unwanted membrane particles in filtration

Membrane filtration is a common way to remove pathogens from materials in which sterility is required.   The hypothesis tested in this report was that

  1. Pieces of membrane may beak off the membrane during the filtration process.
  2. The membrane pieces may seed aggregation of a therapeutic protein particularly prone to aggregation, Keratinocyte Growth Factor.

Filtration through conventional membranes may be stressful

Fluid flow through these filters can sometimes be problematic.   A helpful introduction is a slide player review of drag in fluid dynamics.  Liu and coworkers (2012) tested the hypothesis that particles sloughed off commercial filters could seed aggregation of a protein that typically sterilized with syringe filters.  The tested filter materials included polyether sulfone (PES), polyvinylidene fluoride (PVDF), and cellulose acetate (CA).  Glass microfibers are typically used as a prefilter to remove large particles.


PVDF, PES, and cellulose acetate all appear to have convoluted pores.  One would expect  that the flow through these pores would be turbulent rather than laminar.

Methods used


Liu and coworkers counted particles > 1 μm in diameter with  flow microscopy.  A zetasizer was used to measure particle size distribution.  Interestingly, this device is also used to assess purity of protein crystals in X-ray crystallography.  Filter rinses with and without the test protein were not reported in the main publication.   In terms of building a better filter sterilization system, such results may have been interesting.  Their main hypothesis, from which they did not stray, was that pieces of membrane break off and seed protein aggregate formation.

A 10 mM sodium citrate buffer, pH 6.2, was used in this study.  This buffer was filtered through a Whatman 0.2 μm nylon filter.  The authors claimed that nylon filters remove particles from the buffer without shedding them.   Many times buffer is flushed through a filter prior to filtering protein solutions.  This is to flush out any impurities that might affect the protein solution being filtered.

This image is derived from Figures 1 and 2 of the publication.  Each experiment was performed in replicates of three.  Rounded red squares indicate qualitative differences in with and without pre-flushing.  The blue squares are meant to draw our attention to differences in Y-axis scales.    Except for GE PES, the glass fiber prefilters do not lower the amount of particulates the the nylon membrane pre-filtered citrate buffer.

Keratinocyte growth factor

KGF, aka fibroblast growth factor 7,  is produced by mesenchymal cell in response to injury (Yen 2013).  In a paracrine manner, it binds to the  FGF receptor 2  on nearby  epithelial cells, including epidermal keratinocytes, intestinal epithelial cells, and hepatocytes.  Another review (Finch 2013) covers KGF function and potential therapies.

KGF basic research
– Control of epithelial cell-specific FGFR2b expression
– KGF/FGFR2b signalling mechanisms
– FGFR2b signalling in tumourigenesis and tumour suppression
– KGF family members as pre-synaptic organizing molecules
Pre-clinical models of KGF protective and regenerative effects
– Palliative care in the oncology setting
– Intestinal mucosal injury
– Graft-versus-host disease (GVHD) and immune reconstitution
– Pulmonary disease
– Dermatological applications
– Urothelium
– Pericardium

Aggregation of recombinate KGF

Aggregation of recombinate human  KGF (rhKGF) has been described by Chen (1994).  Elevated temperature was proposed to result in conformational changes that result in disulfide bonds between monomers.  These aggregates are initially soluble, but can grow into precipitates.   Stabilizers include citrate buffer, heparin, and  anionic polymers.  These molecules were proposed to act by reducing charge repulsion on the protein.

Membrane particle and KGF interactions

Liu and coworkers discussed hydrophobic regions of their test protein KGF binding to the filters in their introduction.  They also mentioned coating PVDF and PES with compounds to make the filters more hydrophilic.    The RCSB database of protein structure allows one to create hydrophobicity surface plots of proteins for which there are crystal structures.  In this rendering aromatic  (hydrophobic) amino acids such as tyrosine and tryptophan are dark green.  Charged amino acids lysine and glutamate are rendered in red.


Figure 3 was labeled as “without glass fiber prefilters, but was probably a mistake.  The authors mentioned substantial loss of KGF on filters with the glass fiber prefilters.


Aggregation takes time, and agitation

KGF concentrates apparently do not go “bad” right after filter sterilization.  This process takes time.  From the work of Chen (1994) we know that disulfide cross liniking of recombinant human KGF is part of the process.  The positively charged heparin binding sites may also be part of the aggregation process.  The soluble aggregates may grow so large that they precipitate from solution.  Citrate can greatly decrease the rate of aggregation of KGF at elevated temperatures.

Narrowing the list of filters to study sterilization induced KGF aggregation

  • PES showed little protein absorption to filters (data not shown).
  • The particles leaching from Whatman, Millipore and GE PES filers varied greatly. this is a convenient way to generate three different particle loads.

KGF aggregates may arise three different ways

  • Protein-protein interactions which may be removed by centrifugation. Centrifuge KGF to remove seed aggregates and then let incubate.  The >1μ m particles should be fewer in number.  
  • Pieces of membrane that bind to KGF during the filtration process and are ripped out of the membrane as a unit.  Filtering the KGF through a low binding PES filter should increase the number of  >1 μm particles.
  • Pieces of membrane encounter and bind after filtration.  Filter  citrate buffer and add that buffer, with particles from the filter, to solutions of KGF that have not been filter sterilized.  There should be an increase in  >1μ m particles


A visual and some data

Liu (1994) made mention of the KGF’s isoelectric point being pI 9.9.  The pI is the pH at which a substance as a net charge of zero.   Interestingly, the pI of silica is listed as 1.7-3.5.  This means that at pH 6.2 the silica (glass fibers) carry a negative charge and the KGF a positive charge.  Liu (1994) noted a significant protein loss on the glass pre-filters meant to prevent clogging of the 0.22 μm filters to remove pathogens.

The same KGF structure was colored according to atom type.  Blue is nitrogen, red is oxygen, and yellow is sulfur.  Disulfide bonds are mediated by  S-S bonds in cysteine side chains.  The hypothetical aggregate was produced by lining up red and blue regions of the surface of the monomer and is strictly hypothetical.


The increase in > 1  μm particles after one day

  • Note the scale of the Y-axis ranges from 0 to 3,000,000 particles per mL after one day of storage.  Except for GE PES filters, the range of particles per mL was up to 4,000.  Centrifugation seems to help, but there still seems to be about 300,000 particles per mL.
  • Agitation increases the number of particles in at least some of the samples.  The error is so large in these experiments; definitive statements are hard to make.  One can imagine that agitation simply speeds up the encounter of seed aggregates with monomers.
  • Again, the error bars are really large.

Cavitation, the elephant in the filter

Cavitation is the formation of vapor cavities in a liquid due regions of negative pressure.  These vapor  cavities are steam, water in the gas phase.  Water boils at a lower temperature on a high mountain top than it does at sea level.   Inconsistencies in traditional filter matrices can cause regional vacuums.  The pressure drops may be magnified when the biologist becomes impatient and applies too much pressure when forcing fluid through the filter.   Further damage may occur when these vapor bubbles burst.  Some of the large error seen in the Liu study may have been due to variations in the amount of pressure applied to filter the solutions.


These are screen captures taken from a YouTube video.  Initially, fluid flows easily through the filer.  As the process continues, massive amounts of foam may be generated.  The impatient biologist convinces him or herself that these bubbles came exclusively from the atmosphere.   Either way, they are destructive, as an elephant on a freshly planted field, to your purified protein.

Laminar flow membranes a solution?

NANOPEC has a method of fabricating alumina membranes with  regularly spaced, straight pores.  The low pI of silica glass  fibers proved to be a problem for KGF,  the pI of alumina is 7-8.  Metal oxide materials may be passivated with oxyanions to reduce binding of many different solutes.


NANOPEC’s pores with smooth sides greatly simplifies matters  of flow to Poiseuille’s law.  Naturally, the flow per unit area depends on the number of pores per unit area.


Chen BL, Arakawa T, Morris CF, Kenney WC, Wells CM, Pitt CG.(1994) Aggregation pathway of recombinant human keratinocyte growth factor and its stabilization. Pharm Res. 11(11):1581-7.

Finch PW, Mark Cross LJ, McAuley DF, Farrell CL. (2013) Palifermin for the protection and regeneration of epithelial tissues following injury: new findings in basic research and pre-clinical models. J Cell Mol Med.17(9):1065-87.

Liu L, Randolph TW, Carpenter JF.(2012) Particles shed from syringe filters and their effects on agitation-induced protein aggregation. J Pharm Sci. 101(8):2952-9.

Yen TT, Thao DT, Thuoc TL.(2014) An overview on keratinocyte growth factor: from the molecular properties to clinical applications. Protein Pept Lett.21(3):306-17.

Leave a Reply