Product-Contaminant Complexes How to recognize them How to

www.validated.com PSG–100912

Product-Contaminant Complexes How to recognize them

How to manage them"

Pete Gagnon, Validated Biosystems

Bio Process Network Conference, Melbourne, October 12-14, 2010


Product-contaminant complexes

Purification process development commonly begins with an assumption that the product and contaminants are separate, wholly independent entities.

Recent publications reveal that this assumption is invalid, probably always, and often to a high degree.

Host cell protein clearance during protein A chromatography: development of an improved column wash step, A. Shukla, P. Hinckley, Biotechnol. Progr., 24 (2008) 1115-1121 Evicting hitchhiker antigens from purified antibodies, K. Luhrs, D. Harris, S. Summers, M. Parseghian, J. Chromatogr. B, 877 (2009) 1543–1552 Production of biobetter IgG with enhanced wash and elution of protein A. www.validated.com/revalbio/pdffiles/Biobetter.pdf IgM purification: the fine print. www.validated.com/revalbio/pdffiles/MSS10B.pdf Dissociation of naturally occurring antibody-contaminant complexes with hydroxyapatite. www.validated.com/revalbio/pdffiles/DcplxHA.pdf Chromatographic behavior of IgM:DNA complexes, P. Gagnon, F. Hensel, S. Lee, S. Zaidi, J. Chromatogr. A, JCA-10-1683


Product-contaminant complexes

This presentation will focus on IgM-DNA complexes.

IgM. First, because its unique ability to discriminate glycosylation variants makes it a compelling candidate in cancer therapy, AIDS vaccines, treatment of infectious disease, and differentiation of stem cells. Second, because it has a documented tendency to form stable complexes with DNA.

DNA. First, because of the clinical importance of achieving low levels for all in vivo diagnostic and therapeutic products. Second, because it is easy to detect and measure by dual-wavelength UV absorption. Third, because it has a documented tendency to form stable complexes with IgM.


Isolation of IgM-DNA complexes

Initial capture on hydroxyapatite

One absorbance unit of DNA at 254 nm corresponds to a concentration of about 50 µg/mL. One absorbance unit of IgM at 280 nm corresponds to about 850 µg/mL. When absorbance at 280 is equivalent to absorbance at 254, as roughly in the IgM peak, about 6% of the mass is DNA. See J. Glasel, Biotechniques 18 (1995) 62 for optical proportioning of DNA and protein.

CHTTM Type II, 40 µm 426 mL, 5 x 22 cm A: 10 mM NaPO4 pH 7 B: 500 mM NaPO4 pH 7 EQ: A Load: 8 L CCS Wash: A Elute: 20 CVLG A > B Clean: B Note the break indicated during the loading phase. The indicated elution volume applies only to the illustrated interval.



Fractogel® TMAE HiCap 118 mL, 2.6 x 22 cm 200 cm/hr A: 25 mM Tris, pH 8 B: A + 1 M NaCl, pH 8 EQ: A Load: 1 L HA1 Wash: A Elute: 10 CVLG A > B

The leading shoulder contains host cell protein contaminants left over from the HA capture step. AX1 contains IgM that is relatively free of DNA, making the important point that only a small proportion of the IgM is heavily complexed with DNA. The contents of fractions AX2 and AX3 are IgM:DNA complexes. They were pooled for subsequent experiments.

Anion exchange fractionation of pool HA1



Fractogel TMAE HiCap 1 mL prepacked 200 cm/hr A: 25 mM Tris, pH 8 B: A + 1 M NaCl EQ: A Load: 10 mL HA2 Wash: A Elute: 10 CVLG A > B

Uncomplexed IgM is virtually absent from HA2. This validates exclusion of HA2 from further processing but highlights the point that complexation causes product loss. The loss however, is not as bad as it appears. Per equivalent mass unit, DNA absorbs about 10 times more UV at 280 nm than protein (assuming a protein extinction coefficient of 1).

Anion exchange fractionation of pool HA2


Characterization of complexes

Reduced SDS-PAGE of HA and AX fractions

LC

HC

HA1 AX1 HA2 AX2 AX3 AX4 5-15% gradient gel Coomassie brilliant blue. All AX fractions are derived from HA1 Note the contaminating protein bands in AX2 and AX3. This cautions that complexes include more than just antibodies and DNA. Histones are likely constituents but any solute able to bind nonspecifically to IgM or DNA is a potential candidate. This elevates the priority for complex-dissociation or removal.



Size distribution of IgM:DNA complexes TSKgel® PWXL 5000 7.8 x 300 mm, 0.25 mL/min 200 mM arginine, 300 mM NaCl, 50 mM MES, pH 6.5 Inject 100 µL TMAE pool The red vertical dashed line indicates elution volume at the peak center for DNA-free IgM.

qPCR and micro-BCA indicate that 24% of the complex mass is DNA. This amounts to about 230 kDa, or a total of about 350 base pairs.

The dominance of the 254 trace indicates that most of the DNA co-elutes with native IgM. This suggests that complexed DNA consists of small fragments. Identical elution volume with native IgM indicates that complexed DNA fragments are probably complexed between the pentamer arms and/or perpendicular to the radius of the pentamer disk.



The 254/280 ratio of the very large aggregate peak in the no-2° wash eluate shows the presence of ~50,000 ppm DNA. Very large aggregates disproportionately increase patient risk and are a key focus of regulatory agencies. Note that even the monomer peak carries an elevated DNA load (no-2° wash).

Protein A-purified IgG, with and without a secondary wash TSKgel® PWXL 5000 7.8 x 300 mm, 0.25 mL/min 200 mM arginine, 50 mM MES, 5 mM EDTA, pH 6.5 Inj. vol. 100 µL 254 nm was chosen for DNA instead of 260 because it coincides with a protein absorbance minimum that permits better differentiation of DNA from protein. The 254/280 ratio for pure IgG is about 0.45. The ratio for pure DNA is about 2.0


Chromatographic behavior

How does complexation affect affinity chromatography? CaptureSelect® anti-lambda light chain. EQ: 50 mM Hepes, 100 mM NaCl, pH 7 (HBS) Load: TMAE complexes, NaCl added to ~1.5 M Wash1: HBS Wash2: 2M urea, 1.5 M NaCl, 10 mM EDTA, pH 7 Wash3: HBS Elute: 500 mM arginine, 100 mM MES, pH 5.5

Complexed DNA apparently blocks immunorecognition by the affinity ligand, which means it could have similar consequences for antibody quantitation by ELISA. Experiments with the sample at 0.3 or 3.0 M NaCl produced essentially the same results. The 254/280 ratio of the eluate shows the NaCl/urea wash was able to reduce DNA to a low level, but as with 2° washes on protein A, was unable to completely eliminate it. Note that urea depresses apparent conductivity, and substantially alters the buffer baseline.



Toso PPG HW75 Post-TMAE complexes in 3.2 M NaCl Equilibrate in 50 mM Hepes, 1 M ammonium sulfate, pH 7 Elute 10 CVLG to 50 mM Hepes, pH 7

Flow-through fractions are DNA dominant. Binding fractions appear to be IgM dominant.

Despite the sample having been applied in 3.2 M NaCl, substantial UV absorbance was unretained. Note that the 254/280 ratio diminishes across the flow-through. The 254/280 ratio of the leading shoulder on the elution peak also indicates the presence of complexes. DNA apparently reduces surface hydrophobicity of complexes, which is consistent with the inability of DNA to bind weakly hydrophobic media like PPG.

How does complexation affect HIC?



CIM® SO3, 4 mL/min A: 20 mM Hepes, pH 7 B: A + 1 M NaCl EQ: 100% A Load: TMAE complexes Wash: 100% A Elute: 60 CVLG to 50%B Clean: 100% B

All of the flow-through and the leading side of the elution peak are complexed.

The flow-through spike at the end of the sample load coincides with a decrease in conductivity. This suggests that the spike is caused by increased charge repulsion between exchanger SO3 groups and negatively charged phosphates on DNA fragments complexed to the surface of the IgM. Product recovery is obviously poor.

How does complexation affect cation exchange?



CIM SO3, 4 mL/min A: 20 mM MES, pH 6 B: A + 1 M NaCl EQ: 100% A Load: TMAE complexes Wash: 100% A Elute: 60 CVLG to 50%B Clean: 100% B Note that EDTA absorbs UV at 280 nm and more at 254. Inset bars show relative UV absorbance at 254 nm (red) and 280 nm (blue).

The flow-through and load-end spike is smaller at pH 6. This makes sense. Proteins are more electropositive at lower pH values, which would ameliorate charge repellancy between the exchanger and DNA phosphates. Sample treatment with EDTA causes a substantial change: more 280 absorbance, less 254, and a cleaner sharper peak, all of which suggest decomplexation but raise the question: by what mechanism?

How does complexation affect cation exchange?


Complexation mechanisms

Why, how does EDTA affect DNA complexation with IgM?

The conductivity increase from 10 mM EDTA seems unlikely to significantly dissociate electrostatic interactions, especially when 3.2 M NaCl doesn’t.

Given the obvious implication that metal affinity is involved, the most likely possibility is that EDTA dissociates metal-bridged complexes: IgM-metal-DNA.



Proof of principle for metal-bridge complexation comes from the field of immobilized metal affinity chromatography (IMAC).

Example: Dicarboxyl (iminodiacetic acid) binds Iron III. The immobilized iron coordinates with endotoxin phosphates. This method has been used for endotoxin removal.

IgM carboxyl clusters could equally bind iron (or other metals), and the metal simultaneously bind DNA phosphates.

IgM is documented to bind HA calcium even at 2 M NaCl. HA calcium forms equally strong coordination bonds with DNA phosphate, and also tolerates high salt.



Metal-bridged complexes have characteristics that should favor cation exchange binding over charge complexes. Protein amino groups would be uninvolved, leaving the full positive charge of the protein available for binding the cation exchanger. The positive charge on the metal ion would neutralize some of the negative charge on carboxyl groups with which they coordinate, and reduce repellancy from the exchanger. DNA image modified from Wikipedia



How does complexation affect anion exchange?

TMAE merely reproduces the complex distribution injected onto it. QA dissociates the majority of complexes into their individual components. EDA dissociates complexes almost completely, plus the separation is more than double that achieved by TMAE.

TMAE HC, 1 mL, 5x50 mm CIM QA & CIM EDA, 1 mL All at 1 mL/min A: 50 mM Hepes, pH 7 B: A + 2 M NaCl EQ: 100% A Load: TMAE complexes Wash: 100% A Elute: 20 CVLG to 100%B

EL/ms IgM DNA Delta TMAE 23.7 54.8 31.0 QA 25.8 67.0 41.2 EDA 43.3 117.7 74.8


Decomplexation mechanisms

Why does anion exchange support more effective fractionation and decomplexation than other methods?

Anion exchange is the only method that preferentially binds DNA. This suggests the possibility that it competitively dissociates DNA from bound complexes.*

If this is true, then chemical environments that weaken complexation should enhance dissociation.

The mirror hypothesis, competitive displacement of IgM on methods than preferentially bind IgM, lacks a key enabling feature: DNA is extremely charge dense with no contrary charges. It’s inherent affinity for a high charge density ion exchange surface will be much greater than IgM for a cation exchange support.

*



CIM QA, 4 mL/min Buffer A: 20 mM Tris, pH 8 Buffer B: A + 1 M NaCl EQ: buffer A Load: TMAE complexes Wash1: buffer A Wash2: A + 4 M urea Wash3: buffer A Elute: 60 CVLG to buffer B Urea is a nonionic hydrogen donor/acceptor. As such it can suspend hydrogen bonds but not interfere with electrostatic interactions.

The influence of hydrogen bonding on anion exchange

Loss of hydrogen bonds weakens the association between IgM and DNA. Applying the wash at low conductivity conserves the full strength of the anion exchange groups to competitively dissociate DNA. The combination enhances dissociation efficiency. This supports the notion that AX works by competitive displacement of DNA from IgM.



Why do monolithic anion exchangers support more effective decomplexation than porous particle exchangers?

Binding capacity of large DNA molecules is 10-40 times higher on monoliths than on porous particles, but complexed DNA appears to be in the form of small fragments, so this is an unlikely explanation. Differences in the efficiency of diffusive versus convective transport seem unlikely for the same reason.

However, monoliths have higher ligand density than most porous particle exchangers, and this may enhance their ability to competitively dissociate DNA.

TMAE: 0.22 +/- 0.04 meq/mL QA: 1.0 +/- 0.1 EDA: 1.6 +/- 0.1



Why does EDA decomplex more effectively than QA?

QA bears a single quaternary amine (strong anion exchange group). EDA bears a primary and a secondary amine (two weak anion exchange groups). Adjusting for the higher ligand density of EDA, this comes to 3.2x higher charge density. In addition, less substituted amines such as those on EDA have a higher affinity for DNA. This is believed to result from their stronger hydrogen bonding potential.



CIM EDA, 1 mL, 4 mL/min A: 50 mM Hepes, pH 7 B: A + 2 M NaCl EQ: A Load: TMAE complexes, 0.1 mg/mL, volume as indicated Wash: A Elute: 10 min LG to B Note that the monolith gives greater dissociation with a 20 mg load than the porous particle exchanger gives with 1% of that load (slide 17).

These results show that decomplexation requires excess capacity, which confirms that complexation reduces usable capacity. It also reduces recovery of decomplexed IgM. To optimize both, reduce complex levels as much as possible in previous steps. Mechanistically, these results suggest that DNA reduces the charge potential on the surface of the exchanger, gradually reducing its decomplexation ability.

Does complexation influence usable column capacity?


DBC of decomplexed IgM

All samples were loaded by in-line dilution, with an over-buffered diluent to achieve the target pH. For example, the pH 6 set was initially buffered with 20 mM Hepes, pH 7, and diluted through the chromatograph with 100 mM MES, pH 6. Final conductivity was determined by the in-line dilution factor. Flow rate for monolith experiments was 4 mL/min; for TMAE, 1 mL/min.

Lower capacity at pH 6 is logical, given that proteins are more electropositive at low pH. This should create a degree of capacity-reducing charge repulsion. By the same model however, capacity should have been highest at pH 8. A complete explanation for the discrepancy is presently lacking but may involve the solubility characteristics of this particular IgM. IgMs commonly exhibit limited solubility at fairly moderate pH values, such as pH 6 or pH 8. Note that lower conductivity increased capacity, but it caused solubility and backpressure problems.


Conclusions

Electrostatic interactions Hydrophobic

interactions

Metal affinity

Hydrogen bonding

Protein (beta sheet)

Contaminant

DNA forms stable complexes with IgM via electrostatic interactions, hydrogen bonding, and metal affinity. To the extent that complexes also contain other proteins, additional opportunities exist for hydrophobic interactions. These mechanisms likely work cooperatively, accounting for the remarkable stability of contaminant complexes.


Conclusions

Contaminant complexes depress the ability of every purification method to achieve its goals.

They can prevent binding to CX, HIC, and affinity media, causing product losses on loading.

They reduce usable binding capacity and decomplexation efficiency on AX. They are invisible to SEC.

They depress product recovery, through discarding complex fractions, especially on porous particle resins.

They act as Trojan Horses, transporting suites of otherwise unrelated contaminants straight through the gates of multistep purification procedures.


Conclusions

Early detection is essential to effective complex management.

Multiple wavelength detection is effortless and provides a good first approximation of the proportion of DNA across a chromatogram.

Beyond that, look for unexpected product in the flow-through of CX, HIC, or affinity; or in the high salt strip of AX.

If complexation is confirmed, be prepared to accept that it may become the dominant consideration in overall process design. Reduce DNA levels to the lowest level possible, as early as possible in a process to ensure that later methods can achieve clinical target levels.


Conclusions

Decomplexation can be enhanced to varying degrees by tactical application of dissociating agents.

Use EDTA to suspend metal affinity stabilization. This can help with affinity, HIC, and CX, but not AX because EDTA binds to AX columns.

Use urea to weaken or suspend hydrogen bonding and hydrophobic interactions. This can help with all methods except HIC. It is especially effective in conjunction with AX, particularly on monoliths.

Use high salt concentrations to weaken electrostatic interactions. This can help with HIC and affinity, but not ion exchange.


Conclusions

Restrict decomplexing agents to a pre-elution wash if you can.

Include them in the elution only when necessary since they will certainly complicate gradient development and may interfere with subsequent processing steps, requiring addition of a diafiltration step to the overall process.

Think carefully about including decomplexing agents in the sample since they may interfere with binding. Salt, for example will interfere with ion exchange, urea with HIC.

If a decomplexing agent enhances sample binding, for example salt with HIC, apply as much as provides a worthwhile result. Equally if the agent is neutral, like EDTA with HIC. Expect combinations to be more effective than single agents.


Conclusions

Keep in mind that DNA may provide a useful marker and model for other important contaminant classes:

Endotoxins are likewise heavily phosphorylated, as are lipid enveloped viruses. Decomplexing treatments that reduce DNA are likely to reduce these other classes.

On the other side of the coin, keep in mind that heavy DNA complexation may be a warning that a given product has an elevated affinity for endotoxin. Some proteins are able to scavenge endotoxins from buffers, thereby concentrating them and producing mysteriously high endotoxin levels in previously clean product.

Documents

Product-Contaminant Complexes How to recognize them How to