Effect of flux and shear rate on E. coli recovery in tangential flow filtration through a single hollow fiber

Pathogenic bacteria which enter a viable but non‐culturable (VBNC) state impede efforts to reach detectable concentrations required for PCR methods. This motivated a strategy for tangential flow filtration to concentrate bacteria in aqueous samples while maintaining the bacteria in a viable state, maximizing their recovery and achieving high fluxes through a single hollow fiber membrane. Filtrations were carried out for green fluorescent protein (GFP) E. coli at high shear rates (up to 27,000 sec−1) through 0.2 μm cut‐off polyethersulfone (PES) microfilter membranes or 50 kDa polysulfone (PS) ultrafilter membranes. High shear minimized bacterial attachment on membrane surfaces, which would otherwise occur due to forced convection of the particles to the membrane surface at high flux conditions. Single fiber filter modules were constructed to facilitate concentration of Escherichia coli at fluxes ranging from 55 to 4500 L m−2 h−1. The effect of high shear rates on bacterial viability was found to be minimal with bacterial losses during filtration caused principally by their accumulation on the membrane surface. Recoveries of 90% were achievable at high shear rates when the average flux was ≤300 L m−2 h−1. This corresponded to a 3‐h filtration time for a 225 mL sample through a single hollow fiber. Detectable bacteria concentrations of 1800 colony‐forming unit (CFU)/mL were achieved for starting concentrations of 140 CFU/mL.


| INTRODUCTION
More than 65 known pathogenic species of bacteria can enter a viable but non-culturable state (VBNC) when under environmental stressors. 1,2In the VBNC state, these bacteria will not multiply in enrichment media or on agar plates, but may still pose an infection risk if consumed in food. 1 With millions of Americans contracting foodborne illnesses annually, there is a need for methods which can concentrate these VBNC bacteria to levels that enable detection.[5] One strategy for concentrating bacteria, in the absence of enrichment media, is through tangential flow filtration (TFF) with utilization of crossflow along the surface of a filter membrane to reduce losses of materials/particulates at the membrane surface of the ultrafilter or microfilter.Without crossflow or backflush, previous work using TFF to concentrate bacteria had mixed results.7][8] Without these measures, <16% of sample bacteria were recovered from water samples by Peskoller et al. and <1% were recovered from some buffer-based samples by Zuponcic et al. 7,8 Bacterial removal from surfaces under shear flow was therefore investigated since previous studies for various microorganisms and surfaces showed that shear provided by crossflow could enhance their recovery. 9A range of wall shear rates prevented bacterial attachment in these studies, but generally rates greater than $10,000 s À1 prevented attachment of bacteria flowing along a surface. 10,11In one case, shear rates >12,000 s À1 were capable of removing adhered Pseudomonas fluorescens from stainless steel. 10In another case, wall shear rates >16,000 s À1 were capable of removing more than 90% of Bacillus cereus on glass surfaces after 5 min. 12A review of the literature on TFF concentration of bacteria revealed a trend wherein higher recoveries occurred at higher shear rates (Table 1).However, membrane surface areas and flux rates also varied widely and hence, it remained unclear to what extent shear rate, flux rate, and surface area affected bacterial accumulation on the membrane surface.The research reported here was designed to answer this question over a wide range of shear rate and transmembrane flux as well as to seek conditions that kept filtration time low.
Here, a novel approach to concentrating bacteria in TFF systems was explored by evaluating recovery of E. coli modified to express green fluorescent protein (GFP).We chose GFP E. coli instead of VBNC bacteria, since GFP E. coli may be directly observed using confocal microscopy, and since these bacteria could be cultured and counted to test for viability.In order to experimentally simulate VBNC conditions we suspended the E. coli in phosphate buffer saline (PBS) that lacked nutrients for cell growth, so that the bacteria remained viable, but did not multiply during the course of a 3.5 h experiment.
Accumulation patterns of the GFP E. coli on single hollow fibers, maintaining the same surface area for each fiber, was measured at high shear rates (27,000 s À1 ; Re = 850, laminar flow) at fluxes of 55 to 4500 L m À2 h À1 , respectively.The objective was to determine filtration conditions which achieve high bacterial recoveries while maximizing flux (such that total concentration time is minimized).We hypothesized, and later confirmed, that the applied shear rates and pressures would not be sufficient to cause loss of bacterial viability, with the main cause of bacterial losses, if they occur, being due to their accumulation on the membrane surface.It was also hypothesized, and later confirmed, that increasing flux would increase losses of bacteria (decrease recovery) because of increased convection of bacteria to the filter surface.Understanding how these losses could be mitigated by high shear rates has the potential to enable concentration, and therefore detection, of dilute levels of VBNC microorganisms, as well as applications to concentration and recovery of other micron and submicron sized particles.
T A B L E 1 Literature comparison of hollow fiber filter module configurations, shear rates, fluxes, and recoveries.

| Single fiber membrane module construction
Given the wide range of transmembrane fluxes to be evaluated in this research, two membrane types were tested while keeping constant surface area: a single hollow fiber of 0.2 μm pore size polyethersulfone (PES) microfilter (sourced from product D02-P20U-05-N, Spectrum Labs) and a single hollow fiber of 50 kDa MWCO polysulfone (PS) ultrafilter (sourced from product X15S-300-04S, Spectrum Labs).Both membranes had same fiber inner diameter (ID = 0.5 mm), and modules were constructed with membranes of same final length (14.5 cm), as indicated in providing outlets for permeate flow.The final length of the assembled housing was 14.5 cm.After assembly, a single hollow fiber was threaded down the length of the housing and glued in place using 5-min quick set epoxy (product 1,395,391, Loctite).The epoxy was applied around the fiber and within the opening of the housing to form a seal.The epoxy was left to cure overnight after which excess fiber was removed with a razor blade.Prior to concentrating bacteria, each module was evaluated for water permeability at various transmembrane pressures (TMP) as shown in Figure 2.
Clean, undamaged single fibers demonstrated a linear TMP versus flux relationship, similar to that observed with unmodified commercial hollow fiber membranes (Figure 2).Modules which deviated from this linear behavior were discarded.

| Filtration system setup
Assembly of the filtration system is shown in Figure 3.It required a conical vessel with a male luer lock outlet, precision peristaltic tubing F I G U R E 3 Single hollow fiber filtration system consisting of (left to right) sample reservoir, pump, pressure sensors, and single hollow fiber assembly.
(Mettler Toledo, 30,113,829, RS-232 interface with Excel) was used to measure permeate weight throughout the experiment, and data was recorded every 5 s.Permeate flow rates were calculated by weight differences over one-minute intervals, and from that, permeate flux was calculated according to the membrane filtration area.

| Filtration system operation
Prior to concentrating any sample, sample holder and tubing were autoclaved at 121 for 30 min.Once the filtration system was set-up, 70% ethanol (v/v) was pumped through all tubing, fittings, and the filter module-keeping a contact time of 5 min.Following the ethanol flush, the system was flushed with sterile deionized water, and then flushed with phosphate-buffer saline (PBS).
For concentration, samples were poured into the sample holder vessel, and the peristaltic pump was started at its lowest setting (8 mL/min) and ramped up to 20 mL/min within 2 min.High transmembrane pressures were achieved by applying a clamp to the retentate tubing after the retentate sensor.When retentate sample volume was lower than 20 mL, it was collected directly from the retentate outlet tubing into 50 mL sterile centrifuge tubes for analysis.Final sample volumes ranged from 5 to 20 mL.Filtration times ranged from 15 min to more than 12 h depending on the membrane type and pressures applied.During filtration, the pressure remained relatively steady (within $5 psi); however, occasional pressure spikes did occur, predominately in the single fiber ultrafilters (50 kDa MWCO) rather than in the single fiber microfilters (0.2 μm pore).
The isolated pressure spikes (from 1 to 4 psi) in an ultrafiltration membrane can be seen in Figure 4d, and these might be caused by a transient membrane blockage by the bacteria being forced against membrane at high flux conditions (288 L m À2 h À1 ), with the bacteria then quickly removed from membrane's surface by the high shear rates of 27,000 sec À1 .
After each concentration run, the modules were set aside for microscopy analysis.After imaging, the tubing, sensors, and fittings were cleaned with 70% ethanol and sterile DI water, prior to storage.

| Sample preparation
E. coli cultured for 24 h was diluted in PBS to make a sample containing 2 log colony-forming unit per ml (CFU/mL) E. coli.Prior to concentration, a 10 mL aliquot was set aside to obtain an estimate of growth during the time period of the experiment.Starting volume for each concentration was 225 mL-this is the FDA Bacteriological Analytical Manual (BAM) recommended volume for a 25-gram food sample preparation. 16lection of PBS as the medium for these experiments provided an isotonic solution for the E. coli without providing a carbon source for growth.Without a carbon source, recoveries calculated from the filtration system did not need to be adjusted to account for growth (some filtrations required more than 12 h).

| Cell recovery calculation
Plating of the 10 mL sample aliquot was performed in triplicate to verify that E. coli growth did not occur in PBS over the course of the filtrations.Samples of this aliquot prior to and following the filtration were plated, then plate counts were compared.A paired t-test determined there was no significant growth in the PBS over the course of these filtrations (i.e., growth = 0 in Equation 1).Bacterial recovery was defined as the percent of sample bacteria that were eluted from the filtration system versus the starting CFU count (Equation 1):

| Fluorescence microscopy of filter surfaces
Following a filtration experiment, the module was detached from the filtration system, and 2 mm of the PEEK housing was cut from each side of the hollow fiber (HF) assembly to remove the glue seal.The housing was unscrewed to expose the hollow fiber membrane.
The fiber was cut into three 4 to 5 cm long sections, placed on glass slides, and labeled for flow direction and filter location (i.e., feed section: 4 to 5 cm HF membrane cut close feed inlet; middle section: piece of 4 to 5 cm membrane cut from the mid-length of HF; outlet section: piece of 4 to 5 cm membrane cut close to the HF retentate outlet).Fiber sections were coated in quickset epoxy (same adhesive used for HF housing assembly) and allowed to cure for 20 min.
Coated fiber pieces were cut lengthwise with a sharp precision knife to expose the membrane's lumen surface.
Images of the filter surface were taken on an inverted fluorescence microscope (Eclipse TE2000-U, Nikon) using a FITC filter set to excite and detect GFP (filter excitation at 480 nm, emission at 515 nm).Images were taken at 20Â magnification (Plan Fluor 20Â ELWD objective, Nikon) using a 2 s camera exposure time (Orca-Flash4.0LT+,model C11440, Hamamatsu).

| Single fiber membrane and flux performance
The filtration system comprised custom single fiber filter modules, sourced from commercially available fibers, placed within PEEK tubing and fittings (Figure 1).Two membrane types were studied: where γw in Equation ( 2) is the shear rate (1/s) at the fiber wall and dh is the hydraulic (inner) diameter of the fiber (cm).

| Pressures, flow rates, and shear stresses during concentration of E. coli
Pressures over the course of concentration were monitored at the inlet (P ret,in ) and outlets (P ret,out ) on the retentate and permeate sides of the module using sensors attached to a data acquisition system.
The average TMP (transmembrane pressure) during the filtration was calculated by Equation 3 as: where the permeate pressure (P permeate ) remained at atmospheric levels throughout all experiments.Changes in permeate weight were recorded over 1 min intervals and used to calculate permeate flux in a mass/time basis.
Samples comprised of GFP E. coli suspended in 225 mL PBS (target of 2 log CFU/mL initial concentration).Concentration proceeded at 20 mL/min inlet flow rates-corresponding to an inlet wall shear rate of 27,000 s À1 (Equation 2), for a Newtonian fluid with the viscosity of water at 20 C. This shear stress corresponded to 27 N/m 2 , with shear stress (τ) represented by Equation (4).
Shear stresses at this order of magnitude have been reported as sufficient to remove many species of bacteria from surfaces and prevent cell adhesion. 12,17Even though it has been hypothesized that shear forces to this order of magnitude can reduce bacterial losses to the membrane surface, perpendicular flux to the membrane in TFF carries the bacteria to the membrane surface, and needs to be considered as well.The impact of increasing flux on bacterial recovery at these high shear rates was tested by increasing pressure on the retentate side of the membrane by constricting the outlet retentate tubing with a clamp.
In these cases, higher TMP resulted in higher permeate

| RESULTS AND DISCUSSION
Comparison of microfiltration (MF) and ultrafiltration (UF) hollow fiber membranes showed high shear rates that corresponded to high cell recovery, and not surprisingly, fluxes that were 10Â higher for the MF membrane than for the UF membrane (compare Figure 4a,b to Figure 4c,d).Most importantly, the results showed that the pressures (up to 25-30 psi) and shear rates (up to 27,000 sec À1 ) were not sufficient to decrease the viability of the bacteria being processed through HF microfiltration or ultrafiltration.Evidence that showed bacteria remained viable is based on absence of bacterial growth in PBS, and yet still high recoveries of viable microorganisms were achieved as indicated by plating of retentates (Figure 5).

| Pressure
Pressure had a significant impact on flux.In 0.2 μm (size cut-off) hollow fiber, small pressure differences of only a few psi of between $6 to 9 psi increased flux from 1618 L m À2 h À1 (Figure 4a) to 4549 L m À2 h À1 (Figure 4b), corresponding to permeate flowrates of 6 and 17 mL/min, respectively (Figure 4a,b).In contrast, for the 50 kDa ultrafilters, pressure changes higher than 10 psi were needed to increase flux from 55 to 288 L m À2 h À1 , corresponding to a permeate flow increase of only 1 mL/min (Figure 4c,d).The difference in the change in flux for microfiltration compared to ultrafiltration membranes is attributable to the differences in membrane permeability determined by pore size and overall porosity. 18Permeate pressure in all experiments were atmospheric.

| Shear rate effects
Shear rates at the module inlets were the same for all experiments because feed flow rate (20 mL/min) and fiber inner diameter (0.5 mm) remained constant.Differences in flux between the modules resulted in differing retentate outlet flow rates (therefore lowering shear rates and shear stress between the inlet and outlet).Outlet shear rates were lowest for the highest flux conditions-that is, microfilters under high pressure (Figure 4b).In the ultrafilter modules, flux was sufficiently small that outlet shear rates effectively matched inlet shear rates (Figure 4c,d).

| Flux effects
Decreasing permeate flux over time could, in general, indicate pore blockage occurring over the course of the filtration and/or formation of a protein layer on the membrane surface. 19During unclamped (low pressure and low flux) microfiltration experiments, the permeate flow rate declined linearly with total filtrate volume.1][22] Without a clamp on the retentate tubing, microfiltration was completed in 30-40 min.
Recoveries under these conditions ranged from 14% to 32% (n = 3; Figure 5a).In the microfilters, clamping the retentate tubing resulted in a significant change in retentate outlet pressures from nearly 0 to >4 psi (compare Figure 4a,b), resulting in a significant increase in flux when the retentate outlet pressure was higher.This result suggests that when the retentate tubing was not clamped (Figure 4a), an outlet portion of the membrane (under low pressure conditions) may not be contributing to overall filter flux, which was redistributed through the microfilter when retentate pressure was higher.This redistribution of flux would be manifested in decreased recovery.
It was hypothesized that higher fluxes would reduce recovery of bacteria because there is increased convection of fluid, carrying bacteria, to the membrane surface.At higher flux and pressure conditions, microfilter concentration runs were completed in 14-16 min (n = 4).
Recoveries at higher flux for microfiltration decreased by more than 95% to a range of 2%-5% (Figure 5a).These results provided evidence that flux may affect recovery even at high shear conditions when a microfiltration membrane is used.The opposite was observed for the ultrafiltration membrane, Figure 5b.
Given the results from the microfiltration modules, it was expected that a decrease pressure that also decreased the flux should also result in greater recoveries for ultrafiltration.However, recoveries actually dropped to 56% of initial cell titers for the unclamped ultrafilters when run times were up to 12 h.The cause of cell loss in this case was starvation of the microorganisms in the PBS over the >12-h filtrations, since PBS does not contain a carbon source or other microbial nutrients.This was confirmed in bacterial viability tests for retentate samples taken after 13 to 27 h.
Even though flux can affect bacterial recovery, evidence that high flux did not significantly impact the viability of the bacteria is given by microfiltrations carried out at high flux conditions (4549 L m À2 h À1 ) where the bacteria attached to the membrane (as visualized in Figure 6b) in a viable state (as indicated by GFP fluorescence).In addition, ultrafiltration at higher flux and high shear gave 93% recovery of viable bacteria (Figures 5b and 7c).

| Comparison of membranes
Concentrations using the 50 kDa ultrafilters (Figure 4c,d) had much lower average flux than those in the microfilters (Figure 4a,b).In ultrafilters, when applying pressures of 18 psi transmembrane pressure, the concentration run was completed in 3-3.2 h rather the 10-15 min obtained for in microfilters.Recoveries in ultrafilters increased from 67% to 93% as flux increased and run time decreased from 15 to 3 h (Figure 5b).The ability to recover high levels of E. coli in a viable state (capable of producing colonies on agar), in 50 kDa ultrafiltrations at both high pressure and high shear conditions, was evidence that the pressures and shear forces of the ultrafiltration were not sufficient to kill bacteria.These results confirmed the hypothesis that the filtration conditions studied in this work did not affect bacterial viability and that high shear rates remove retained particles when the ratio of particle size to membrane pore size or porosity exceeds 100Â.This conclusion is based on comparison of 0.2 μm microfiltration membranes with particle/pore ratio of 1/0.2 (or about 5) versus 50 kDa ultrafiltration membranes where the porosity cut-off coincides to about 5 to 10 nm giving a particle/porosity ratio of 100 or more.
Comparison between micro-and ultrafiltrations at high flux conditions, showed that for a microfilter at high flux (4549 L m À2 h À1 ), the outlet shear rates were lowest (shear rates decreased from 27,000 sec À1 at inlet to 7500 À1 to outlet, Figure 4b), while in the ultrafilter modules, overall flux was sufficiently small that reduction in retentate volume/ pass gave outlet shear rates that effectively matched inlet shear rates (Figure 4d).The combination of high shear and flux conditions (288 L m À2 h À1 ) achieved in the ultrafilter modules favored higher bacterial recovery among all module configurations evaluated.F I G U R E 6 Micrographs of hollow fiber membrane lumen interiors post-filtration.The left column shows the inlet region of the membrane, the middle column shows the center, and the right column shows the outlet regions of the fibers.E. coli appear as green dots on the filter surface.

| Filter surface microscopy
We used fluorescence microscopy to examine the inner surfaces of the hollow fiber membranes after E. coli concentration (Figure 6).We hypothesized most of the bacteria would collect near the inlet of the fiber where pressure, and thus flux, should be highest.Using FITC excitation and detection wavelengths, GFP-producing E. coli were visible as small dots at a 20Â magnification.
In microfilters at low pressures (Figure 6a) greater numbers of bacteria were visible near the center of the fiber versus the areas closest to the retentate inlet and outlet.At higher pressure and flux conditions in the microfilters (Figure 6b), greater numbers of CFUs collected at the fiber inlet and outlet regions-creating a more even distribution of bacteria down the length of the module.
Lower CFU accumulation near the entrance region could be explained by increased shear forces at the fiber wall due to nonuniform flow conditions.The entrance length for development of flow in these modules (Reynolds number = 850) was about 2 cm (corresponds to a normalized distance from the entrance of 0.14 as indicated in Figure 7a).Additionally, the pattern of accumulation in the middle region of the fiber was not constant.Rather, regions of the membrane containing dozens of CFU were immediately adjacent to several other sections containing only a few CFU.This was observed at both high-and low-pressure conditions (Figure 7a-c).This pattern corresponds to regions of locally higher flux in the hollow fiber due to nonhomogeneous membrane porosity and/or particle accumulation at the membrane surface.
For the 50 kDa ultrafiltration membrane where concentrations were carried out at high pressure, the recoveries were the highest and with only a few microorganisms visible along the entire length of the membrane (Figure 6d).Likewise, for the lower pressure ultrafiltration runs, almost no microorganisms were observed on the filter surface (Figures 6c and 7c) indicating absence of retention of microorganisms.
Of all tested conditions, the ultrafilters had the lowest average fluxes; and hence, lower transmembrane fluxes resulted in higher retentate outlet flow rates which in turn coincided with maintaining higher wall shear stresses along the entire length of the hollow fibers and corresponded to high recoveries of viable bacteria, that is, little or no accumulation of viable bacteria on the membrane (Figure 7c).Microscope images of dissected hollow fibers as a function of fractional distance from the inlet were consistent with recoveries determined by plating out retentate samples.The plate counts from the microfiltration hollow fiber membrane sections showed that significant numbers of cells were retained on the membrane along the length of the hollow fiber (Figure 7a,b).The microorganisms accumulated preferentially at a fractional distance of 0.5 at the lower flux (Figure 7a) and distributed more evenly along the hollow fiber length at the higher flux (Figure 7b).The accumulation was measurably larger at the higher flux (Figure 7b).

| CONCLUSIONS
The developed single hollow fiber housing enabled investigation of recovery of viable E. coli at high flux and high shear rate conditions.
Recoveries near 90% were achievable at high shear rates when average flux was 0.5 mL min À1 cm À2 (300 L m À2 h À1 ).This corresponded to a $3-h filtration time for a 225 mL sample.
The ability to recover >90% of E. coli in a viable state was evidence that the pressures and shear forces of the filtration were not sufficient to destroy bacteria.Filtrations producing the highest recoveries (67%-93%) displayed only a few CFU on the membrane surface post-filtration.Conversely, the microfiltration membranes producing the lowest recoveries (<5%) had the greatest CFU counts on the membrane surfaces.This confirms that lower recoveries coincide with retention of bacteria on the membranes.The results also show that the primary driver of decreased recovery from the filtration system was not cell death, but rather the increasing flux that led to bacterial retention on the membrane surface.
Although inlet shear forces were equal for all filtrations, hollow fiber ultrafiltration membranes had the advantage of allowing concentration to occur at both higher shear rate (27,000 sec À1 ) and low average flux (<300 L m À2 h À1 ), resulting in steady transmembrane pressure across the length of the filter.These conditions of higher shear rate and low flux resulted in dampening of bacterial accumulation at the filtration module entrances.Taken together, these observations lead to the conclusion that the transmembrane flux is a key factor in controlling bacterial losses even at shear forces that are sufficient to remove bacteria from membrane surfaces.Our findings suggest recovery to be a function of biophysical rather than biological phenomena, and hence apply to hollow fiber filtration of other types of similarly sized particles.

Figure
Figure 1a,b, and were uniquely suited to achieve high shear rates while avoiding excessive overall volumetric flowrates.Single hollow fiber filter membranes were threaded into a housing constructed from PEEK tubing and fittings.To prepare the housing, three lengths of PEEK tubing of 1/16 00 ID, 1/8 00 OD (Idex, MFR#1534, Cole-Parmer), were cut to the dimensions indicated in Figure 1a.Flangeless ferrules and nuts (Idex, MFR#XP-335X, Cole-Parmer) attached the tubing to two T-shaped fittings (Idex, MFR#P-713, Cole-Parmer) Tygon A-60, MFLX06402-16, Cole-Parmer), a peristaltic pump (Masterflex L/S drive with Easy-Load II pump head, Cole-Parmer), and a preassembled hollow fiber filter module, as shown in Figure 1a,b.Pressure sensors with monitor (PendoTECH, PMAT4A monitor and transmitter) were installed for inlet pressure measurements in the feed, retentate and permeate tubing.Connection of tubing to the sample holder required a female luer to hose barb adapter (1/8 00 ID, MFLX4550205, Cole-Parmer).Additional female and male barbed luer adapters (1/8 00 ID, MFLX4550533, Cole-Parmer) interfaced tubing with the pressure sensors.To attach tubing to the filtration module, a 1 /4-28 female screw to female luer adapter (P-658, Idex, F I G U R E 1 Assembly of a single hollow fiber filter module.(a) Three lengths of PEEK tubing, of measurements described, were attached to two PEEK tee assembly fittings using PEEK flangeless ferrules and nuts.(b) Lateral and front views of assembled single hollow fiber membrane, threaded into the housing, and glued with epoxy at each of the ends of the PEEK tubing.Cole-Parmer) was attached at either end of the module.From these, the male luer to barb adapters fastened the tubing.A hose barb in the lid of the sample holder allowed attachment of retentate tubing for recirculation.Pressure data was automatically recorded every 2 s and time stamped by PMAT system software (PendoTECH, PMAT-GUI) as a 4 to 20 mA signal into Excel, through a R-232 interface between the pressure monitor and a laptop.A balance with 0.01 g read-out F I G U R E 2 Water permeate flux as a function of transmembrane pressure for (a) 0.2 μm polyethersulfone (PES) single hollow fiber membrane, and (b) 50 kDa polysulfone (PS) single hollow fiber membrane.Different transmembrane pressures were achieved by adjusting a clamp on the retentate tubing (after the pressure sensor).

(1) a 0. 2
μm pore size polyethersulfone membrane and (2) a 50 kDa MWCO polysulfone membrane.All modules were constructed with the same length and fiber inner diameter, to maintain both a constant surface area and equal inlet shear rates across all experiments.For a given average fluid velocity, v ave (=feed flow rate/cross-sectional area), shear rate at the wall of the membrane entrance was described by: F I G U R E 4 Pressures and permeate flow rates over time for microfilters (a, b) and ultrafilters (c, d) at high and low pressures.Inlet shear rates were the same at 27,000 s À1 in all cases.(a) average flux of 1618 L m À2 h À1 ; (b) 4549 L m À2 h À1 compared to average flux of (c) 55 L m À2 h À1 and (d) 288 L m À2 h À1 , respectively.
flux.A set of experiments without the tubing clamp (i.e., lower TMP and lower flux conditions) were carried out for comparison.Recorded transmembrane flux and pressure data are shown in Figure 4 for the 0.2 μm cut-off microfilter at (a) 1618 L m À2 h À1 and (b) 4549 L m À2 h À1 , compared to 50 kDa cut-off membrane which exhibited significantly lower fluxes of (c) 55 and (d) 288 L m À2 h À1 , respectively, at the same conditions.

3. 5 |
Bacterial viability in phosphate-buffer saline (PBS) Plate counts from PBS aliquots at the start of the experiments were compared to plate counts from the same aliquots at the end of the experiments.Concentration times in the unclamped ultrafilters ranged widely, from 13 h to nearly 27 h (Figure 4c).All other concentrations were completed in under 3.5 h.A paired t-test confirmed (n = 6, p-value = 0.03) that cell viability dropped in PBS during the long, low pressure (unclamped) ultrafilter experiments.On average, only 80% of the E. coli survived in the PBS for these longer experiments.In summary, the high pressures and shear rates in this work were not sufficient to reduce bacterial viability, but filtrations in non-nutritive PBS that lasted more than 12 h reduced bacterial viability.Although the lack of nutrient availability reduced recoveries for extended duration filtrations, the use of PBS without a carbon source also provides a situation more closely mimicking concentration of viable but nonculturable microorganisms where bacteria do not multiply in enrichment media.

F I G U R E 7
Micrograph CFU counts down the length of bisected hollow fibers.Filled points indicate low pressure and flux conditions.Open points indicate high pressure and flux conditions.(a, b): Microfilters at these two conditions are shown separately; (c): Ultrafilter data is overlaid.Note that scale of h, axis is expanded 6Â compared to a and b, since accumulation of bacteria during ultrafiltration is much lower than for microfiltration.