HYDROLOGICAL PROCESSESHydrol. Process. (2013)Published online in Wiley Online Library(wileyonlinelibrary.com) DOI: 10.1002/hyp.9733

Escherichia coli strains harvested from springs in Kampala,Uganda: cell characterization and transport in saturated

porous media

George Lutterodt,1* Jan Willem A. Foppen2 and Stefan Uhlenbrook2,31 Central University College, Department of Civil Engineering, Miotso Campus, P.O Box 2305, Tema, Ghana

2 UNESCO-IHE Institute for Water Education, PO Box 3015, 2601, DA Delft, The Netherlands3 Delft University of Technology, Department of Water Resources, PO Box 5408, 2600, GA Delft, The Netherlands

*CofE-m

Co

Abstract:

We hypothesized that the transport of Escherichia coli strains harvested from springs could be characterized by a similar set ofcell characteristics and transport parameters. The hypothesis was tested by sampling springs throughout the Lubigi catchment inKampala, Uganda. Chemo-physical parameters in addition to total coliform concentrations were determined. Furthermore, E. colistrains were harvested, and cell properties determined. Column experiments in saturated quartz columns of 7 cm height wereconducted to determine transport parameters of selected E. coli strains. Using a two-site non-equilibrium sorption model,transport was modelled by fitting breakthrough data in HYDRUS 1D. Results indicated faecal contamination of the springs withhigh concentrations of total coliforms, chloride and nitrate. Furthermore, the maximum relative E. coli concentrations (C/C0)max

in the column experiments were high. Compared with our previous work on E. coli strains, collected from a pasture and from zooanimals, attachment was low. Modelling revealed that both equilibrium and kinetic sorption were not important under conditionsemployed in the experiments. These observations are explained by the way in which the strains were harvested: from terminationpoints of flow lines (springs). Such strains may possess characteristics that might have influenced their transport in the subsurfaceleading to their low attachment efficiency and possibly contributing to the lack of influence of equilibrium and kinetic sorptioncharacteristics. There was no significant correlation between cell properties and transport parameters. Furthermore, 58% of thetested strains were of the O21:H7 serotype, and all definable serotypes identified were associated with diseases. We speculate thatthis serotype may possess characteristics that allow preferential transport through the aquifers of the area. We demonstrated thatbacteria harvested from termination points of flow lines compared with those obtained from pollution sources, which have notundergone transport yet, present a good option for the assessment of bacteria transport characteristics in aquifers. Copyright ©2013 John Wiley & Sons, Ltd.

KEY WORDS Escherichia coli; attachment efficiency; faecal pollution; springs; Kampala; Uganda

Received 6 August 2012; Accepted 21 January 2013

INTRODUCTION

Groundwater is an immensely important resource, as itprovides more than one-third of the world’s drinking water(Morris et al., 2003). An important characteristic that hasoften been associated with groundwater is the assumptionthat, generally, the resource is free of pathogenic micro-organisms (Bhattacharjee et al., 2002). Although ground-water is assumed to be free from contamination bymicrobialpathogens, research reported in the literature (Macler andMerkle, 2000; Powell et al., 2003) has implicated microbialpathogens in groundwater as the cause of many water-bornedisease outbreaks. According to theWHO, an estimated onebillion people lack access to an improved water supply, andtwo million deaths are attributable to unsafe drinking water,sanitation and hygiene, with many countries still reportingcholera to the WHO (WHO, 2004). The use of shallow

orrespondence to: G. Lutterodt, Central University College, DepartmentCivil Engineering, Miotso Campus, PO Box 2305 Tema, Ghana.ail: glutterodt@central.edu.gh

pyright © 2013 John Wiley & Sons, Ltd.

groundwater for drinking and domestic purposes is acommon feature in many developing countries (Kulabakoet al., 2007). In Kampala, the capital of Uganda, protectedsprings within the shallow aquifer are a major source ofwater supply (Kulabako et al., 2007). The springs aresusceptible to pollution related to anthropogenic activities,even when protected (Kulabako et al., 2008). Previousworks undertaken on the protected springs in the area haveindicated widespread faecal contamination (Byamukamaet al., 2000; Howard et al., 2003; Kulabako et al, 2007).Because of the importance of Escherichia coli as a

faecal indicator organism, considerable attention has beengiven to understanding their transport and fate insaturated porous media (e.g. Foppen and Schijven,2006; Bolster et al., 2010; Schinner et al., 2010). Inmost of these studies, bacteria strains isolated fromdifferent sources, for example from zoo animals (Foppenet al., 2010), swine lagoons (Bolster et al., 2010), dairy cowmanure and sewerage (Haznedaroglu et al., 2008), or a soilof a pasture used for cattle grazing (Yang et al., 2008;Lutterodt et al., 2009a,2009b; Foppen et al., 2010) wereused. To our knowledge, there is no study reported in the

G. LUTTERODT, J. W. A. FOPPEN AND S. UHLENBROOK

literature that focuses on the transport of E. coli strainsisolated after they have undergone transport in an aquifer.These strains may start transport among a wide variety ofmicroorganisms that have infiltrated from the surface intothe groundwater system, and after considerable transporthas taken place, they may end up as one of the fewmicroorganisms remaining in the system. In other words,the attachment efficiencies of this type of E. coli will belower than of strains harvested from other sources. Resultsof transport experiments from such strains can perhaps beused to calculate more realistic groundwater protectionzones surrounding abstraction points. Our objectives weretwofold: (1) to present an assessment of the chemical,physical and bacteriological status of the springs in theKampala area of Uganda and (2) to characterize the transportof E. coli strains isolated from these springs, whenconsiderable transport through the aquifer has already takenplace. The underlying hypothesis was that transport of suchgroup of E. coli strains could possibly be characterized by arather homogeneous set of surface characteristics and transportparameters together with low attachment efficiency values.

MATERIALS AND METHODS

Study area

The Lubigi catchment (Figure 1) in the Kampala areaof Uganda is characterized by a weathered crystallinerock aquifer system and is comprised of a shallow aquiferand a deeper aquifer within the fissured crystallinebedrock. The shallow aquifer is in the upper layer orregolith and consists of poorly sorted sand and clays,derived from weathering of bedrock (Taylor and Howard,1998). The area is low lying with a high water table (lessthan 1.5m below the surface) in the weathered regolith(Kulabako et al., 2007), and site investigations showedthe presence of preferential flow paths, although it is notknown how far these extend (Howard et al., 2003).Previous studies have shown that the top of the regolith iscomposed of fine material, with increasingly coarser(sandy clay) material found at depth with productiveaquifers commonly associated with these layers (Howardet al., 2003; Kulabako et al., 2007). Recharge tends tooccur during two distinct wet seasons, although Kampalaexperiences rainfall throughout the year due to itsproximity to Lake Victoria (Howard et al., 2003; Nyenjeet al., 2010). Types of land use within the Lubigicatchment include formal residential, low-lying informalsettlements with high population densities (slums), small-scale industrial areas and urban agricultural farms.

Sample collection and analyses

To assess the state of pollution of springs in theKampala area, a total of 77 springs located within thevarious land use types were sampled (Figure 1).

Analyses of chemo-physical parameters

Temperature and Electrical Conductivity (EC) weredetermined using a conductivity metre (Cond340i, WTW

Copyright © 2013 John Wiley & Sons, Ltd.

GmbH, Weilheim Germany) calibrated at 25 �C, and thepH was determined using a pH metre (pH340i, WTWGmbH, Weilheim Germany). To analyse nitrate andchloride, as chemical indicators of contamination, 250mlof spring water was collected in sterile polypropylenebottles, and by means of a syringe, 25ml of sample wasfiltered through 0.45 mm cellulose acetate filter intoscintillation vials. Samples were then stored in a coolbox and transported to the Environmental Engineering andPublic Health laboratory of the University of Makerere andstored at �20 �C until they were transported to theUNESCO-IHE laboratory in Delft, the Netherlands, wherenitrate and chloride were analysed with an ICS-1000 AS14A Column (Dionex, Benelux BV).

Microbiological analyses and E. coli isolation

Total coliforms (TC) in spring waters were analysed byfilter pressing 100ml of spring water, and the filter paperwas placed on site on a Chromocult™ agar (Merck,Whitehouse Station, NJ) plate. Within 4 h, agar plateswere transported to the Environmental Engineering andPublic Health laboratory of Makerere University andincubated at 37 �C for 24 h. The number of TC on theplates was counted, and the purple coloured colonies onthe Chromocult agar plates allowed for the detection andisolation of E. coli from other types of bacteria speciesgrowing on the agar plates. To isolate E. coli, a steriletoothpick was used to pick a single colony of E. coli fromthe agar plates and inoculated into 5ml of nutrient brothfollowed by incubation at 37 �C for 24 h. Then, 1–2ml offreshly grown E. coli cells were inoculated into sterilevials (Microbank™-Dry, PRO-LAB DIAGNOSTICS,Toronto, Canada) containing porous beads saturated withcryopreservative (cryovials), which serve as carriers tosupport microorganisms. The vials were then stored at�20 �C and transported to the UNESCO-IHE laboratoryin Delft, the Netherlands.

Bacteria growth and cell characterization

To conduct experiments, a sterile forceps was used topick a bead from the cryovial, placed into 25ml ofnutrient broth and then incubated for 24 h at 37 �C whileshaking at 150 rpm on an orbital shaker to obtain a cellconcentration of ~109 cells/ml. Bacteria were washed andcentrifuged (4600 rpm; Universal Centrifuge Z383;HERMLE Labortechnik GmbH, Wehingen, Germany)three times in artificial ground water (AGW; Lutterodt etal., 2009a, 2012), which was prepared by dissolving526mg/l CaCl2.2H2O and 184mg/l MgSO4.7H2O, andbuffering with 8.5mg/l KH2PO4, 21.75mg/l K2HPO4 and17.7mg/l Na2HPO4. The final pH-value of the suspen-sions ranged from 6.6 to 6.8, whereas the electricalconductivity ranged from 980 to 1024 mS/cm.To determine motility, a 2-ml fresh culture was

centrifuged (14000 g) and washed three times in AGW,and by means of a sterile toothpick, cells were pickedfrom the remaining pellet in the test tube and inoculated atthe centre of petri dishes containing 0.35% Chromocult

Hydrol. Process. (2013)DOI: 10.1002/hyp

Figure 1. Map showing spring sampling locations in the study area (Lubigi catchment) in Kampala, Uganda

TRANSPORT OF BACTERIA STRAINS ISOLATED FROM SPRINGS

agar. The plates were incubated at 37 �C for 24 h; afterwhich, growth and diameter of migration was measuredas motility (Ulett et al., 2006).Hydrophobicity was determined with the microbial

adhesion to hydrocarbons method (Pembrey et al., 1999;Walker et al., 2005), where percentage partitioning ofcells into dodecane was measured as cell hydrophobicity.To do this, 4ml of bacteria suspension of known opticaldensity (OD) and 1ml of dodecane were vigorouslymixed in a test tube for 2min and left to stand for 15minto allow phase separation. Then, the OD of the aqueousphase was determined by a spectrophotometer (Cecil1021, Cecil Instruments Inc, Cambridge, England) at anabsorbance of 254 nm, and the percentage of cellspartitioned into the hydrophobic substance was reportedas hydrophobicity percentage.To determine cell aggregation, 15ml of freshly grown

bacteria were centrifuged (14000 g) and washed threetimes in AGW and then allowed to stand for 180min at a

Copyright © 2013 John Wiley & Sons, Ltd.

temperature of 4 �C. Then, at a depth of 1 cm below thesurface of the suspension, a 1ml sample was taken. TheOD of the samples was measured at 254 nm, and the auto-aggregation was determined as the ratio of the final overthe initial OD (in %).To determine the zeta potential, a zeta-metre similar to the

onemade byNeihof (1969)was used.Movement of bacteriawas visible on a video screen attached to a camera mountedon top of a light microscope (Olympus EHT) in phasecontrast mode (Foppen et al., 2007a,2007b). Bacteriamobility values were obtained from measurements on atleast 50 bacteria cells. Velocity measurements were used tocalculate the zeta potential with the Smoluchowksi equation(von Smoluchowski, 1921).To measure the average equivalent spherical diameter

of the cells, a light microscope (Olympus BX51) in phasecontrast mode, with a camera (Olympus DP2) mountedon top and connected to a computer, was used to takeimages of cells. At least, 50 images were imported into an

Hydrol. Process. (2013)DOI: 10.1002/hyp

G. LUTTERODT, J. W. A. FOPPEN AND S. UHLENBROOK

image processing programme (DP-Soft 2), and the averagecell width and cell length were measured. The equivalentspherical diameter was determined as the geometric mean ofaverage length and width (Rijnaarts et al., 1993).

Serotyping

Pure cultures of E. coli were grown on Chromocult agarand sent to the Dutch National Institute of Public Health andEnvironmental Hygiene (RIVM), Bilthoven, the Nether-lands, where serotyping was performed according tostandard procedures. For O typing, the strains were testedagainst antisera from O1 to O181 using the classicalapproach (Guinée et al., 1972). For H typing, the strainswere incubated at 30 �C, and the test was performed in smallglass tubes against H antisera H1 until H56. Details of themethod are described in Ewing (1986).

Porous medium and transport experiments

Porous medium comprising 99.1% pure quartz sand(Kristall-quartz sand, Dorsilit, Germany) was used in thecolumn experiments to study bacteria transport character-istics. The range of grain sizes and mean grain size weredetermined by carrying out sieve analysis. Grain size rangedfrom 180 to 500mm with a computed grain size uniformityof 1.5, and the mean of the grain size distribution, dc, wasdetermined at 356mm using the mean weight diametermethod (Van Bavel, 1950). With this grain size, weexcluded straining as a possible retention mechanism inour column: assuming a bacteria equivalent sphericaldiameter of 1.5mm, the ratio of colloid to grain diameterwas 0.004,whichwaswell below the ratio (0.007) for whichstraining was observed by Bradford et al. (2007) forcarboxyl latex microspheres with a diameter of 1.1mmsuspended in solutions with ionic strengths up to 31mM(the ionic strength of the solutions we used was 4.7mMonly). Total porosity was determined gravimetrically to be0.39. The quartz sand was rinsed sequentially with acetone,hexane and concentrated HCl, followed by repeated rinsingwith de-mineralized water until the electrical conductivitywas close to zero (Li et al., 2004). We used pure quartz sandto create an environment in which unfavourable attachment(negatively charged E. coli bacteria and negatively chargedquartz sand) in the primary energy minimum (e.g. Foppenet al., 2007a) and/or secondary energyminimum (e.g.Redmanet al., 2004) would prevail. We assumed that the effect ofgeochemical heterogeneities, or conditions of favourableattachment, on bacteria transport are well known (e.g. Ryanand Elimelech, 1996; Lutterodt et al., 2012).To assess the transport characteristics and attachment

variations among E. coli strains isolated after they hadundergone transport in an aquifer, 40 of the 71E. coli strainswere randomly selected for the column experiments. Theseexperiments were conducted in borosilicate glass columnswith an inner diameter of 2.5 cm (Omnifit, Cambridge, UK)with polyethylene frits (25mm pore diameter) and oneadjustable endpiece. The column was packed wet with thequartz sand with vibration to minimize any layering or airentrapment. Column sediment length was 7 cm. All column

Copyright © 2013 John Wiley & Sons, Ltd.

experiments were conducted in AGW at a volumetric flowrate of ~0.26 pore volume (PV) per minute (Darcyvelocity = 0.71 cm/min). We chose this velocity to enablesubstantial interaction of the cells with the quartz grains andalso to enable comparison of the results with our previouswork (Foppen et al., 2010), in which we studied thetransport of E. coli strains isolated from the faeces anddifferent parts of zoo animals and six strains isolated from asoil pasture (Lutterodt et al., 2009a,2009b).Prior to each experiment, and in order to remove

retained cells from the previous experiment, the columnwas rinsed with 1 PVof 1.9M HCl, immediately followedby a pulse of 1.5M NaOH to restore pH. The acid flushlowers the pH to less than 1, which is well below theknown acid resistance of E. coli (pH 2.5–4.5; e.g. Smallet al., 1994; Sayed and Foster, 2009). As a result, cellsdisintegrate, and most of the cell components will beflushed out of the column. To check whether the flushwith HCl followed by NaOH had indeed removed allbacterial cells, at the beginning of each experiment,effluent samples were plated in triplicate. All plates of allexperiments were negative, indicating that after theprevious experiment, all viable bacterial cells had indeedbeen removed from the column. We realized that thecleaning method we used had the inherent disadvantageof bacterial cell parts of the previous experiment stillbeing present inside the column. Therefore, to assess theeffect of the cleaning method on bacteria breakthrough,we repeated the transport experiments of 12 strains, andused the Student’s t-test to compare the mean values ofthe sticking efficiency, derived from the HYDRUS curvefitting (see the next section), of the repeats. There was nosignificant difference between the two groups of experi-ments. In addition, we assessed the possible effect ofbio-clogging on the hydraulic properties (porosity andflux rate) of the column by measuring the volumetricflow rate before and after each experiment. Resultsindicated negligible changes in the volumetric flux rate.The outcomes of these two control parameters convincedus that that the possible influence on bacterial retention ofcells remaining in the column in between experimentswas negligible, and that the experimental set-up we hadchosen was capable of acquiring transport parameters ofthe E. coli strains we used.After washing, the column was equilibrated by flushing

with 50–60PVof AGW to restore pH and EC. To conductbacteria transport experiments, a suspension ofE. coliwith aconcentration of ~108–109 cells/ml was flushed through thecolumn for 4min (approximately equal to one pore volume)followed by a flush of E. coli-free AGW. E. coliconcentrations in the shallow Kampala aquifer weremeasured to be as high as 107 cells/100ml (Kulabakoet al., 2007), so the concentrations we used were on the highside, and we considered these to be representative for aworst case scenario. Furthermore, an E coli suspensionpulse of 1 PV was considered sufficient to acquire transportparameters (Foppen et al., 2010). The E. coli concentrationwas determined using ODmeasurements (at 254 nm) with a1 cm flow-through quartz cuvette and a spectrophotometer

Hydrol. Process. (2013)DOI: 10.1002/hyp

TRANSPORT OF BACTERIA STRAINS ISOLATED FROM SPRINGS

(Cecil 1021, Cecil Instruments Inc., Cambridge, England).Cell numbers were deduced from a polynomial equation aftercalibration with plate counts on Chromocult™ agar (Merck,Whitehouse Station, NJ), using the serial dilution method. Todo this, seven tenfold dilutions of an undiluted E. colisuspension were prepared (in triplicate). The OD of alldilutions was measured, and the 107-fold dilution was platedin triplicate. The relation between the number of cells andmeasured OD was best fitted (R2 = 0.99) by a second-orderpolynomial.

Transport model

The 1D (macroscopic) mass balance equation formobile bacteria suspended in the aqueous phase exclud-ing bacteria growth and decay is normally expressed as(Corapcioglu and Haridas, 1984, 1985; Foppen et al.,2007a,2007b)

@C

@t¼ D

@2C

@x2� v

@C

@x� rbulk

θ@S

@t(1)

where C is the mass concentration of suspended bacteriain the aqueous phase (number of cells/ml), t is the time(min), D is the hydrodynamic dispersion coefficient (cm2/min), v is the pore water velocity (cm/min), S is the totalretained bacteria concentration (number of cells/g), rbulkis the bulk density (g/ml), x is the distance from thecolumn inlet (cm) and θ is the volume occupied by thefluid per total volume of medium (�).The model describes the interaction of mass between

the aqueous phase and the solid phase by a first-orderkinetic reaction together with instantaneous equilibriumsorption. The model has been successfully used (Chenet al., 2003; Mallén et al., 2005) and assumes thatinteraction can be divided into two fractions

S ¼ Se þ Sk (2)

where Se is the mass adsorbed at equilibrium sites (type-1sites) and assumed to be instantaneous and reversible, i.e.sites at which attachment and detachment are relative fastcompared with the flow velocity (Schijven and Hassaniza-deh, 2000) and allows equilibrium to occur, Sk is the massadsorbed at kinetically controlled sites (type-2 sites), whichis assumed to be a time-dependent irreversible process orkinetically limited relative to flow velocity and involvesconstant attachment and detachment rates (Schijven andHassanizadeh, 2000). The relation between the totaladsorbed mass and the mass adsorbed at type-1 and type-2sorption sites, respectively, is

Se ¼ fksC (3)

Sk ¼ 1� fð ÞS (4)

where f represents the fraction of exchange sites assumed tobe in equilibrium with the solution phase (�) and ks is anempirical coefficient (ml/cells). The time rate of change ofmass at type-1 sites is given by

Copyright © 2013 John Wiley & Sons, Ltd.

@Se@t

¼ fks@C

@t(5)

and for type-2 sites, the time rate of change of mass is givenby (Lee et al., 2009):

@Sk@t

¼ o 1� fð ÞksC � Sk½ � (6)

whereo is a first order rate constant (min�1). The parametersf, o and ks were obtained by fitting the model to measuredbacteria breakthrough curves in HYDRUS-1D (Šimůneket al., 2009). Assuming that the bacteria attachmentefficiency, a (�), and the attachment rate coefficient, ka(min�1), are time and space invariant during a columnexperiment and within the entire column, acan be computedusing the relation (e.g. Tufenkji and Elimelech, 2004)

ka ¼ 3 1� θð Þ2dc

�va (7)

whereby � is the single collector contact efficiency (�)and dc is the mean grain diameter (cm). To determine kausing the values of the parameters modelled withHYDRUS, we wrote the time rate of change of Sk as(Cameron and Klute, 1977)

@Sk@t

¼ kaθ

rbulkC � kdSk (8)

where kd is the detachment rate coefficient (min�1).Combining Equations (6) and (8) yields for the first andsecond term on the right side of Equations (6) and (8):

kaθ

rbulk¼ o 1� fð Þks½ � ; so ka ¼ o 1� fð Þks½ � rbulk

θ(9)

and

kd ¼ o (10)

Inserting (9) in (7) yields

ka ¼ 3 1� θð Þ2dc

�va ¼ o 1� fð Þks½ � rbulkθ

(11)

From this equation, the bacteria attachment efficiency,a, can be computed. We determined � with a correlationequation (e.g. Rajagopalan and Tien, 1976; Tufenkjiand Elimelech, 2004), assuming a particle density = 1055kg/m3 (Lutterodt et al., 2009a), fluid density = 1000 kg/m3, fluid viscosity = 1.005� 103 kg/m s, temperature =278K and Hamaker constant = 6.5� 10�21 J (Walkeret al., 2004).To compare the importance of equilibrium sorption

with kinetic attachment, we determined the totalbacteria mass fraction involved in both types ofinteractions with the quartz grains. The total bacteriamass fraction involved in equilibrium sorption wasdetermined as

Hydrol. Process. (2013)DOI: 10.1002/hyp

G. LUTTERODT, J. W. A. FOPPEN AND S. UHLENBROOK

Ztf

t0

fksC tð Þdt

C 0V(12)

where C0 and V are, respectively, the initial cell concentra-tion and total volume of bacteria suspension loaded and t0and tf the respective times at the beginning and end ofbreakthrough. The total bacteria mass fraction involved inkinetic attachment was determined from the measureddifference between total bacteria mass input and totalbacteria mass output. Total bacteria mass output wasobtained by the product of the area under respectivebreakthrough curve and the volumetric flow rate.

Statistical analysis

Correlation between cell characteristics and transportparameters obtained from HYDRUS modelling weredetermined with the Pearson correlation test. Also,Fisher’s least significant difference was applied inANOVA to compare the mean values of transportcharacteristics and also measured cell properties ofidentified serogroups. All statistical analyses wereperformed using SPSS 14 (SPSS Inc. Chicago).

RESULTS

Chemo-physical and bacteriological characterization ofthe springs in Kampala

Temperature values for all springs were 24.0� 0.54 �C,whereas pH values were uniform and low (mean value of5.14 and a standard deviation (SD) of 0.36). Nitrate andchloride concentrations of the samples were high andranged from 4.7 to 135.8mg/l and from 13.9 to 151.2mg/l, respectively (Table I; Figure 2a). On the basis ofchemistry, we concluded that the spring samples werepolluted, likely resulting from the infiltration of wastewater. With the exception of two springs (K03, NaK-1),all springs we had sampled contained TC with concentra-tions ranging from 2 to 1.9� 105 cells/100ml (at a lowerdetection limit of 2 CFU/100ml). A good relationbetween nitrate and chloride content was observed(r= 0.87, p= 0.00), which indicated their possible releasefrom the same source. Correlation analysis revealed a lackof association between TC and chemical parameters(chloride and nitrate; Figure 2b).

Bacteria characterization

Serotyping. The E. coli serotyping showed four differ-ent O serogroups (O14, O21, O91 and O108), whereas 16of the strains were O-untypable and six strains were nottyped. The results indicated that 58% of the serotypedE. coli strains exhibited the same serotype (O21:H7),whereas the remaining identified serotypes were O108:H45, O14:H15, O14:H4 and O91:H28. Identified H typeswith untypable O-serogroups were H31, H8 and H6. For

Copyright © 2013 John Wiley & Sons, Ltd.

some of the strains, both the O serogroup and H type wereuntypable (O-nt: H?; see Table I).

Cell characteristics. Cell characteristics (motility,hydrophobicity, cell aggregation, zeta potential andaverage cell size) of 40 randomly selected E. coli strainswere determined. Cell motility measured as the growthand diameter of migration on agar plates ranged between0.2 cm (SL43 and FA01) to 9 cm (SL21 and H31;Table II) (SD= 2.59 cm) The hydrophobicity varied fromless than 1% (FR05) to 44.6% (SL20) (SD= 9%), andauto-aggregation ranged from no auto-aggregation (SL41)to 64% (SL14) (SD= 18%), respectively. Average cellsurface charge ranged from �12.4mV for FR04 to�31.2mV for IN01 (SD=�5mV), whereas the averagesize ranged from 1.76 mm for SL08 to 3.22mm for FR04(SD= 0.26mm). Apart from the measured average cellsize, the SDs from the mean values of all measured cellcharacteristics indicated significant variations. From theseresults, we concluded that the cell properties variedconsiderably among the strains.

Transport experiments and HYDRUS modelling. Tocharacterize the transport of the E. coli strains isolated fromsprings, 40 randomly selected strains were flushed throughsaturated quartz columns of 7 cm height. Maximum peakrelative breakthrough concentrations, (C/C0)max, rangedfrom 0.06 to 0.89 representing a 1 log unit variation, with 33out of the 40 strains tested (82%) having a (C/C0)max valuegreater than 0.5 (Table II; Figure 3).Transport parameters were determined by fitting

(Figure 4) a two-site non-equilibrium model to measuredbreakthrough data using HYDRUS-1D. The resultingmodel parameters f, o , ks and the dispersivity Dx (cm) aregiven in Table II, and with these, a-values weredetermined (Table II). Generally, a-values were low (forall strains, a ≤ order 10�2), with the lowest values in theorder of 10�7. Comparison between the mass of bacteriainvolved in equilibrium sorption and in kinetic attachmentrevealed two groups of bacteria strains (Figure 5):

1. A group of strains with approximately equivalent (low)mass fractions involved in both equilibrium sorptionand kinetic attachment (group I), and

2. A group of strains with a high fraction of their cellsinvolved in kinetic attachment, whereas the fraction ofcells involved in equilibrium sorption was low andcomparable with the group I strains (group II).

Although it is well known that bacterial attachment ispredominantly kinetic, most of the strains (33 out of 40 or82%) belonged to group I. Their breakthrough wassomewhat retarded, whereas their maximum relativebreakthrough concentrations were above 0.5; their stick-ing efficiency values were usually in the order of 10�3

and 10�4. The group II strains (7 out of 40 or 18%) wereall subject to considerable kinetic attachment, and theirmaximum relative breakthrough concentrations weretherefore rather low (between 0.06 and 0.48).

Hydrol. Process. (2013)DOI: 10.1002/hyp

Table

I.Chemo-physical

andbacteriologicalcharacteristicsof

thesprings

IDLocation

pHEC

Temperature

TC

Cl

NO3

Strainisolated

from

spring

Serotype

(mS/cm)

(�C)

(cells/100

ml)

(mg/l)

(mg/l)

North

East

K01

00� 20.9180

32� 34.7870

4.89

238

231�10

322.2

63.19

SL01-SL03

O21:H7

K02

00� 20.8650

32� 34.8680

6.02

267

251�10

322.1

51.37

SL04-SL06

O21:H7

K03

00� 22.0140

32� 34.9900

5.3

127

256�10

44.7

13.91

NSI

K04

00� 22.0130

32� 34.6590

4.94

148

258�10

410.6

23.56

SL23

O-ntH?

K05

00� 21.5920

32� 35.3750

5.09

105.4

251�10

47.4

24.33

SL07-SL08

O21:H7

K06

00� 22.0060

32� 36.0060

5.01

159

231�10

510.3

34.91

SL09-SL11

O21:H7

SL16-SL17

O21:H7

K07

00� 21.3790

32� 35.5130

4.16

187

231�10

59.1

26.05

SL12-SL13

O21:H7

SL18

O21:H7

K08

00� 21.1950

32� 35.4640

4.8

196

23.5

2�10

516.5

42.82

SL14

O21:H7

SL25

O-nt:H

?K09

00� 21.2960

32� 35.4730

5.12

214.6

23.6

1�10

517.5

36.9

SL28

O14:H4

K10

00� 21.1080

32� 33.3520

4.4

553.8

23.8

1�10

562.2

119.09

SL15

O21:H7

SL24

O-nt:H

31K11

00� 21.1250

32� 33.2730

4.85

528.3

23.6

2�10

060.5

112.02

SL29

O-nt:H

8K12

00� 21.2100

32� 33.1400

4.8

440.4

24.4

1�10

338.9

82.82

SL27

O-nt:H

8K13

00� 20.7370

32� 32.9650

4.84

245.6

23.4

3�10

320.7

63.98

SL26

O-nt:H

?K14

00� 20.6160

32� 33.4030

4.87

285.3

231�10

526.2

62.36

SL21

O-nt:H

31K15

00� 20.6180

32� 33.5150

5.18

163.2

23.5

2�10

412

25.19

SL19

O-nt:H

31K16

00� 20.5490

32� 32.7430

5.06

308

24.6

4�10

129.1

74.28

NSI

K17

00� 20.6000

32� 32.8490

4.9

177.6

23.8

1�10

416.9

38.68

SL22

O-nt:H

6K18

00� 20.7180

32� 32.8990

4.74

199.5

24TNTC

26.2

48.71

SL20

O-nt:H

8K19

00� 20.7350

32� 32.9640

4.95

275.1

23.6

7�10

320.8

52.22

NSI

K20

00� 20.6390

32� 32.4360

4.96

327

24.1

4�10

326.7

61.93

SL31

O21:H7

K21

00� 20.4760

32� 32.1290

4.71

254

23.7

5�10

422.9

58.63

NSI

K22

00� 20.3300

32� 32.8690

5.1

285

24.3

7�10

422.6

44.4

NSI

K23

00� 20.2640

32� 32.9080

5.16

230

23.9

9�10

333.5

54.53

NSI

K24

00� 20.2280

32� 32.9220

5.06

306

23.9

3�10

321.3

58.01

NSI

K25

00� 20.7630

32� 33.3040

4.83

382

24.7

1�10

537.6

66.76

NSI

K26

00� 20.4670

32� 33.4520

4.94

341

24.4

3�10

326.6

54.39

NSI

K27

00� 20.2590

32� 33.5640

4.94

258

242�10

525.7

60.61

NSI

K28

00� 20.1400

32� 33.5090

4.95

614

24.9

2�10

561

118.45

SL33

O14:H15

K29

00� 20.0620

32� 33.5190

5.25

761

24.3

TNTC

72.9

151.18

NSI

K30

00� 19.9120

32� 33.5940

5.4

171.2

24.3

2�10

2135.8

148.94

NSI

K31

00� 19.9000

32� 33.6770

5.33

37.2

243�10

317.6

33.72

SL30

O21:H7

K32

00� 19.8000

32� 33.5810

6.03

92.7

24.4

2�10

420.3

32.9

NSI

K33

00� 19.7000

32� 33.7510

5.02

65.7

24.4

2�10

027

74.18

NSI

K34

00� 18.9540

32� 33.1650

4.95

33.2

23.4

1�10

36.6

19.85

FR04-FR05

O21:H7

Contin

ues

TRANSPORT OF BACTERIA STRAINS ISOLATED FROM SPRINGS

Copyright © 2013 John Wiley & Sons, Ltd. Hydrol. Process. (2013)DOI: 10.1002/hyp

Table

I.(Contin

ued)

IDLocation

pHEC

Tem

perature

TC

Cl

NO3

Strainisolated

from

spring

Serotype

(mS/cm)

(�C)

(cells/100

ml)

(mg/l)

(mg/l)

North

East

FR07-FR08

O21:H7

FR09

O-nt:H

31K35

00� 19.0050

32� 32.9640

4.8

232

23.7

1�10

124.6

56.28

FR01

O-nt:H

31FR02

O21:H7

FR06

andFR10

O21:H7

FR11

O-nt:H31

K36

00� 19.0100

32� 32.8060

5.24

139.1

23.6

2�10

38.6

22.01

FA01

O21:H7

FA03

O108:H45

FA04-FA06

Notyping

K37

00� 18.8620

32� 33.0010

4.98

175.9

23.6

2�10

512.1

23.07

FA02

O108:H45

FA05,FA07,FA08

Notyping

K38

00� 18.8620

32� 33.1540

5.2

315

23.8

1�10

324.8

79.81

FR12-FR14

Notyping

NaK

-100

� 20.4770

32� 35.9900

5.5

103.8

24.4

0�10

06.7

14.21

NSI

NaK

-200

� 20.7560

32� 36.5900

5.07

400

24.4

1�10

431.1

92.05

IN01

O-nt:H

8IN

06O21:H7

NaK

-300

� 19.8400

32� 36.2610

5.88

164.5

24.9

2�10

310.1

21.68

IN03

O21:H7

IN05

O-nt:H

8IN

08Notyping

NaK

-400

� 19.8120

32� 36.2310

5.4

174

24.7

9�10

312.7

27.18

IN02

O-nt:H

8IN

04andIN

07O21:H7

K43

00� 21.5600

32� 34.8190

4.91

284

23.9

8�10

324.4

80.01

NSI

K44

00� 20.7930

32� 34.2700

4.98

352

24.6

6�10

331.3

76.36

NSI

K45

00� 20.4730

32� 34.3460

5.3

422

23.9

6�10

338.4

82.48

SL37-SL38

O21:H7

K46

00� 21.7010

32� 32.4290

5.02

166

23.3

1�10

45.7

12.18

SL39

O21:H7

K47

00� 21.6260

32� 32.2330

4.98

162

23.5

3�10

321.4

19.39

SL40

O21:H7

K48

00� 21.4460

32� 32.3890

4.95

158

23.7

6�10

415.8

34.68

NSI

K51

00� 21.0240

32� 32.7450

4.91

211

23.5

1�10

422

42.02

SL41

O21:H7

K52

00� 21.0850

32� 32.9340

5.21

311

23.8

1�10

521.7

54.34

NSI

K53

00� 22.8460

32� 34.6550

5.24

403

24.2

1�10

236.5

73.47

SL42

O21:H7

K54

00� 22.4830

32� 34.5150

5.25

316

24.3

1�10

516.6

53.48

NSI

K55

00� 22.3510

32� 34.5190

5.48

266

24.5

1�10

418.5

80.35

SL43

O21:H7

K56

00� 22.2590

32� 34.5670

5.24

317

242�10

424.8

64.34

NSI

K57

00� 22.1970

32� 34.5780

5.14

212

24.5

5�10

313.4

37.21

SL44

O21:H7

K58

00� 22.1800

32� 34.3090

5.14

232

24.3

5�10

416.6

50.26

NSI

K59

00� 21.7770

32� 34.4890

5.5

432

24.8

2�10

217.6

30.23

SL45-SL46

O91:H28

K60

00� 21.1980

32� 34.5960

5225

23.9

2�10

319.6

36.03

SL34

O21:H7

SL35

O14:H15

K61

00� 18.8060

32� 33.8930

5.2

563

23.9

3�10

337.1

116.38

NSI

K62

00� 18.5050

32� 33.8860

5.3

493

23.9

TNTC

38.4

88.67

NSI

K63

00� 18.4960

32� 33.8720

5.15

335

23.6

1�10

320.3

55.13

NSI

K64

00� 18.4880

32� 33.8590

5.4

265

23.4

1�10

419

45.77

NSI

G. LUTTERODT, J. W. A. FOPPEN AND S. UHLENBROOK

Copyright © 2013 John Wiley & Sons, Ltd. Hydrol. Process. (2013DOI: 10.1002/hyp

)

K65

00� 18.4600

32� 33.8150

5.2

246

23.3

1�10

418.2

41.81

NSI

K66

00� 18.2850

32� 34.5480

5.73

750

24.6

2�10

458.3

94.54

NSI

K67

00� 18.2000

32� 34.5410

5.3

988

24.2

TNTC

90.8

142.82

SL36

O14:H15

K68

00� 18.1860

32� 34.5450

6.01

874

23.9

1�10

364.7

82.93

NSI

K69

00� 18.1130

32� 34.5110

6.5

609

25.1

TNTC

52.3

69.69

NSI

K70

00� 20.0500

32� 34.4680

5.69

327

23.9

4�10

511.3

28.65

NSI

K71

00� 20.2000

32� 34.4750

5.2

187

23.7

1�10

421

43.23

NSI

K72

00� 18.2390

32� 34.9620

5.2

221

23.3

2�10

314.7

49.7

NSI

K73

00� 18.1430

32� 34.9450

5.35

350

23.6

1�10

433.4

59.05

NSI

K74

00� 17.5130

32� 35.1660

4.98

356

23.9

TNTC

31.5

82.1

NSI

K75

00� 17.4880

32� 35.2240

4.83

231

23.9

2�10

320.9

64.84

NSI

K76

00� 17.4500

32� 35.2980

5.18

203

24.3

9�10

312.5

29.8

NSI

K77

00� 17.4170

32� 35.3980

5.12

199

23.1

6�10

318.8

44.72

NSI

TNTC,toonumerousto

count;TC,totalcoliforms.ID

,SpringIdentifi

catio

n,EC,ElectricalConductivity.

0.E+00

5.E+04

1.E+05

2.E+05

0 40 80 120 160

TC

(#

cells

/100

ml)

Concentration (mg/l)

Chloride

Nitrate

0

40

80

120

160

0 50 100 150

NO

3 (m

g/l)

Cl (mg/l)

(b)

(a)

Figure 2. Relation between parameters (a) nitrate and chloride concentra-tions in the studied spring waters (r= 0.87, p= 0.00) and (b) total coliforms(TC) and chemical parameters – nitrate and chloride (nitrate: r=�0.1,

p= 0.2, chloride: r=�0.09, p= 0.23)

TRANSPORT OF BACTERIA STRAINS ISOLATED FROM SPRINGS

Copyright © 2013 John Wiley & Sons, Ltd.

Correlation among cell properties and model para-meters. The relations between measured cell properties,parameters obtained from HYDRUS modelling and alsomaximum peak relative breakthrough concentrations wereanalysed using Pearson’s correlation test. Generally,correlations were low and statistically insignificant(Table III). However, direct and inverse statisticallysignificant correlations were observed for the relationsbetween model parameters a and o and between a andf, respectively. Also, the relationship between ks and(C/C0)max was statistically significant. This was notsurprising because the empirical coefficient, ks, deter-mined to a certain extent the magnitude of the attachmentrate coefficient (Equation 11). Four identified O-serogroupswere grouped, and one way ANOVA was applied tocompare mean values of transport characteristics andmeasured cell properties between the groups. Resultsindicated that with the exception of f, which showed asignificantly (p= 0.021) different mean value betweenserogroups O21 and O108, mean values of transportcharacteristics for the serogroups (O14, O21, O91 andO108) were not significantly different between the groupswith high p values (p> 0.05) (for a, p=0.254; for o,p = 0.605; for (C/C0)max, p = 0.657; for ks, p = 0.833).Average values of cell properties grouped according to

Hydrol. Process. (2013)DOI: 10.1002/hyp

Table

II.Measuredpropertiesof

Escherichia

colistrains(leftpart)andparametersobtained

byfitting

breakthrough

curves

inHYDRUS-1D

software(right

part)

MeasuredE.colistrain

property

Param

etersobtained

from

modellin

g

Strain

Serotype

Motility

(cm)

Hydrophobicity

(%)

Cellaggregation

(%)

Zeta

potential(m

V)

Average

size

(mm)

(C/C

0) m

ax

(�)

Dx

(cm)

a (�)

f(�

)o

(min

�1)

k s(m

l/cells)

R2

FR01

Ont

H31

5.50

1.89

1.44

�23.21

2.41

0.71

0.18

1.4�10

�7

0.008

1.9�10

�6

2.31

0.98

FR02

O21

H7

7.45

4.04

47.02

�27.53

2.26

0.67

0.02

1.1�10

�2

0.119

1.7�10

00.19

0.98

FR03

O21

H7

6.90

1.33

15.60

�27.25

2.45

0.38

0.12

2.0�10

�3

0.023

9.1�10

�3

6.36

0.99

FR04

O21

H7

5.60

6.20

1.98

�12.39

3.22

0.62

0.35

6.6�10

�3

0.000

1.3�10

00.17

0.98

FR05

O21

H7

6.25

0.21

2.77

�18.96

2.12

0.64

0.17

7.0�10

�3

0.176

1.2�10

00.18

0.99

FR06

O21

H7

8.45

7.99

15.24

�19.79

2.47

0.48

0.62

1.1�10

�3

0.287

8.9�10

�2

0.51

0.98

FR07

O21

H7

7.25

29.86

21.20

�15.40

2.46

0.61

0.02

5.1�10

�3

0.223

9.2�10

�1

0.20

0.99

FR08

O21

H7

6.95

8.78

54.37

�13.89

1.83

0.89

0.00

2.3�10

�3

0.609

7.8�10

�1

0.19

0.98

IN01

Ont

H8

7.65

6.88

0.70

�31.18

2.44

0.81

0.03

1.1�10

�2

0.013

2.6�10

00.13

0.98

IN02

Ont

H8

6.75

3.43

1.52

�26.53

1.92

0.74

0.00

7.2�10

�3

0.069

1.4�10

00.14

0.99

IN03

O21

H7

0.35

0.59

1.50

�25.74

1.85

0.68

0.00

1.1�10

�2

0.115

1.7�10

00.19

0.98

IN05

Ont

H8

8.15

19.40

2.40

�25.79

2.39

0.71

0.01

1.2�10

�2

0.000

2.0�10

00.17

0.98

IN06

O21

H7

5.95

7.25

6.16

�30.65

2.12

0.71

0.00

1.1�10

�2

0.052

2.0�10

00.15

0.99

IN07

O21

H7

6.15

13.40

0.95

�21.10

2.16

0.63

0.21

2.3�10

�4

0.517

3.9�10

�2

0.32

0.99

SL01

O21

H7

7.30

7.44

1.85

�20.60

2.17

0.67

0.22

8.7�10

�3

0.421

2.3�10

00.18

0.99

SL02

O21

H7

6.85

11.58

28.24

�13.89

2.57

0.33

0.08

2.5�10

�3

0.067

3.4�10

�2

2.22

0.99

SL03

O21

H7

6.05

12.69

1.98

�17.50

2.23

0.68

0.18

3.6�10

�4

0.628

1.1�10

�1

0.22

0.99

SL04

O21

H7

6.70

13.77

26.19

�18.96

2.23

0.15

0.26

5.3�10

�3

0.046

3.5�10

�2

4.31

0.95

SL08

O21

H7

2.10

7.18

0.73

�26.68

1.76

0.66

0.26

1.5�10

�6

0.998

9.0�10

�2

0.18

0.99

SL12

O21

H7

4.00

11.72

1.50

�25.25

2.17

0.67

0.20

1.0�10

�4

0.514

2.0�10

�2

0.27

0.98

SL14

O21

H7

8.60

12.17

63.87

�21.30

2.38

0.18

0.20

3.6�10

�3

0.011

6.4�10

�3

15.8

0.99

SL16

O21

H7

7.10

13.97

2.05

�18.11

2.39

0.32

0.10

2.3�10

�3

0.025

1.3�10

�2

5.03

0.99

SL20

Ont

H8

5.65

44.63

2.60

�21.46

2.34

0.80

0.17

1.6�10

�7

0.870

2.1�10

�4

0.16

0.98

SL21

Ont

H31

9.00

5.11

0.39

�19.60

2.37

0.65

0.03

6.1�10

�3

0.338

1.0�10

10.26

0.99

SL22

Ont

H6

4.75

11.40

0.00

�17.73

2.17

0.57

0.14

3.0�10

�3

0.446

5.1�10

�1

0.28

0.98

SL28

O14

H4

6.25

4.22

3.91

�18.28

1.84

0.70

0.24

5.3�10

�3

0.352

1.0�10

00.20

0.99

SL30

O21

H7

5.30

9.50

1.70

�12.76

2.30

0.06

0.05

6.9�10

�3

0.006

8.8�10

�3

21.5

0.96

SL31

O21

H7

7.05

6.21

54.56

�19.00

2.44

0.72

0.11

1.5�10

�3

0.629

9.0�10

�1

0.13

0.99

SL33

O14

H15

6.65

6.92

0.45

�20.50

2.39

0.71

0.13

4.6�10

�4

0.740

2.7�10

�1

0.19

0.99

SL37

O21

H7

2.45

7.41

44.14

�20.08

1.82

0.67

0.00

6.8�10

�4

0.951

1.7�10

00.20

0.98

SL40

O21

H7

4.05

25.82

0.96

�21.17

2.34

0.74

0.22

1.7�10

�3

0.686

8.6�10

�1

0.18

0.99

SL41

O21

H7

0.50

10.74

0.00

�18.64

2.38

0.69

0.01

4.7�10

�3

0.384

1.0�10

00.21

0.99

SL42

O21

H7

6.70

22.71

36.17

�16.20

2.18

0.66

0.11

8.6�10

�4

0.298

7.3�10

�2

0.45

0.99

SL43

O21

H7

0.20

2.97

4.17

�19.65

1.98

0.69

0.04

1.8�10

�3

0.760

1.0�10

00.19

0.99

SL44

O21

H7

1.30

5.45

0.68

�14.30

2.32

0.67

0.00

5.8�10

�3

0.360

1.2�10

00.22

0.99

SL45

O91

H28

1.00

27.30

3.00

�26.79

1.86

0.80

0.16

1.9�10

�7

1.000

3.3�10

�1

0.14

0.98

G. LUTTERODT, J. W. A. FOPPEN AND S. UHLENBROOK

Copyright © 2013 John Wiley & Sons, Ltd. Hydrol. Process. (2013DOI: 10.1002/hy

)p

SL46

O91

H28

8.70

16.53

3.35

�14.74

2.45

0.54

0.16

7.9�10

�4

0.509

1.1�10

�1

0.41

0.99

FA01

O21

H7

0.20

11.88

0.38

�12.75

2.19

0.61

0.18

5.4�10

�4

0.252

2.9�10

�2

0.66

0.99

FA02

O108H45

6.15

19.36

0.85

�16.12

2.14

0.59

0.22

4.4�10

�4

0.747

1.9�10

�1

0.25

0.99

FA03

O108H45

8.00

3.51

0.83

�26.37

2.41

0.74

0.22

1.2�10

�6

0.998

1.4�10

�1

0.15

0.99

a,computedsticking

efficiency

(�).

k s,em

piricalcoefficient(m

l/cells).

o,first-orderrate

coefficientforonesite

ortwosite

non-equilib

rium

adsorptio

n,masstransfer

coefficientforsolute

exchange

betweenmobile,andim

mobile

liquidregions(m

in�1).

f,fractio

nof

adsorptio

nsiteswith

instantaneoussorptio

nwhenchem

ical

non-equilib

rium

isconsidered

(�).

Dx,dispersivity

(cm).

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8

C/C

0

Pore Volume (-)

FR01

FR02

IN01

IN03

FR04

SL01

SL20

Input Mass

Figure 3. Representative breakthrough curves of Escherichia coli strainswith high (C/C0)max

TRANSPORT OF BACTERIA STRAINS ISOLATED FROM SPRINGS

Copyright © 2013 John Wiley & Sons, Ltd.

the four O-serogroups did not show any significantdifference among their mean values (for cell aggregation,p=0.531; motility, p= 0.156, zeta potential, p= 0.534;hydrophobicity, p=0.483, size, p=0.818).

DISCUSSION

Chemo-physical and bacteriological quality of the springwaters in the Kampala area

An important conclusion from our work was thatalmost all springs we sampled had high concentrations ofTC, nitrate and chloride, whereby nitrate and chloridewere correlated (r= 0.87, p= 0.00). This suggested thatwaste water, which is so abundantly present and disposedof in Kampala was the source of contamination of thesprings. Our findings are consistent with observationsmade by Howard et al. (2003) and Kulabako et al. (2007),who both identified multiple sources of anthropogeniccontamination of the shallow aquifer from which most ofthe springs tap their water. For example, solid wastedumps, pit latrines, unlined grey water channels or greywater polluted unlined storm water drainage channels areall prevalent in most of the Kampala area. In addition,spring protection was not optimal as protective fenceswere broken and faulty, thus allowing free-range cattleand other domestic animals to access the site and therebyincreasing the vulnerability of the springs to faecalcontamination.Although the length of groundwater flow paths in the

area is unknown, the non-responsiveness of springdischarge to rainfall observed by Kulabako et al. (2007)suggests that the springs in the area may be at the end ofregional flow lines. The likely sources of pollution (highnitrate and TC) in the area may therefore come from acombination of local sources within the vicinity of thesprings (Nsubuga et al., 2004; Kulabako et al., 2007) andpollutants transported from a wider area outside thecatchment (Kulabako et al., 2007).The lack of direct correlation between EC and TC

suggested the possibility of multiple sources of faecal

Hydrol. Process. (2013)DOI: 10.1002/hyp

0

0.2

0.4

0.6

0.8

1

0 1 2 3 4 5 6

C/C

0C

/C0

C/C

0

C/C

0C

/C0

C/C

0

PV(-) PV(-)

PV(-) PV(-)

PV(-) PV(-)

Observed

HYDRUS Fitted

R2=0.98

R2=0.99

R2=0.99

R2=0.99

R2=0.98

R2=0.98

(a) SL01

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8

Observed

HYDRUS Fitted(b) SL44

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8

Observed

HYDRUS Fitted(c) FA01

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8

Observed

HYDRUS Fitted

(d) IN03

0

0.2

0.4

0.6

0.8

1

0 2 4 6 8

Observed

HYDRUS Fitted

(e) FR01

0

0.2

0.4

0.6

0.8

1

0 1 2 3 4 5 6

Observed

HYDRUS Fitted

(f) IN05

Figure 4. Representative breakthrough curves fitted with HYDRUS-1D to obtain modelling parameters (f, o and ks)

0.0

0.2

0.4

0.6

0.8

1.0

0.0 0.2 0.4 0.6 0.8 1.0

No

n-e

qu

ilib

riu

m E

. co

li m

ass

(Fra

ctio

n o

f in

pu

t m

ass)

Equilibrium E. coli mass (Fraction of input mass)

Group I

Group II

Figure 5. Relation between fraction of cells with kinetic and equilibriumtransport characteristics, indicating two groups. Group I, strains withequivalent (low) mass fractions involved in both equilibrium sorption andkinetic attachment and group II, strains with a high fraction of cells

involved in kinetic attachment

G. LUTTERODT, J. W. A. FOPPEN AND S. UHLENBROOK

Copyright © 2013 John Wiley & Sons, Ltd.

pollution leaching into the aquifers and polluting groundand spring waters.

Low sticking efficiencies and inter-strain attachment variability

An important observation from the transport experi-ments and modelling exercise is the set of low stickingefficiency values we measured. Even though experimentalconditions employed were similar to that of our previouswork (Foppen et al., 2010), the attachment efficiencyvalues measured (10�7 to 10�2) are much lower. To ourknowledge, a-values in the orders of 10�6 and 10�7

measured have not been recorded over transport distancesless than 1m. These values are only comparable with thesegment sticking efficiencies (Lutterodt et al., 2011) andthe so-called minimum sticking efficiency extrapolated inour previous works (Lutterodt et al., 2009b, 2011, 2012).The low attachment characteristics (low sticking effi-ciency and high effluent breakthrough) of the E. coli

Hydrol. Process. (2013)DOI: 10.1002/hyp

Table III. Correlation matrix for (C/C0)max, measured cell properties and model parameters (a, f, o, ks)

C/C0–max a f o ks Cell aggregation Motility Zeta potential Hydrophobicity

a 0.006f 0.476 �0.638o 0.444 0.796 �0.215ks �0.789 0.059 �0.380 �0.326Cell aggregation �0.180 �0.059 �0.052 �0.023 0.219Motility �0.059 �0.024 0.018 �0.099 0.054 0.154Zeta potential 0.215 0.368 �0.179 0.394 �0.116 �0.068 �0.369Hydrophobicity 0.048 �0.284 0.269 �0.275 �0.044 �0.023 �0.092 �0.046Average size �0.071 0.117 �0.263 0.073 �0.009 �0.080 0.074 �0.257 �0.050

Statistically significant (p ≤ 0.01) correlations are in bold.

TRANSPORT OF BACTERIA STRAINS ISOLATED FROM SPRINGS

strains may be compared with observations made byDong et al., (2002), who observed high effluent recoveryand low values of collision efficiency for two indigenousgroundwater bacterial strains when transported throughgranular porous sediment.Although it is difficult to explain these observations from

E. coli surface characteristics (Table III), the strains we usedwere isolated from flow lines at their termination points(springs), and such strains may therefore possess certain cellcharacteristics that might have influenced their selectivetransport in the subsurface giving rise to their low attachmentefficiencies. Although an overall 5 log unit inter-strainattachment variation was observed between the highest andlowest attaching strains, for substantial numbers of the strainssticking efficiency values varied by an order of 2 log units(10�2 to 10�4). These transport characteristics in terms ofretention variability were similar to those reported in theliterature (Haznedaroglu et al., 2009; Bolster et al., 2010;Foppen et al., 2010). Foppen et al. (2007a) and Bolster et al.(2010) reported inter-strain and intra-strain attachmentvariations of ~2 log units. This observed variability inattachment was likely due to differences in those cellproperties that promote transport and attachment of the cells.

Correlating cell properties and model parameters

In contrast to observations made by some workers (e.g.Walker et al., 2005; Bolster et al., 2006, 2010; Jacobs et al.,2007), who reported that hydrophobicity and zeta potentialplay an important role in bacteria transport, but consistentwith our previous observations (Lutterodt et al., 2009a;Foppen et al., 2010) and observations made by Bolster et al.(2009, 2010), our results demonstrated that there was nostatistically significant correlation between measured cellproperties (zeta potential, motility, cell size, cell aggregationand hydrophobicity) and transport parameters (f, o and ksand (C/C0)max). Bolster et al. (2010) explained thatadditional confounding factors may be present whencorrelating E. coli transport to surface properties. Inaddition, various researchers (Bradford and Bettahar,2005; Tong and Johnson, 2007; Castro and Tufenkji,2008; Bolster et al., 2009; Haznedaroglu et al., 2009;Lutterodt et al., 2009a; Schinner et al., 2010) observedtransport and attachment variability among bio-colloids andattributed their findings to heterogeneity in bio-colloid

Copyright © 2013 John Wiley & Sons, Ltd.

population, such as differences in cell surface charges(Baygents et al., 1998; Haznedaroglu et al.,2009), variablesurface properties (Li et al, 2004; Tufenkji and Elimelech,2005a,2005b; Tong and Johnson, 2007), variations inlipopolysaccharide (LPS) coating (Simoni et al., 1998),differences in the ability to express an outer membrane autotransporter protein-Antigen 43 (Lutterodt et al., 2009a),distribution of interaction potential within bio-colloidpopulations (Li et al., 2004) and straining (Bradford andBettahar, 2005). However, the direct and inverse statisticallysignificant correlations observed for the relations between aand o, between a and f, and between a and ks, respectively,are obvious because these parameters have a mathematicalrelationship with a in Equation (11).Even though one porous medium was used throughout

the experiments, the parameter f varied. Because of thehigh R2 values of modelling results (Table II), we ruledout improper fitting of model parameters as the cause ofvariability in f. This result is consistent with observationsmade by Chen et al. (2003) and suggests the existence ofdifferent binding sites at the interface between the E. colicells and quartz grains we used. The non-correlation off with any of the cell properties further supports theobservation that none of the measured cell propertiesaffected the transport and attachment of E. coli to thequartz sand used. In contrast to Foppen et al. (2010), therewas no correlation between a and (C/C0)max, even thoughthe results showed an inverse and statistically significantcorrelation between ks and (C/C0)max. Although theextended tails observed in the breakthrough curves(Figure 5) of almost all E. coli strains might havecontributed to the lack of correlation between a and (C/C0), we attributed the non-correlation between the twoparameters to the fact that all strains were not onlysubjected to kinetic attachment but also to equilibriumsorption. This type of sorption was absent in the E. colistrains obtained from the Rotterdam Zoo in the work ofFoppen et al. (2010). In addition, Foppen and Schijven(2006) reported that on the basis of a number of studies,equilibrium sorption in E. coli transport was of littlesignificance (Alexander and Seiler, 1983; Havemeisterand Riemer, 1985; Sinton et al., 1997; Sinton et al., 2000;Powelson and Mills, 2001; Pang et al., 2003, both in:Merkli, 1975; Matthess et al., 1985; Champ andSchroeter, 1988; Matthess et al., 1988). It must be

Hydrol. Process. (2013)DOI: 10.1002/hyp

G. LUTTERODT, J. W. A. FOPPEN AND S. UHLENBROOK

emphasized here that the role of the two retentionmechanisms we observed was limited for most of thestrains, and perhaps the mixing of influences of dispersionand attachment in these short duration pulse experimentsconfounded expected relationships. We hypothesize thatthe presence of some equilibrium sorption could havebeen inherent to the approach we took in collecting E. colistrains, isolated from flow lines at their termination points(springs), and these strains may therefore have possessedcertain typical transport characteristics that becameapparent in our columns. For two of the strains, SL04and SL30, the total mass of equilibrium and kineticallyadsorbed fractions exceeded 100% (Figure 5). Thisobservation can be explained by the methods applied tomeasure the mass fractions involved in the two types ofinteractions; the equilibrium mass fraction was model-derived, whereas the kinetic attachment was based onmeasured values. This might have led to the anomaly inresults obtained for the two strains.

E. coli serotypes harvested from spring waters in Kampala

Typing results indicated that 58% of the strains wetested were of the O21:H7 serotype. The O antigen ofE. coli consists of many repeats of an oligosaccharide unitand forms part of the LPS molecule protruding from theouter membrane of an E. coli cell into its immediateenvironment. The O antigen determines the serogroup ofan E. coli, and the specific combination with the flagellar(H) antigen determines the serotype of an isolate (Stenutzet al., 2006). LPS is anchored in the outer membrane ofE. coli and occupies 75% of the surface of the gram-negative bacterium, and E. coli is known to have some106 LPS molecules per cell (Caroff and Karibian, 2003).The interaction of E. coli with its immediate surroundingmay, to a certain extent, therefore be controlled by thesemolecules. For instance, Foppen et al. (2010) found anassociation between E. coli serotype on the one hand andattachment efficiency on the other. Although the transportmechanisms in the Kampala aquifers of the strains weused are unknown, the specific serotype O21:H7 maypossibly possess certain characteristics that allow itspreferential transport through the aquifers in the Kampalaarea, so that some 60% of the strains we harvestedpossessed this particular serotype.According to Garabal et al. (1996), E. coli strains

belonging to a specific serotype or serogroup do notthemselves confer virulence; rather, it has often beendemonstrated that serotyping can be used as an indicatorfor pathogenicity, because there is a high positivecorrelation between certain serotypes and (entero)pathogenicity. We identified five serogroups, i.e. O21,O14, O91, O108, and a group with O-undefinableserotypes. It is interesting to note that all definable O-serogroups were associated with (diarrhoeal) diseases.Jenkins et al. (2006) and Kang et al. (2001) reportedisolation of E. coli O21 from patients with diarrhoea,whereas Knobl et al. (2001) also identified a pathogenicE. coli isolate of serotype O21. Furthermore, E. coli

Copyright © 2013 John Wiley & Sons, Ltd.

belonging to O14 and O91 serogroups is known to beassociated with urinary tract infections and diarrhoea withlife-threatening complications (Bettelheim, 1978; Stenutzet al., 2006). Finally, E. coli O108 has been isolated fromfaeces of adults with urinary tract infections (Bettelheim,1978). The observation that all definable serogroups andserotypes of E. coli strains isolated from springs in theKampala area could be associated with diseases includingdiarrhoea and urinary tract infections has importantimplications for the use of these springs for drinkingpurposes. Our finding supports the suggested linkbetween cholera outbreak and use of contaminated watersfetched from protected springs (Howard et al., 2002). Thetransport of both disease causing and commensal bacterialstrains through the aquifers of the area renders untreatedspring waters unsafe for drinking. The finding of thiswork further buttresses the emphasis on the vulnerabilityof the shallow groundwater in Kampala to faecalcontamination (Howard et al., 2003).

CONCLUSIONS

On the basis of our work from the Kampala springs, weconcluded the following:

• Almost all springs we sampled had high concentrationsof thermotolerant coliforms, nitrate and chloride,whereby nitrate and chloride were correlated. Thissuggested that waste water, which is so abundantlypresent and disposed of in Kampala was the source ofcontamination of the springs.

• In general, under the experimental conditions weemployed, bacterial attachment efficiencies variedbetween 10�2 and 10�7. We considered these efficien-cies to be low. Interactions with the sediment, either inthe form of equilibrium sorption or kinetic attachmentwere relatively unimportant. From this result, we inferthat to determine realistic travel distances of bacteria inaquifers, originating from waste water infiltration, theuse of E. coli strains harvested from endpoints of flowlines might serve as a representative worst case proxy.

• Apart from statistically significant correlations betweenmathematically linked model parameters, there was norelationship amongmeasured cell properties and betweenmodel parameters and cell properties. In other words,identifying causal relationships between E. coli surfacecharacteristics and retention of E. coli in sediment upontransport in groundwater still remains a challenge.

• Only a limited number of different E. coli serotypeswere found in the spring waters we sampled. However,all definable serotypes are associated with diarrhoeaand/or urinary tract infections, and this renders the useof untreated spring waters for drinking purposes unsafe.

ACKNOWLEDGEMENTS

Our sincere gratitude goes to Dr Robinah Kulabako of thePublic Health and Environmental Engineering Depart-ment of the faculty of Technology, University ofMakerere, Uganda. Appreciation also goes to Peter

Hydrol. Process. (2013)DOI: 10.1002/hyp

TRANSPORT OF BACTERIA STRAINS ISOLATED FROM SPRINGS

Kiyaga who was so helpful in finding the springs in theKampala area. We would also like to thank the entirelaboratory staff of UNESCO-IHE for their immensesupport during the experiments. This research was fundedby the Netherlands Ministry of Development Cooperation(DGIS) through the UNESCO-IHE Partnership ResearchFund. It was carried out jointly by UNESCO-IHE,Makerere University and the Kampala City Council inthe framework of the Research Project ‘Addressing theSanitation Crisis in Unsewered Slum Areas of AfricanMega-cities’ (SCUSA). It has not been subjected to peerand/or policy review by DGIS and, therefore, does notnecessarily reflect the view of DGIS.

REFERENCES

Alexander I, Seiler KP. 1983. Lebensdauer und transport von bakterien intypischen grundwasserleitern. Munchener Schotterebene. DVGW-Schr.,35, 113-125(ZfGW, Frankfurt/Main)

Baygents JC, Glynn JR, Albinger O, Biesemeyer BK, Ogden KL ArnoldRG. 1998. Variation of surface charge density in monoclonalpopulations: Implications for transport through porous media. Environ-mental Science and Technology 32: 1596–1603

Bettelheim KA. 1978. The sources of ’OH’ serotypes of Escherichia coli.Journal of Hygiene, Cambridge 80: 83–113.

Bhattacharjee S, Ryan JN, Elimelech M. 2002. Virus transport inphysically and geochemically heterogeneously subsurface porousmedia. Journal of Contaminant Hydrology 57: 161–187.

Bolster CH, Walker SL, Cook KL. 2006. Comparison of Escherichia coliand Campylobacter jejuni transport in saturated porous media. Journalof Environmental Quality 35: 1018–1025.

Bolster CH, Haznedaroglu B, Walker SL. 2009. Diversity in cell propertiesand transport behaviour among 12 different environmental Escherichia coliisolates. Journal of Environmental Quality 38: 465–472.

Bolster CH, Cook KL, Marcus IM, Haznedaroglu BZ, Walker SL. 2010.Correlating transport behaviorwith cell properties for eight porcineEscherichiacoli Isolates. Environmental Science and Technology 44: 5008–5014.

Bradford S, Bettahar M. 2005. Straining, attachment and detachment ofcryptosporidium oocysts in saturated porous media. Journal ofEnvironmental Quality 34: 469–478

Bradford S, Torkzaban S, Walker SL. 2007. Coupling physical andchemical mechanisms of colloid straining in saturated porous media.Water Research 41: 3012–3024.

Byamukama D, Kansiime F, Mach RL, Farnleitner AH. 2000. Determin-ation of Escherichia coli contamination with chromucult coliform agarshowed a high level of discrimination efficiency for differing fecalpollution levels in tropical of Kampala, Uganda. Applied andEnvironmental Microbiology 66: 864–868.

Cameron DR, Klute A. 1977. Convective-dispersive solute transport witha combined equilibrium and kinetic adsorption model. Water ResourcesResearch 13: 183–188.

Caroff M, Karibian D. 2003. Structure of bacterial lipopolysaccharides.Carbohydrate Research 338: 2431–2447.

Castro FD, Tufenkji N. 2008. Role of Oxygen Tension on the Transportand Retention of Two Pathogenic Bacteria in Saturated Porous MediaEnvironmental Science and Technology. 42: 9178–9183

Champ DR, Schroeter J. 1988. Bacteria transport in fractured rock- A fieldscale tracer test at the Chalk River Nuclear Laboratories. Water Scienceand Technology 20: 81–87.

Chen G, Rickhold M, Stevett KA. 2003. Equilibrium and kinetic sorptionof bacteria on alluvial sand and surface thermodynamic interpretation.Research in Microbiology 154: 175–181.

Corapcioglu MY, Haridas A. 1984. Transport and fate of microorganisms inporousmedia: a theoretical investigation. Journal of Hydrology 72: 149–169.

Corapcioglu MY, Haridas A. 1985. Microbial transport in soils andgroundwater: a numerical model. Advances in Water Resources 8: 188–200.

Dong H, Rothmel R, Onstott TC, Fuller ME, DeFlaun MF, Streger SH,Dunlap R, Fletcher M. 2002. Simultaneous Transport of TwoBacterial Strains in Intact Cores from Oyster, Virginia: BiologicalEffects and Numerical Modeling. Environmental and AppliedMicrobiology 68: 2120–2132

Ewing WH. 1986. Edwards and Ewing’s Identification of Enterobacter-iaceae. 4th Edition. Elsevier Publishing Company: New York; 130.

Copyright © 2013 John Wiley & Sons, Ltd.

Foppen JWA, Schijven JF. 2006. Transport and survival of Escherichiacoli and thermotolerant coliforms in aquifers under saturated conditions– a review. Water Research 40: 401–426

Foppen JWA, van Herwerden M, Schijven J. 2007a. Transport ofEscherichia coli in saturated porous media: dual mode deposition andintra-population heterogeneity. Water Research 41: 1743–1753.

Foppen JWA, van Herwerden M, Schijven JF. 2007b. Measuring andmodeling of straining of Escherichia coli in saturated porous media.Journal of Contaminant Hydrology 93: 236–254.

Foppen JW, Lutterodt G, Roling WFM, Uhlenbrook S. 2010. Towardsunderstanding inter-strain attachment variations of Escherichia coli duringtransport in saturated porous quartz sand.Water Research 44: 1202–1212.

Garabal JI, Gonzalez EA, Vazquez F, Blanco J, Blanco M, Blanco JE.1996. Serogroups of Escherichia coli isolated from piglets in Spain.Veterinary Microbiology 48(1-2): 113–123.

Guinée PAM, Agterberg CM, Jansen WJ. 1972. Escherichia coli OAntigen Typing by means of a mechanized micro technique. AmericanSociety for Microbiology 24: 127–131.

Havemeister G, Riemer R. 1985. Laborversuche zum transport-verhaltenvon bakterien. umweltbundesambt materialien, 2/85, 27-32

Haznedaroglu BZ, Bolster CH, Walker SL. 2008. The role of starvation onEscherichia coli adhesion and transport in saturated porous media.Water Research 42: 1547–1554.

Haznedaroglu BZ, Kim, HN, Bradford SA, Walker, S. 2009. RelativeTransport Behavior of Escherichia coli O157:H7 and Salmonellaenterica Serovar Pullorum in Packed Bed Column Systems: Influence ofSolution Chemistry and Cell Concentration., Environmental Scienceand Technology 43: 1838–1844

HowardG,Teuton J, LuyimaP,OdongoR. 2002.Water usage patterns in low-income urban communities in Uganda: implications for surveillance.International Journal of Environmental Health Research 12: 63–73.

Howard G, Pedley S, Barrett M, Nalubega M, Johal K. 2003. Risk factorscontributing to microbiological contamination of shallow groundwaterin Kampala, Uganda Water Research 37: 3421–3429.

Jacobs A, Lagolie F, Herry JM, Debroux M. 2007. Kinetic adhesion ofbacterial cells to sand: cell surface properties and adhesion rate. Colloidand Surfaces B Interfaces 59: 35–45.

Jenkins C, Chart H, Willshaw GA, Cheasty T, Smith HR. 2006.Genotyping of enteroaggregative Escherichia coli and identification oftarget genes for the detection of both typical and atypical strains.Diagnostic Microbiology and Infectious Disease 55: 13–19.

Kang G, Pulimood AB, Mathan MM, Mathan VI. 2001. Enteroaggrega-tive Escherichia coli infection in a rabbit model. Pathology 33: 341–346

Knobl T, Baccaro MR, Moreno AM, Gomes TA, Vieira MA, Ferreira CS,Ferreira AJ. 2001. Virulence properties of Escherichia coli isolated fromostriches with respiratory disease. Veterinary Microbiology 83: 71–80.

Kulabako NR, Nalubega M, Thunvik R. 2007. Study of the impact of landuse and hydrogeological settings on the shallow groundwater quality ina peri-urban area of Kampala, Uganda. Science of the TotalEnvironment 381: 180–199.

Kulabako NR, Nalubega M, Thunvik R. 2008. Phosphorus transport inshallow ground-water in peri-urban Kampala, Uganda: Results from fieldand laboratory measure-ments. Environmental Geology 53: 1535–1551.

Lee S-J, Chung S-G, Kim D-J, Lee CE, Choi J-W. 2009. New method fordetermination of equilibrium/kinetic sorption parameters. CurrentApplied Physics 9: 1323–1325.

Li X, Scheibe TD, Johnson WP. 2004. Apparent decreases in colloiddeposition rate coefficients with distance of transport under unfavorabledeposition conditions: a general phenomenon. Environmental Scienceand Technology 38: 5616–5625.

Lutterodt G, Basnet M, Foppen JWA, Uhlenbrook S. 2009a. The effect ofsurface characteristics on the transport of multiple Escherichia coli isolatesin large scale columns of quartz sand. Water Research 43: 595–605.

Lutterodt G, Basnet M, Foppen JWA, Uhlenbrook S. 2009b. Determiningminimum sticking efficiencies of six environmental Escherichia coliisolates. Journal of Contaminant Hydrology 110: 110–117.

Lutterodt G, Foppen JWA, Maksoud A, Uhlenbrook S. 2011. Transport ofEscherichia coli in 25m quartz sand columns. Journal of ContaminantHydrology 119: 80–88.

Lutterodt G, Foppen JWA, Uhlenbrook S. 2012. Transport of Escherichiacoli strains isolated from natural spring water. Journal of ContaminantHydrology 140-141: 12–20

Macler BA, Merkle JC. 2000. Current knowledge on groundwatermicrobial pathogens and their control. Hydrogeology Journal 8: 29–40.

Mallén G, Maloszewski P, Flynn R, Rossi P, Engel M, Seiler K-P. 2005.Determination of bacterial and viral transport parameters in a gravelaquifer assuming linear kinetic sorption and desorption. Journal ofHydrology 306: 21–36.

Hydrol. Process. (2013)DOI: 10.1002/hyp

G. LUTTERODT, J. W. A. FOPPEN AND S. UHLENBROOK

Matthess G, Foster SSD, Skinner AC. 1985. International Association ofHydrogeologists, Unesco., & International Union of GeologicalSciences. Theoretical background, hydrogeology, and practice ofgroundwater protection zones. Hannover, West Germany, HeiseISBN/ISSN: 3922705103.

Matthess G, Pekdeger A, Schroeter J. 1988. Persistence and transport ofbacteria and viruses in groundwater- a conceptual evaluation. Journal ofContaminant Hydrology 2: 171–188.

Merkli B. 1975. Untersuchungen ueber mechanismen und kinetik derelimination von bakterien und viren im grundwasser. Diss. ETH Zurich5420.

Morris BL, Lawrence ARL, Chilton PC, Adams B, Calow RC, KlinckBA. 2003. Groundwater and its susceptibility to degradation. A Globalassessment of the problem and options for management. Early warningand assessment report Series, RS. 03-3. United Nations EnvironmentProgramme/DEWA, Nairobi, Kenya.

Neihof R. 1969. Microelectrophoresis apparatus employing palladiumelectrodes. Journal of Colloid and Interface Science 30: 128–133.

Nsubuga FB, Kansiime J, Okot-Okumu J. 2004. Pollution of protectedsprings in relation to high and low density settlements in Kampala—Uganda. Physics and Chemistry of the Earth 29: 1153–1159

Nyenje PM, Foppen, JW, Uhlenbrook S, Kulabako R, Muwanga A. 2010.Eutrophication and nutrient release in urban areas of sub-Saharan Africa— A review. Science of the Total Environment 408: 447–455.

Pang L, Close M, Goltz M, Sinton L, Davies H, Hall C, Stanton G. 2003.Estimation of septic tank setback distances based on transport of E. coliand F-RNA phages. Environment International 29: 907–921.

Pembrey RS, Marshall KC, Schneider RP. 1999. Cell surface analysistechniques: What do cell preparation protocols do to cell surfaceproperties? Applied and Environmental Microbiology 65: 2877–2894.

Powell KP, Taylor RG, Cronin AA, Barrett MH, Pedley S, Sellwood J,Trowsdale SA, Lerner DN. 2003. Microbial contamination of two urbansandstone aquifers in the UK. Water Research 37: 339–352.

Powelson DK, Mills AL. 2001. Transport of Escherichia coli in sandcolumns with constant and changing water contents. Journal ofEnvironmental Quality 30: 238–245.

Rajagopalan R, Tien C. 1976. Trajectory analysis of deep-bed filtrationwith sphere-in-cell porous media model. AICHE Journal 22, 523–533

Redman JA, Walker SL, Elimelech M. 2004. Bacterial Adhesion andTransport in Porous Media: Role of the Secondary Energy Minimum.Environmental Science and Technology 38: 1777–1785.

Rijnaarts H, Norde W, Bouwer EJ, Yklema JI, Zehnder AJB. 1993.Bacterial adhesion under static and dynamic conditions. Applied andEnvironmental Microbiology 59: 3255–3265.

Ryan J, Elimelech M. 1996. Review: Colloid mobilization andtransport ingroundwater.Colloids Surf A: Physicochemical Engineering Aspects 107: 1–56.

Sayed AK, Foster JW. 2009. A 750 bp sensory integration region directsglobal control of the Escherichia coli GadE acid resistance regulator.Molecular Microbiology 71(6): 1435–1450.

Schijven JF, Hassanizadeh SM. 2000. Removal of Viruses by SoilPassage: Overview of Modeling, Processes, and Parameters CriticalReviews in Environmental Science and Technology. 30: 49–127

Schinner T, Letzner A, Liedtke S, Castro FD, Eydelnant IA, Tufenkji N.2010. Transport of selected bacteria pathogens in agricultural soil andquartz sand. Water Research 44: 1182–1192.

Copyright © 2013 John Wiley & Sons, Ltd.

Simoni SF, Harms H, Bosma TNP, Zehnder AJB. 1998. Populationheterogeneity affects transport of bacteria through sand columns at lowflow rates. Environmental Science and Technology 32: 2100–2105

Šimůnek J, Šejna M, Saito H, Sakai M, van Genuchten MTh. 2009. TheHYDRUS-1D Software package for simulating the one-dimensionalmovement of water, heat, and multiple solutes in variably-saturatedmedia version 4.08. Department of Environmental Sciences, Universityof California Riverside, Riverside California.

Sinton LW, Noonman MJ, Finaly RK, Pang L, Scott DM. 1997. Transportof bacteria and bacteriophages in irrigated effluent into and through analluvial aquifer. Water, Air, and Soil Pollution 98: 17–42.

Sinton LW, Finaly, RK, Pang L, Close M. 2000. Transport and attenuationof bacteria and bacteriophages in alluvial gravel aquifer. New ZealandJournal of Marine and Freshwater Research 34: 175–186.

Small P, Blankenhorn D, Welty D, Zinser E, Slonczewski J. 1994. Acidand Base Resistance in Escherichia coli and Shigella flexneri: Role ofrpoS and Growth pH. Journal of Bacteriology 176: 1729–1737.

von Smoluchowski M. 1921. Electrische Endosmose und Strömungs-ströme. In Handbuch der Elektrizität und des Magnetismus, (ed.).L Greatz Leipzig: Barth, Vol. 2 p 366.

StenutzR,WeintraubA,WidmalmG. 2006. The structures of Escherichia coliO-polysaccharide antigens. FEMS Microbiology Reviews 30(2006): 382–403.

Taylor RG, Howard KWF. 1998. The dynamics of groundwater flow in theregolith of Uganda. International Contributions to Hydrogeology 18: 97–113.

Tong M, Johnson W. 2007. Colloid population heterogeneity drives hyper-exponential deviation from classic colloid filtration theory. Environ-mental Science and Technology 41: 493–499

Tufenkji N, Elimelech M. 2004. Correlation equation for predictingsingle-collector efficiency in physicochemical filtration in saturatedporous media. Environmental Science and Technology 38: 529–536.

Tufenkji N, Elimelech M. 2005a. Breakdown of colloid filtration theory:Role of the secondary Energy minimum and surface charge hetero-geneities. Langmuir 21: 841–852.

Tufenkji N, Elimelech M. 2005b. Spatial distribution of Cryptosporidiumoocysts in porous media; Evidence of dual mode deposition.Environmental Science and Technology 39: 3620–3629.

Ulett GC, Webb RI, Schembri MA. 2006. Antigen-43-mediatedautoaggregation impairs motility in Escherichia coli. Microbiology152: 2101–2110.

van Bavel CHM. 1950. Mean weight diameter of soil aggregates asa statistical index of aggregation. Soil Science Society of AmericaJournal 14 No C: 20–23.

Walker SL, Redman JA, Elimelech M. 2004. Role of cell surfacelipopolysaccharides in Escherichia coli K12 adhesion and transport.Langmuir 20: 7736–7746.

Walker SL, Redman JA, Elimelech M. 2005. Influence of growth phase onbacterial deposition: Interaction mechanisms in packed-bed column andradial stagnation point flow systems. Environmental Science andTechnology 39: 6405–6411.

World Health Organization. 2004. World Health Report. http://www.who.int/whr/2004/en/report04 en.pdf

Yang H-H, Morrow JB, Grasso D, Vinopal RT, Dechesne A, Smets B.2008. Antecedent growth conditions alter retention of environmentalEscherichia coli isolates in transiently wetted porous media. Environ-mental Science and Technology 42: 9310–9316.

Hydrol. Process. (2013)DOI: 10.1002/hyp