a living organism (a cell) which is involved in the production of the biological product. Pharmaceutical: word that started with very small organic molecules and kept going with even smaller molecules (antibodies). The future is going toward gene therapies and cellular therapies. In the FDA definition the human source is included in the natural sources. Biotherapeutics: something eventually produced by some organism that we can either extract or guide. PROTEIN STRUCTURE Proteins are extremely large molecules. The primary structure is the simple sequence of the aa. The secondary structure does not involve the entire structure of the protein but only one part. It indicates the region where the aa have organized themselves according to some familiar structures. The 3rd structure is the entire structure. It’s the total arrangement in space. The third structure is the one necessary to have bioactivity! The 1st and 2nd aren't enough in order to have a function. The quaternary structure doesn’t always exist. It’s typical of those proteins that perform a specific activity. 2 RECOMBINANT DNA TECHONOLOGY DNA sequences that code a specific gene of interest. We cannot produce synthetically a protein, but we can produce a cell that can produce a protein by inserting the necessary DNA sequence. We have many organisms that we can use to inject the DNA sequence. They produce the same basic proteins but with some differences (ex. some are highly glycosylated). LIVING ORGANISMS FOR PRODUCING BIOLOGICS To produce biologics we can use humans, for instance by taking proteins from the blood (such as albumin), or we can use genetically modified organisms (CHO cells are today the basic tool that is used). Monoclonal antibodies are our main interest, and they are taken by using the CHO cells. Cells are cultivated in the bioreactor. PROTEINS Insulin is a protein characterized by 2 chains: A and B. There are 21 aa in the A chain and 32 aa in the B chain. There are some interactions/bounds between the 2 chains or within one same chain. In the protein ribonuclease some regions are organized as alfa helix, some as beta sheets. 3 There are different steps in the production of a protein. After transcription (DNA in RNA) and translation (RNA into an amino acid sequence) the cell machinery takes care of post translational modifications:  Protein folding in Endoplasmic Reticulum (the secondary and tertiary structures are formed to make the protein bioactive),  Glycosylation in Golgi Apparatus,  Others. CLASSIFICATION OF GLYCOFORMS In the Golgi apparatus happens the glycosylation , a fundamental step that guarantees the correct functionality of the protein. There are different steps and the process is overall very complicated. It’s possible to observe in mAbs 8 typical glycan structures:  G0, G1, G2: addition of 0, 1 or 2 Galactose (Gal).  G0F,G1F, G2F: addition of none, 1 or 2 Fucose (Fuc) groups.  A1, A2: addition of the Sialic acid (Neu5Ac). Note: acid = charged. These 8 structures are the main components of the glycosylation patterns that we observe in pdf 15. This is very important: the same protein might have different types of glycoforms! They might be slightly or very different. So same protein, same structure but different glycosylation. So, even if the primary structure is the same, there might be a different glycoform. These are called variants of the same protein. CHARGE VARIANTS There are also charge variants . The protein is a sequence of aa; all the different reactions (e.g. Acetylation, Amidation, Oxidation etc.) can change the protein somewhere in the structure, producing a variant (charge). 4 So, if I feed the protein population in an ion exchange chromatography, first the proteins with a certain type of charge will be eluted, then the others. Conclusion: proteins have a heterogeneous nature . When a protein is described by its primary structure it’s still not completely identified. It may have differences in the charges or in the glycoforms due to post translational transformations. Different variants: the properties will be affected! But to this day we don’t know exactly how much a variant affects the bioactivity. When I produce a protein with recombinant technology, I actually produce a mixture! My mixture will be characterized by a certain distribution, so a % of proteins will have a certain charge and a % a certain type of glycosylation and so on. Once I have done all the tests and proved that the protein isn’t toxic and it has a therapeutic effect, the FDA might accept it and give me the authorization to produce it and commercialize it. But I must make sure that I always get the same distribution of different variants (glycoforms and charges). It is very difficult to control the post translational transformation. Now, to a certain extent, it’s possible to remove the charge variants not desired with chromatography, but it’s also expensive (I’m trying to distinguish very similar proteins). It’s not possible to separate one glycoform from another one. Note: in the world of small molecules the bio series are generics, they are a drug and they have one specific molecule. When the patent on a certain molecule expires, another company has the right to patent the molecule. This doesn’t apply in the case of proteins. It’s not a bio series. It’s impossible to produce one exact single protein! There’s a mixture, a distribution (charge variant or glycoform) and in order to prove that one protein produced by a company is identical to the protein of another company several clinical tests must be taken. Since the subject is a distribution, the main focus is to prove if they have the same activity function. If yes, then it’s a biosimilar. e.g. Herceptin: This is a rare case where one drug has been fractionated into the different charge variants and each of them have been tested. There are in total four charge variants. The activity of two molecules are very different. We cannot claim that charge variants have the same bioactivity. We have to reproduce the production conditions in the large scale that we have used in the small scale to produce the amount of material which went to the clinical test. We need to obtain the same material with its variant’s distributions. 5 STRUCTURAL CHANGES Post translational modifications might involve other processes such as the proteolytic processing (enzyme cuts some aa), the disulfide bond formation, low molecular weight (LMW, proteins which have been depredated) forms or aggregation. Aggregates are constituted by only a few proteins (oligomers) and are often reversable. But, if at some point the oligomers start aggregating with each other they stare growing more. This process here can lead to the formation of high molecular weight (HMW) aggregates. If they grow even more to 100 microns, they become very dangerous. We need to prevent these processes or to remove the aggregates by purification. Note: these processes do not happen in the cell, but after the protein is expressed outside by the cell and is in the supernatant. These aggregates affect the immune response of the organism and the efficacy of the cell, this is why they have to be removed. Generally speaking, this happens also inside the cell (e.g. formation of fibrils, which are aggregates of amyloid that work in the brain. The consequence is probably the Alzheimer’s disease). PROTEINS DISPLAY MICRO HETEROGENEITY 6 Non-exhaustive list and description of some typical product CQAs and some examples of CPPs that could affect them Purity of Biologics (Critical Quality Attributes) 7 SCHEME OF A TYPICAL PROCESS FOR THE MANUFACTURING OF MONOCLONAL ANTIBODIES 1. Cell culture: we are talking about antibodies. The glycosylation of the protein in fundamental to understand the distribution of the drug in the body and the kinetics of the permanence of the effects: therefore, glycosylation is directly correlated to the effect and efficiency of the drug. Mammalian cells are the only ones that satisfy this request. 2. Primary recovery (centrifugation – filtration): it’s the simplest unit. The cells are removed and separated from the supernatant. If the proteins are expressed outside the cells in the supernatant, the cells get directly removed. If the type of protein is instead expressed inside the cell, the cell must be first broken. 3. The clear supernatant (the one without cells) goes to the protein capture step. It still contains the impurities (DNA, other proteins, cells fragment etc.). The target protein must be separated by the rest. In this level, at 1L of supernatant there might be only 1 g of target proteins (antibodies in this case). The separation process is very difficult! The product that we desire is available in a very small amount, so we cannot accept to lose any part of it. In the protein capture process, it’s important to recover 90% of the quantity available. Since the antibodies will be injected in the patient, a purification process is needed. So, the process must be a high recovery and high purity process. We want to collect only the designed protein and ALL of them. It’s very easy to obtain high purity with low recovery, and vice versa . It’s very difficult to have both purity and recovery simultaneously. - Recovery: fraction of the designed product in the supernatant that is recovered. - Purity: grams of the target product in the overall quantity recovered. This is the reason why the manufacturing process has the capture and polishing steps . Capture: recovery. All the monoclonal antibodies are captured + other molecules not needed. After the protein capture the purity is still too low (e.g. 80%). 4. in the polishing step the much less material reduced in the capture step is purified. Something might get lost. If we lose some product, we lose it in this step. After each of these steps there are the viral clearance: one is based on pH and the other one on filtration. These steps are fundamental because in the different steps some viruses might be introduced. If we go into the formulation of the drug with some viruses present everything goes wrong! It’s very dangerous. All different type of viruses will be removed: some don’t survive very acidic environments and others are removed in function of their dimensions through filtration. We must reduce their concentration at least of 10 order of magnitude. We can't ever get to zero. 5. formulation : the drug is formulated in a proper medium to be delivered to the patient. The delivery can be oral, intravenous injection etc. Generally speaking, proteins are too delicate to go through oral or nasal delivery. Note : during the viral clearance 1 the low pH deprotonates the charges present in the proteins. The unfolded protein exposes the hydrophobic portions in water. The regions that are more hydrophobic, once opened, might interact with other hydrophobic portions in other proteins forming aggregates. 8 Process design: discussion basis Integrated continuous and batch processes B. The Traditional Fed-Batch Process is the one used for monoclonal antibodies. The bioreactor receives the media (nutrients for the fermentation process). The bioreactor goes on for 15 days, after which is emptied, and the harvest is filled. After centrifugation there is the clarified harvest vessel. The next step is the flow to the capture unit, which has a much smaller volume, since a big part is eluted. Through the capture step we concentrate, and the quantity is smaller. All these processes are batched. A. The Traditional Perfusion Process has a continuous bioreactor, with a complete flow from the media to the bioreactor. There is a flow going in the bioreactor and a flow going out in the harvest tank. Each one of these preparations is one at a time in the fed batch, meanwhile in the perfusion one it’s all one flow. The MF Feed is needed because before feeding the harvest to the capture we must remove the cells. This operation can be done in the bioreactor or in a separate step. C. In the new platform everything is continuous: it’s a mix between A and B. End – to -end continuous integrated production 9 LECTURE 2 FUNDAMENTALS OF CHROMATOGRAPHY Both the capture and the polishing step will constitute the downstream product, and both use chromatography (there are different types). PRINCIPLE OF CHROMATOGRAPHY There’s a column filled with a specific solid, which has the capability of adsorbing more or less different chemical species which are present in the feed depending on the affinity of those molecules with the solid phase. The feed, which is a mixture of several different components, is push in the column by adding an eluent. The feed is eluent + a mixture of different components : each component goes through the column but have a different affinity absorbing in different places along the column. Each different component goes through the column, and it’s characterized by a different affinity with the solid. To flow through the column a certain time is needed: first the component adsorbs to the solid phase, then dis-adsorbs and goes down, then adsorbs again and so on. A tracer is a component which has zero affinity to the solid , so it flows through the column without spending any time in the solid phase. All the other components have a sort of affinity: those that adsorbs the most will come out later, meanwhile the ones with lower affinity will come out faster. The affinity of a chemical species to the solid is proportional to the retention time , which is the time needed for the chemical species to show up outside the column. ↑ affinity, ↑ absorp
�on, ↑ retardation time. By collecting at the right time the different fractions, we can divide the mixture’s components: pure red, then pure brown, then pure yellow. By plotting as f(t) the concentration in the outlet of the column we obtain a signal called chromatogram. Chromatogram: it’s a graph that displays the concentration as a function of time at the outlet of the column. It works with UV detector: three peaks = three different molecules. Note: we don’t know which are the different chemical components. We take some samples and analyze them with some analytic tools. The UV signal is proportional to the concentration of the different molecules present on the system depending upon of their response to the UV signal. If two components come out at the same time, they will show up together and in the UV signal will be seen only one peak. Saying that there is only one component would be wrong. Equal affinity = equal retention time = same peak. This is why there’s a second step: the fractions relative to the different peaks will be analyzed, for instance, with mass spectroscopy and distinguished. Another analytical tool would be to inject the fractions in another chromatography column with a different stationary phase. I hope that the two different components that had the same affinity with the first solid phase, will have different affinity for the solid phase of the second chromatography. Orthogonality of stationary phases! A stationary phase separates molecules based on their affinity in respect to specific chemical species of the stationary phase. If there are two stationary phases which separate chemical species based on different affinities definitions, like same charge or same size, we call these stationary phases orthogonal. 10 Orthogonal: the separation is based on chemical physical characteristics which are different. If we have instead two stationary phases but they both separate chemical species based on the same chemical physical characteristics (e.g. the charge) they are not orthogonal. The same physics is used to separate the chemical species. In order to separate chemical species as much as possible it’s better to use orthogonal chromatography. Chromatography is used as an analytical tool combined with analytical standards, not as a production tool. Its job is to tell how much there is of different molecules. Example: bio- separation There’s a mixture of 3 proteins. Albumin (S) is the one that comes out immediately, to the point that the peak is barely visible at the beginning. Now, Lysozyme (L) and chymotrypsin (C) are not completely separated in the middle. This is a problem: which one is coming out in that moment? This is due to the dispersion phenomenon : same molecules do not have the same retention time. The dispersion phenomenon might be due to some mixing phenomenon such as turbulence, and it’s what causes the enlargement of the peaks. In the example, the dispersion phenomenon in not sufficient to create any overlapping between S and C, in fact there is a base line separation. The peak goes down to the baseline and then reappears as another peak. Therefore, the area under the peak is for sure S. when there isn’t the base line separation it’s not possible to tell which one is what. It’s not possible to make soon a prediction of the quantity C and L. Some estimations and approximations must be done. The dots are the gradients present in the elution phase. Stationary phase: Adsorbents The stationary phase can be formed either of spheres of same size, or of particles that have a much rougher form and overall different geometry. The monobeads are better performing but more expensive, the random geometry particles are cheaper but less performing. There is a trade off between cost and efficiency. 11 PACKED COLUMNS The stationary phase is filled with particles (the spheres in the figure). Each one of the particles is porous (white regions in the figure), so the liquid can enter into the particles and in the pores. The solid is where the molecules adsorb. Now, molecules adsorb not only on the external surface (which is very small) but on the internal one, which is very wide. We can distinguish two porosities in the stationary phase: 1. Interparticle porosity: voids in between the particles, 2. Intraparticle porosity: voids inside the particles. The surface available for absorption is the one of each of the small pieces that are the walls of the pores. The sum of this area is >> of the area of the sphere, to the point that this last one can even be neglected. This is done on purpose. The adsorption surface must be as big as possible. The capacity of the stationary phase is the quantity of molecules that it can absorb, which is therefore proportional to the surface of the stationary phase. Porous materials have surfaces of 100 m 2 per gram. One good adsorbent might be able to adsorb on the surface about 10% of its own weight. 10 grams of stationary phase might be able to adsorb 2 grams of molecules. ������ = ������������������������������������������������������������������������������ ������������������������������ ������ ������ ������ = ������������������������������������������������������������������������������ ������������������������������������������������ ������������������������������������������������ ������������������������������������ ������ ∗= ������ + (������ − ������)������ ������ ������ = ������ ������ : ������������������������������������������������������������������ ������������������������������������������������ ������ = ������ ������: ������������������������������������������������������������������������ ������������������������������������������������ ������������= ������������ ∗ ������ = ������������ ∗ ������ : ������������������������������������������������������ ������������������������ Where: - V: total volume of the column. - Q: volumetric flowrate (pump). - S: column cross section. - ε: bed porosity/the packed column. - εp: porosity of the particles. - ε*: total porosity: inter + intra voids. 40% of the volume is usually occupied by the interparticle voids, 60% by particles. εp ≈ 20%, so 80% of the particle is occupied by the solid. - Q: volumetric flow rate controlled by the pump. [L/min], [ml/second], [m 3/s]. - S: column cross section. - u: superficial velocity that we would have by feeding that volumetric flow rate in an empty column. It’s a fictional velocity: in reality, most of the cross section in occupied by the solid. The liquid has to cross a certain surface. - v: interstitial velocity. It’s the real velocity (the flow in between the particles). The tracer moves with velocity u, since it wouldn’t go through the voids. - t0: residence time. It’s the time that a tracer would take to cross the column (time needed to cover L with u). Note: the residence time depends on which voids we are talking about. - t0 = Vxε*/Q or t 0 = Vxε/Q - If ε* = 1 → all voids → t 0 = L/u (empty column), - If the tracer is small enough to enter the intraparticle voids, we use ε*. - If the molecule is very big and can’t enter intravoids, it’s ε. 12 HIGH PRESURE LIQUID CHROMATOGRAPHY A computer controls and processes data. The column is the smallest object in the apparatus. The feed, before going in the column, starts in a pump that starts the flow rate into the column and is composed by different components. The detector gives back the peaks by detecting the UV signal. The fraction collector is used to recognize the components of the single peaks. Only a fraction is taken, the rest is wasted. In each point along the chromatogram is obtained an actual chemical analysis (name of those chemical species). PHASE EQUILIBRIUM: LINEAR ADSORPTION ISOTHERM We talked about affinity of the molecules (blue in the figure) to a certain solid phase. The affinity translates into an absorption process on the surface: a new phase is formed. The new phase is called 2D adsorbed phase and it has different characteristics compared to the 3D fluid phase (T, P, V, c). The concentration in the liquid phase is somehow related to the concentration of the same component in the solid phase. The adsorbed molecules don’t enter into the bulk of the solid but create a new phase on its surface. When we talk about the concentration, we refer to the entire particle volume, not only to the surface. Note: particle volume = adsorbent solid. The model is the LINEAR ADSORPTION ISOTHERM. Isotherm because it’s at the same T. The linear adsorption equilibrium indicates that the concentration in the solid is proportional to the concentration in the liquid. The constant which defines the linear relation is the Henry constant H i. ↑ the concentration in the liquid ↑ the quantity of the component in the solid. This linear adsorption isotherm applies to diluted systems! If the solid gets completely covered there is saturation : so, by increasing the concentration in the liquid the 2D adsorbed phase does not increase anymore. In a diluted system the concentration in the liquid phase is very small, therefore the concentration in the solid is very small, with no interaction between the molecules on the solid. They are very far from each other and they don’t affect each other. There is no interaction among the molecules in the adsorbed phase. The only interaction is between the solid and the liquid. The Henry constant is what defines the affinity of a certain component i to a certain solid. Different components will have different affinities to the solid phase. High affinity = high Henry constant . 13 If there are two different components there is selectivity. - If H 2>>H 1 then S 21 > 1 - If H 2 q ) there would be adsorption. - If the concentration that we would have at equilibrium is smaller than what we currently have ( q*< q ), there would be de-adsorption . Adsorption Modeling: Hindered Mass Transfer (for computing K m) Here K m is computed in a more structured way, in order to account for the fact that the adsorbed molecules are proteins (j different types), which are big enough to have a dimension comparable to the one of the pores. This model takes in account the hindrance and transport of the protein trough the pores of the stationary phase. There can be processes that can decrease the section of the pores, making the available mass transfer coefficient smaller. - At the beginning, when there isn’t anything adsorbed, the pore is free and empty, and we have the maximum value of K. In fact, if all the q j are zero we get K mmax . - If we are at saturation conditions, km = K max xS 1, where S 1 is the fraction of mass transfer coefficient remaining at saturation. As the concentration in the solid phase increases from zero to the saturation value, the pore’s mouth becomes smaller and smaller, reducing the mass transfer coefficient, which goes from a very big value to S 1* K mmax . - Note: S 2 is another parameter that can be used when we want to modulate the variation of K m, from k mmax to km max xS 1 to be linear or maybe non linear. 36 ADSORPTION MODELING: SHRINKING CORE MODEL (da non usare nel modello in silica) Model I The adsorption of the mAb to the protein A is very very fast: as soon as it sees the protein A, the mAb binds to it. At the same time, the diffusion of the mAb inside the pores is kind of slow, because it’s a big protein and, once inside the pore, it moves slowly. The protein doesn’t travel much inside the pore, but it adsorbs instantaneously. The mAb diffuses from the bulk of the liquid phase through the surface of the particle, crossing the boundary layer (the film of liquid stagnant around the particle), to r p, which is the surface of the particle . It starts then diffusing inside and adsorbing simultaneously. However, we can use an approximation that sees the adsorption processes are very fast compared to the diffusion process . There is inside a shrinking core: the mAb immediately sees the protein A and adsorbs saturating the layer, then comes more and more inside making the adsorption layer bigger and bigger. We can simplify the general rate model saying that we have a diffusion without adsorption in the external boundary layer. In the shrinking core model, we can divide the particle in two regions: 1. the front one, where the particle is saturated. All the ligands of protein A have been saturated by mAb. There cannot be any more adsorption. In the grey region there cannot be any adsorption because here q=q*, there is saturation, 2. internal region where the mAb has not yet arrived. Here adsorption doesn’t occur because there is no mAb into the pore! There aren’t any antibodies in the liquid phase. The adsorption happens only at the interface between the core and the shell . The velocity is dictated by how fast the protein arrives from outside. In the core there is no protein, so the concentration of the protein in the core (from 0 to R) is zero, therefore also the concentration in the solid is zero (the protein hasn’t arrived yet in this region). In the shell the concentration in the solid phase is equal to the saturation concentration, so the protein diffuses through the shell without adsorption because the stationary phase is already saturated. There is also the boundary layer of the polymer around the particle. Diffusion is through the boundary layer, which is the stagnant liquid around the particle. Therefore, the diffusion is from the bulk of the liquid that flows, through the boundary layer to the surface of the particle. There are “three” diffusion processes. 37 One is defined by K f, which is the mass transfer coefficient in the liquid phase. Here there is no adsorption because there is no stationary phase, only stagnant liquid. Here the concentration is c. The concentration becomes C p(rp) at the surface of the particle. The second diffusion process is inside the cell. It’s pure diffusion, therefore the gradient is linear and there is no adsorption. The concentration isn’t anymore c p* but becomes zero at the core-shell interface: here all the protein is adsorbed by the stationary phase. The concentration of the protein remains 0 in all the regions inside the core. This situation is described with the mass balance in the shell. The flux within the shell remains constant. C*? At the surface of the particle the flux of the protein coming in is equal to K f(c-c*) which is the flux computed through the stagnant liquid film. But if we compute the same flux coming in or leaving the surface of the particle to go inside the cell, this will be given by D pdc/dr. dc/dr can be computed from the c written above. R is the location of the boundary between the core and the shell. The assumptions are that there is no protein in the core, the stationary phase is saturated by the proteins in the shell. But, nothing has been said about the adsorption process. In the shrinking core model all the flux of proteins coming from outside are adsorbed instantaneously at the interface between the core and the shell and are saturated. Model II ( With this model here we compute dq/dt.) - Kf is the film mass coefficient in the boundary layer. - Kp is the mass transfer coefficient in the saturated shell where nothing occurs. - Km is the total mass transfer coefficient is the some of the reciprocal of the mass transfer coefficient in the field and in the saturated shell. - X is the saturation coefficient. 38 THE TWO SITES HIERARCHICAL ADSORPTION MODEL This model takes in consideration the fact that proteins are adsorbed by protein A. Each ligand of protein A can capture two mAb. There are 2 sites. From a physical point of view the two sites of protein A have the same affinity towards the mAb. They have the same Henry constant. Therefore, the mAb at first can bind to one protein or the other with equal probability. But, once it binds, the other one becomes less probable; the second site becomes weaker. When one of the two is already occupied it’s more difficult to occupy the second site due to steric hindrance, provided by the mAb adsorbed on the other site. INTEGRATING A CHROMATOGRAM The process performance parameters (yield, purity and mass balance) describe how the process is performed. The yield is how much of the product that has been injected into the column is recovered. It’s the rate between of the red fraction and the area of the peak. The purity refers to the product fraction: in the fraction recovered how much product there is? How many impurities? The purity is the ratio between the amount of pure product (target product) and the total area underneath the red peak (where there might be are also blue and yellow parts). There is a tradeoff between purity and yield, and it’s the reason why chromatography is complicated. By integrating the concentration of that component (blue, yellow or red) between the two times t start and t end of that particular window, we get the amount of material which is eluting from the column, between the interval of time chosen. In order to make sure that the discretization chosen is good enough we have to make sure that the mass balance is satisfied. 39 FITTING A CHROMATOGRAPHIC MODEL 40 FITTING CHROMATOGRAMS We compute a function which is the different of concentration in the peaks (y k) at different times measured at different times and the concentrations obtained with the model (f k(x)). We want the difference to be as much as possible. We have, for each chromatogram, many points in time or many experiments. The equation is a function of x, where x is some model’s parameter that we don’t know. We calculate it trying to make the difference as small as possible. Fitting chromatograms: estimation of one of the models parameters by fitting the model response to experimental errors. 41 LECTURE 4 PRINCIPLES OF CONTINUOUS CHROMATOGRAPHY OUTLINE  Batch, continuous and counter – current chromatography.  Origin of continuous counter – current chromatography. ADVANTAGES/DISADVANTAGES OF CHROMATOGRAPHY Certain molecules can be separated only in this way, thanks to the high selectivity and low temperature of chromatography. Other techniques that are used for big amounts and are cheap cannot be used for protein’s purification. Chromatography however is very expensive compared to other purifications techniques. We need to make chromatography cheaper and usable for the production of big quantities. WHAT IS CONTINUOUS CHROMATOGRAPHY? Continuous: the feed to be purified is continuously fed into the process where the purification occurs. We have a continuous flow leaving that unit with the purified product. So far, we discussed about batch chromatography: a small amount of the mixture is fed to the column and split into the different peaks, and each one of these can be collected purified. In the flip flop operation, we feed first the mixture that has to be separated and then the eluent to make the recovery of the target product that we have fed. If we want to proceed to purify it more, we need first to clean the chromatographic column in order to re-use it. It’s a batch process. By adding a second column, we can directly feed the purified product eluted from the first one. It’s the same “batch” product but repeated in a sequence of different columns. This system has the same efficiency as the batch chromatography! 42 TRANSFORM BATCH INTO CONTINUOUS CHROMATOGRAPHY We operate many columns in parallel where each one of them works as a batch column. BATCH CHROMATOGRAPHY We feed a pulse (a fine amount of the mixture we want to purify), the stationary phase is fixed and the process is discontinuous. In the example it’s a binary mixture, but in realty we have hundreds of different impurities. CONTINUOUS CHROMATOGRAPHY We feed one pulse to the column, and once the processed is finished and we want to clean the first column, it automatically goes to the second one. WHAT IS COUNTER CURRENT CHROMATOGRAPHY? In counter-current chromatography, stationary and mobile phase move in opposite directions. The components to be separated “see” the adsorbent more often (more stages), spend more time in the chromatographic bed and the separation is improved. In chromatographic equipment, the counter-current movement of stationary and mobile phase can be simulated through port switching. 43 BASICS OF CONTINUOUS COUNTER-CURRENT CHROMATOGRAPHY There are two flows of two phases. For example, a gas phase which contains some pollutant which we want to remove before it goes out to the atmosphere. For doing this we concurrent it with a liquid (e.g. water) that enhances the adsorption of the component, transferring it from the gas phase to the liquid phase. An other example could be the mixture of cold and hot liquids. We want to use the hot stream to warm the cold one. By feeding them with the same direction , all along the length the cold stream becomes hotter and the hot stream becomes colder. The driving force, which is the difference between the two temperatures, is decreasing . Therefore, the efficiency of the process becomes smaller and smaller. On the other end, eventually the two streams become equal. If instead of concurrent we go with countercurrent , by maintaining the same length and cold stream in the same direction, the hot stream is fed in the opposite direction. Again, the hot stream gets colder and the cold stream gets hotter, but in this case the driving force stays constant . The outlet temperature of the cold stream can be close to the temperature of the hot one. We don’t want to heat up the cold stream at the very beginning but at the end, when it’s already slightly warmer. The concurrent and the countercurrent flow have very different efficiencies. This can be useful for chromatography, where we have one protein in the liquid phase and one non protein that we want to adsorb in the solid phase. We don’t want to work concurrently as in the batch system, but we want the chromatography to work counter currently. The liquid and the solid move in two different directions. There isn’t anymore a solid phase! Only two liquid phases, which move counter currently. CONTINUOUS COUNTER – CURRENT CHROMATOGRAPHY Continuous feed of mixture + continuous movement of stationary phase. With the slow solid flow it’s not possible to divide the components. Same problem if the solid flow is too fast. With the right solid/liquid flow ration we get the intermediate absorptivity of the two components that makes it possible to divide them. 44 CONTINUOUS COUNTER CURRENT ADSORPTION In 1942 they tried to create the movement between the stationary phase and the liquid phase by dropping the solid from the bottom and letting it come down, and then moving the liquid phase in the opposite direction. This process was never build in a larger scale because the solid gets disrupted after few cicles. FROM TRUE TO SIMULATED CONTINUOUS CHROMATOGRAPHY The idea is that, if we observe from a fixated point of view (the fan), we will see the solid phase as non moving, therefore the stationary phase exists as a non moving part. But we also would see the inlet of the flow mixture moving and passing by. The trick is to design the process not like a process with a true moving bed but a process with respect to the observer sitting on the solid. THE SIMULATED MOVING BED (SMB) PROCESS To do this we take our simulated column but, instead of having one single entire column, the column is cut in many pieces, obtaining the continuous movement of the liquid (arrow) and a non moving solid. The inlet of the feed, the eluent, the outlet of the extract and the raffinate, are moving. We obtain the movement of the liquid, the movement of the pores of the inlet and outlet flows, and a non moving solid. This is a simulated moving bed, not a real one. Every single small column works as a chromatographic column. In order to have a completely smooth movement, we would need an infinity of this pieces → small columns. It’s the switch movement that simulates the movement of the stationary phase. 45 SMB FOR HYDROCARBON SEPARATIONS First SMB planned to separate p-xylene: it contained 24 columns. At the edges of the cycle represented above is necessary to put multipositions valves, therefore 4 connections in total. In the sixties, back when this was planned, it wasn’t possible to use the valves. He used rotatory valves, which rotated in 24 different positions. SBM SUCCESS STORY FOR SMALL MOLECULES Chiral molecules are the reflection of each other. Even tough they are very similar, in nature they are perceived as very different. In the sixties there was a problem with thalidomide , a drug useful for pain release, but only one of the two enantiomers worked. In the industrial production 50% was one type of molecule and the other 50% the chiral copy. One of the two enantiomers worked against nausea, but the other one had the effect to de-build the babies inside pregnant mothers. Now, to be able to produce pure enantiomers the technology used was the one with a SMB. This went from the small scale to the big scale. The model didn’t have 24 columns but only 8. SBM MODELING AND DESIGN We have the feed (A+B+S). The liquid and the switching of the solid move in the same direction. The yellow component (B) is the less adsorbing one and the one in blue (A) is the component that adsorbs more. 46 We want the liquid rate to be sufficiently high to de-adsorb all the yellow pushing toward the raffinate. But, at the same time, the liquid flow rate shouldn’t be too high, otherwise it de-adsorbs also the blue component, which would pollute the raffinate. We want a flow rate fast enough to move the less absorbable component B) in the raffinate but not too much, otherwise it would bring also the most absorbable component (A) in the raffinate (instead of the extract). The flow rate in section 3 has to be in-between the absorptivity of the yellow and the blue. Same thing for the section 2! The component A has to be left behind and the green mixture has to start again the cycle. In order to model this system, we use an equilibrium theory , which neglects all dispersion effects. There is no mass transfer coefficient (it goes to infinity). TRIANGLE THEORY In the graph m 2 and m 3 are the relative velocities liquid to solid in section 2 and in section 3. The liquid flow rate in section 3 is equal to the liquid flowrate in section 2 plus the liquid flow rate of the feed. The entry is larger than the output (we don’t look at half of the graph; that region is impossible). We divide the graph in four regions: for m 3 we have a minimum and a maximum value, and for m 2 we also have min and max values. If we are between the appropriate values of m 2 and m 3, then the system works (yellow region). - Pure extract : m 2 is in the right interval but m 3 is too high: the raffinate gets polluted by the blue component. Only the extract is pure. - Pure raffinate: m 3 is in the right interval and m 2 isn’t: only the raffinate is pure. - Yellow region : pure extract and pure raffinate. It’s the complete separation region. Which point of the yellow region should we select? The difference between m 2 and m 3 is the feed rate . If I know I can have purity (in recovery, raffinate and extract), I pick the point where I can have the highest feed, which is the point where the difference between m 2 and m 3 is the largest (point most far away from the diagonal). The red point has very small robustness , because even a very small change in m 2 or m 3 might give a vastly different behavior of the system. It’s possible to compute the boundary using the equilibrium theory : the boundary is independent from all resistances or mixing/mass transfer in the solid phase. TRIANGLE THEORY – LINEAR CASE In the linear case the triangle becomes a square triangle and the two vertices are the Henry constant of the two components. In reality, these adsorption isotherms are non-linear. 47 BATCH VS SBM PERFORMANCE TERNARY SEPARATIONS – BATCH CHROMATOGRAPHY What happens if we have three components and need a ternary separation? In the case of batch chromatography it’s easy if I know the timings of the different components. TERNARY SEPARATIONS – CONTINUOUS COUNTER CURRENT In this case the concept of a simulated moving fluid does not work. Limitations of 4-zone SMB 48 Typical bio-purification challenge is ternary Here is represented a chromatogram for analytical applications. There are millions of components and we care only about the mAbs. HCPs are the protein produced by the cells during fermentations to survive. We intend to elute first the mAbs. Separation challenge When we want to separate the mAbs we have: - All the weak impurities (W), which elute before the product. - All the strong impurities like the aggregates (S), which adsorb after the product. With a very narrow window we get very high purity but very low yield . With a very large window we get a high yield but a very low purity (many impurities are recovered). Purification challenge Affinity chromatography might work to make the purification high: a ligand might be so selective that, given a mixture of million of impurities and few mAbs, it captures ALL (max yield) and ONLY the mAbs. Nowadays such a perfect stationary phase doesn’t exist. The second option would be a process like SMB for the continuous counter current that can work with a ternary mixture. 49 Combining batch and SMB ↓ MCSGP (Multi-column Counter-current Solvent Gradient Purification) The idea is of a multi column system where we can have a continuous current, that allows for a purified stream, plus other two streams, one for heavy impurities and the other one for light impurities. The first MCSGP prototype (ETH Zurich, PhD Aumann, 2006) Counter current principle 50 LECTURE 5 AFFINITY (CAPTURE) CHROMATOGRAPHY OUTLINE - Process fundamentals and design. - Comparison of single and multicolumn processes. - Application examples. SEPARATION BY ADSORPTION METHODS – VERMEULEN 1958 Among the unit operations, adsorption may be considered a prototype for all fluid-solid separation operations . When it is conducted under counter-current conditions, the calculation methods required are entirely analogous to those for counter-current absorption or extraction . Often, however, it is most economical to conduct adsorption in a semi-continuous arrangement , in which the solid phase is present as a fixed bed of granular particles. The fluid phase passes through the interstices of this bed at a constant flow rate and for an extended period of time. The concentration gradients in the fluid and solid phases display a transient or unsteady-state behavior , and their evolution depends upon the pertinent material balances, rates, and equilibria. TWO BASIC METHODS FOR FIXED-BED OPERATION – VERMEULEN 1958 AFFINITY CHROMATOGRAPHY FOR CAPTURE The feed is the supernatant, which is the liquid that comes out the bioreactor after we remove the cells. It contains between 3 and 6 grams of mAbs per liter. The idea is to feed the mix supernatant in the column, where there is the stationary phase with ligands. The ligands capture the mAbs and all the rest travel trough the column and gets out. In reality, as soon as the target protein (red curve) shows up in the end, if we continue the operation we’re going to lose the target protein into the outlet stream. We are going to lose either in recovery or yield. Therefore, as soon as it gets there we have to stop and go for the regeneration : change of the pH and elution of all the mAbs that were inside. It would remain the mixture full of 51 impurities. So, we want to do first the wash just to get rid of all the mixture with impurities, then we change pH (going acidic to detach the mAbs) and finally we elute the desired protein and reuse the column again. However, there is a big part of the column that hasn’t been used (orange part). The number of times we can use a column is limited: after a certain amount of cycles we have to throw away the protein A column and replace it. After the wash and elution steps there is the sanitation process, which is very harsh (very low pH) and partially damages the column. At the most 100 cycles can be made. In the batch column we can have high yield by stopping at the very end, but we don’t use the total resin capacity. We can keep on going but we’re going to lose some product (low yield). There’s this trade-off. By using two columns (counter current operation) we can operate like before, but this time without stopping at the breakthrough point since there is the second column. The first column is completely saturated. As always, we want high recovery and full resin capacity utilization. LOAD STRATEGIES AND RESIN COSTS COUNTER CURRENT CAPTURE General scheme: counter current capture 52 The feed goes into the column 1. Before this is completely saturated, there is the breakthrough to the second column. If the third column starts receiving what comes out from the second column, and the first column is done, we can put the first column into regeneration and use a fourth column which was being regenerated. We put the fourth column in the third position, the second becomes the first column and the first one gets regenerated. From left to right: “movement of the solid”. From the right to the left: movement of the liquid. It’s a counter current movement. Wash step after feeding COUNTER CURRENT FOR CAPTURE There are many different processes. The system we are going to consider is the one with two columns. 53 PROCESS PERFORMANCE: PRODUCTIVITY, LOAD/ CAPACITY UTILIZATION In all these chromatographic processes we must consider the performance: yield, productivity, purity (this last one is established and mandatory). The productivity is how much product is produced per unit time per unit column volume. We take one cycle, we calculate how much protein we can purify per unit time per column volume. High productivity means needing less time to purify a certain amount of feed. The load is the quantity of the protein that we load on the column. If we don’t use our column fully but stop the feed before in order to not lose the product, then the load has to be smaller. High load = good capacity utilization = more product per resin. Note: every time we have to go through regeneration. THE COLUMN UTILIZATION PRODUCTIVITY TRADE-OFF If we feed a very low flowrate , we have that the residence time is longer compared to the transport diffusion and transport characteristic time. With low flowrate we have steep breakthrough. With high flowrate the breakthrough is broad . This is important because, to keep the yield constant, we have to stop as soon as there is the feed’s breakthrough. 54 By making the breakthrough course much steeper we decrease the flow rate : when the breakthrough occurs, we are very close to saturation. We get a much better column utilization. Trade-off between column utilization and productivity in the one column model. With two columns or more I don’t care about the flow rate: it can be even high and the curve can be broad and not steep, since all the breakthrough is captured by the next column and nothing is wasted. Example: varying mAb profiles There are several possible degradations reactions that mAb can undergo: this creates different profiles (analytical chromatograms) of the drug. In the graphs we see the profiles of the drug as it is sold to the pharmacy. APPLICATION EXAMPLES PF COUNTER CURRENT CAPTURE Example 1: capture SMB (2c pcc) with protein a (simulation moving bed) In this picture we see the resin capacity utilization in Batch vs. Capture SMB. We see the concentration in the feed of monoclonal antibody as a function of eluted volume. The eluted volume is obtained by multiplying the time with volumetric flow rate of the buffer fed to the column to elute the mixture containing the mAbs out of the capture chromatography. By the looking at the graph we see that at first nothing comes out: all the antibody is captured inside the column. At some point, we have 1% Dynamic Bending Capacity (DBC is the full capacity of the column): at 1% we have to stop if we have a single column otherwise we will loose some Ab. If we stop, we then proceed with the regeneration step, where the amount of material is recovered once we elute out the mAbs that were kept in the column. 55 Instead, if we have a second column we can go on, and what comes out from the first column isn’t lost since it’s taken from the second column. In the breakthrough we have two quantities: one is captured by the first column and the second part is captured by the second column. Nothing is lost. Then we stop because we reach saturation of the second column. We never get the 100% saturation, but we stop at a certain X%. This is just to have a safety margin. So, in this image we see the amount of material that is kept by the first column, the amount of material that would be captured in the first column if we had a second one, and the amount of mAbs captured by the second column. Process animation of CaptureSMB (2C PCC) We start with a feed (mAb + supernatant). The mAbs come down. We place the feed and at a certain point we have the breakthrough. The mAbs first show up at the first column outlet. If we had only one column we should have stopped here. If we have a 2 nd column we can go on: what comes out from the 1 st column goes in the 2 nd one and the material is not lost  the yield remains high. We go on until the first column is saturated with mAbs (95% of saturation to keep a safety margin). At this point we stop, and we proceed with the wash. The wash is done with a buffer which doesn’t contain neither the mAbs or supernatant, and it’s used to clean the column from all the supernatant which is present mostly in the interparticle’s porosity. Once the wash is finished, we go to step 2. In step 2 we start eluting. To do this we change the pH, making it acidic. With low pH the ligand protein A releases the mAbs, which can then be removed from the column. The feed goes to the second column (it never stops). Then we get the eluted product and the second column keeps receiving the feed (the mAbs go further down the column). 56 Then, before we have the breakthrough on the 2 nd column, we enter the 3 rd step. The third step begins before the breakthrough in the second column. After we finish to elute the product we have to sanitize the first column : Cleaning In Place (CIP). To do this we feed a 1mol sodium hydroxide , which is a very strong mixture, capable to completely destroy everything that was still present in the column. Then, we re-equilibrate the column with a buffe r so that the column is ready again to receive the feed. Note: all of this has to happen before we have the breakthrough from the second column. Once the cleaning step is done we re-connect the two columns and we keep on feeding the second column. Now, when the breakthrough goes from the second column the product goes to the first column. Basically, the first column gets the role of the second one, and everything is repeated. MATERIALS AND METHODS We’re talking about analytical methods use by analyzing the protein structure. The UV signal is taken in between the 2 columns. The operation is periodic. 57 RESULTS: BATCH VS 2 COLUMN PCC There are two sensors: one is after the first column and the other is after the second column. In red and blue is displayed the outlet of the two columns (UV signal), which behave in the same way since it’s a periodic continuous operation. The first column gets fed and second one gets regenerated, and after a little while the opposite happens. Performances comparison: Batch VS Continuous (capture SMB) Yield: by increasing the flow rate we always recover a lot of product in both processes. There isn’t much difference between the methods because in the batch we stop as soon as there is the breakthrough , meanwhile in the capture SMB we continue until saturation of the 1 st column and proceed until the breakthrough of the 2 nd one. We do not loose any product in both cases. Load/Cycle: if we use 2 columns, we obtain higher loading and higher column utilization. The higher yield is due to the larger amount of column utilization. In fact, with two columns we produce a larger amount of monoclonal antibody per cycle. Note: there is a limited number of cycles (maximum 100) because the sanitation processes damage a bit the column. The more the product we can recover the better it is. Productivity: it increases for both batch and continuous if we increase the linear flow rate. We pump more feed, but the productivity increases more for continuous. Buffer consumption: if we produce more per unit time, we need to produce less buffer. The buffer must be purified after its use, and this process has a cost ! Therefore, it’s more advantageous to have a continuous process. 58 Product quality HCP : Host Cell Protein. Aggregates, HCO, DNA content: quality of the product : how pure the protein is? There isn’t much difference between the 2 processes. The biggest difference between the processes is in the yield and the column utilization. There aren’t any differences in the product quality parameters. Economic evaluation: scale up model for resin costs 2 velocities linear in the column and 3 possible processes. The economic evaluation depends on how much material we produce. There are three typical situations: 1. Proof of concept, which is when the process is checked to get to a certain productivity. 2. Phase 3: we must produce the amount of material for clinical test. 3. Commercial: larger amount The Improvement we get is different depending upon the amount of material we want to produce because there is a scale effect. Advantage from going to Batch technology to more advantageous one. 59 Example 2 : model based optimization of multi column capture Base-case and constraints Capture Chromatography – Process Design Typically, we take some breakthrough curves. In this case we used two column lengths, 3 values of the flow rate and 3 values of supernatant. The number of experiments was between 10 to 20 to tune the parameters of the model, then we can calibrate the model and finally make comparisons in the process performance. 60 Comparisons of various capture processes Compare a batch (single column) or 2 or 3 or 4 columns  we can predict the performances by using the columns. We can use a single column or we can use two to three columns: there are some differences. But there’s a large difference between the batch and continuous operation. Smaller difference between the different continuous processes. Parator: it’s a plot in which we have summarized more than one parameter which is useful to evaluate the performance. It’s optimal when we have objectives which are contradictory, such as productivity and column utilization (CU). When we use the parator we can see all the optimal points: which one of our objectives is achieved and which one is not  we do not have a single maximum. Yield is always constant and also the product quality is the same  in this plot we can decide if we want to have higher productivity or higher column utilization or if we want to stay in the middle Another comparison in terms of buffer consumption vs productivity. Protein A capture optimization 61 LECTURE 6 BIND AND ELUTE (POLISHING) CHROMATOGRAPHY OUTLINE - Process fundamentals and design, - Comparisons of single and multicolumn processes, - Application example. THE POLISHING STEP We can have polishing with two different modalities (each one in batch or continuous): 1. Bind and Elute: we load the mixture that we want to separate. Here we assume that we want to separate 3 components: red is the product and blue and green are the weak and strong impurities which elute respectively before or after the product. We work with the bind and elute mode: we consider the effect of the modifier on the molecules. We start in the load with a very low concentration of the modifier. All the components adsorb very strongly on the solid. Then we start changing the concentration in the feed of the modifier (ex. Salt in IE). By increasing the modifier concentration, first the less absorbable components will reach a modifier concentration specific for the de-adsorption and then, by increasing a little bit the modifier’s concentration, even the stronger components will de-adsorb. So, by playing with the slope we can separate the different components in a better or worse way. When we hit the value of the modifier that de-adsorbs our component, the component leaves the stationary phase, starts moving along the column and arrives to the outlet. We can say that the performance of the column is related at how steep is the gradient: - Very stiff gradient of the modifier  very different components will come out more or less at the same time  peaks at the outlet of the column are very close. - Low stiffness of the gradient, shallow gradient  very different components will come out at very different times  much separated. Note : in the capture process the quality of the separation was related to the ligand!! In that case it depends on the stationary phase, not on the concentration of the modifier. We can do the bind and elute in a batch or with the continuous mode. 62 2. Flow- through mode : in this case we do not have a gradient, but we flow through the column the capture material (material which leaves the capture step). Here typically we want to remove some DNA, agg