A Metabolic Model for Biological

Phosphorus Removal by
Denitrifying Organisms
T. Kuba, E. Murnleitner, M . C. M. van Loosdrecht,* and J. J. Heijnen
Department of Biochemical Engineering, Delft University of Technology,
Julianalaan 67, 2628 BC Delft, The Netherlands

Received December 13, 1995/Accepted April 29, 1996

A metabolic model for biological phosphorus removal formation are derived from the conversion of glycogen
under denitrifying conditions has been established. The (Arun et al., 1989; Mino et al., 1987; Smolders et al.,
model is based on previous work with aerobic phospho-
1994a). In the aerobic zone oxygen is utilized as an electron
rus removal. The form of the kinetic equations used is the
same as for the aerobic model. The main difference is the acceptor for the oxidation of stored PHA, which leads to
value of PJNADH, ratio i n the electron transport phos- growth, phosphate uptake (polyp regeneration), and synthe-
phorylation with nitrate (S,,,). This value was determined sis of glycogen by the phosphorus-removing bacteria.
independently from batch tests with an enriched culture In conventional processes, oxygen is utilized as an elec-
of denitrifying phosphorus-removing bacteria. The mea-
sured S,,, was approximately 1.0 m o l ATP/mol NADH,. tron acceptor for phosphorus uptake. Nitrate is considered to
This indicates that the energy production efficiency with disturb the phosphorus removal process. However, from a
nitrate compared t o oxygen is approximately 40% lower. microbiological point of view there is no restriction against
These batch tests were also used t o identify a proper set utilization of nitrate instead of oxygen. Recently denitrify-
of kinetic parameters. The obtained model was subse-
quently applied for the simulation of cyclic behavior i n an
ing phosphorus-removing bacteria (DPB) have been found
anaerobic-anoxic sequencing batch reactor at different in activated sludge systems and laboratory cultures (Kerrn-
biomass retention times. The simulation results showed Jespersen and Henze, 1993; Kuba et al., 1993; Vlekke et al.,
that the metabolic model can be used successfully for the 1988; Wanner et al., 1992). In these bacteria nitrate is uti-
denitrifying dephosphatation process. The obtained ki- lized for the oxidation of stored PHA and is removed as
netic parameters for denitrifying enrichment cultures,
however, deviated from those obtained for the aerobic dinitrogen gas from wastewater. Thus besides phosphorus,

-
enrichment cultures. 0 1996 John Wiley & Sons, Inc.
Key words: phosphorus removal denitrifying dephos-
phatation stoichiometry metabolic model sequenc-
nitrogen, which also induces the eutrophication of surface
waters, is removed simultaneously by DPB. Previous re-
search has shown that DPB have similar potential for phos-
ing batch reactor
phorus removal (Kuba et al., 1993) and a similar biological
metabolism based on intracellular PHA and glycogen (Kuba
INTRODUCTION et al., 1994) as the phosphorus-removing bacteria in the
conventional anaerobic-aerobic ( N O ) processes.
In order to prevent surface waters from eutrophication, bio-
Several mathematical models of the phosphorus removal
logical phosphorus removal has been introduced in waste-
process are proposed, but only few studies have so far been
water treatment plants with activated sludge for several de-
made at modeling or denitrifying dephosphatation. For in-
cades. The biological phosphorus removal is caused by bac-
stance, phosphorus-removing bacteria in the Activated
teria which can accumulate phosphate as polyphosphate
Sludge Model No. 2 proposed by the IAWQ Task Group
(polyp) granules in the cells. To achieve phosphorus re-
does not incorporate denitrifying dephosphatation (Gujer et
moval, activated sludge is circulated through anaerobic and
al., 1995). Moreover, the model does not take the intracel-
aerobic zones. In the anaerobic zone where no electron ac-
lular glycogen metabolism into account.
ceptors (like oxygen and nitrate) are present, the phospho-
A structured metabolic model of the phosphorus removal
rus-removing bacteria take up lower fatty acids such as
process under conventional N O conditions has been pro-
acetic acid and store them as polyhydroxy-alkanoates
posed by Smolders et al. (1995a,b). The model is based on
(PHA; for instance, poly-P-hydroxy-butyrate or -valerate).
three anaerobic and five aerobic metabolic reactions, in-
The energy (ATP) for this process is derived from glycogen
cluding the glycogen metabolism. Smolders et al. showed
conversion and the stored polyp by hydrolysis and excretion
that their model is very well capable of describing the com-
of phosphate (van Groenestijn, 1987, 1989; van Veen et al.,
plex conversions of the biological phosphorus removal pro-
1993). The reduction equivalents (NADH,) for the PHA
cess. Their model, developed for the N O process, can be
modified for the denitrifying dephosphatation process, be-
* To whom all correspondence should be addressed. Telephone: 31-15- cause the anaerobic metabolism is completely identical in
2781618; fax: 3 1-15-2782355; e-mail: marK.vl@stm.tudelft.nl both processes, and the difference between the aerobic and

Biotechnology and Bioengineering, Vol. 52, Pp. 685-695 (1996)

0 1996 John Wiley &Sons, Inc. CCC 0006-3592/96/060685-11
anoxic metabolic reactions is only the electron transport aerobic value of 6 = 1.85 mol ATP/mol NADH, (Smolders
phosphorylation: et al., 1994b).
Aerobic: NADH, + 0.50, + GATP + H,O (1)
Anoxic PHB and Nitrate Conversion Rates
Anoxic: NADH, + 0.4HN0, + 6,ATP + 0.2N,
+ 1.2H,O (2) The linear relation for PHB or nitrate conversion rates as a
Smolders et al. (1994b) determined a value of P/NADH, function of growth, polyp formation, and glycogen synthe-
(P/O) ratio (amounts of produced ATP per oxidized sis is derived similarly as described by Smolders et al.
NADH,), 6, from the difference of measured oxygen con- (1994b) for aerobic phosphorus removal.
sumption rates in the absence and presence of phosphorus. From the assumption that no net accumulation of NADH,
In the same way, a value of the P/NADH, ratio under anoxic and ATP takes place, the overall stoichiometric equation of
conditions, 6 , can be determined. By using the obtained 6 , the PHB conversion rate under anoxic conditions, rphb,can
the metabolic model for the conventional aerobic process be expressed as follows, by using three PHB yield coeffi-
can be modified for denitrifying dephosphatation. cients ( l/Yphb.x, l/Yphb-pp, l/Yphb-gl) and one maintenance
In this research, the 6, and a proper set of kinetic param- coefficient (mphb):
eters are determined in batch tests with different phosphorus -)lphb = (l/yphb-x) . r x + (‘/yphb-pp) ’ ‘pp + (l’yphb-gl)
concentrations using enriched DPB sludge in an anaerobic- . rgl + mphb * cx (3)
anoxic (A,) sequencing batch reactor (SBR). The obtained
anoxic model based on the 6, is applied to denitrifying with
dephosphatation by DPB. Cycle behavior of phosphorus in
l/Yphb, = (0.635 + 2.24256,+ ~)/(2.256,+ 0.5) (4)
the A, SBR is simulated under two hfferent biomass reten-
tion time (SRT) conditions to verify the reliability and va- 1/YPh&,, = (6,/EN + 01,)/(2.256, + 0.5) (5)
lidity of the model. Also a comparison with the aerobic (26, + 1.5)/(2.256, + 0.5)
l/Yphb-gl = (6)
model and the limits of the present metabolic model will be
discussed. mphb = mapN/(2.256, + 0.5) (7)
Equation (3) describes the amount of required PHB for bio-
METABOLIC MODEL mass production (rJ, polyp (rpp), and glycogen synthesis
(rgl) and maintenance (mpbb). Yield and maintenance coef-
For the basic derivation of the metabolic model and the
ficients are defined by the metabolic model parameters ( 6 ,
exact reading of most of the reaction stoichiometry under
K, E,, a3,matpN)(Smolders et al., 1994b). The formulas of
denitrifying conditions we refer to Smolders et al. (1994a,b,
these yield coefficients are completely identical for aerobic
1995a,b), who developed a metabolic model for the con-
(Smolders et al., 1994b) and anoxic conditions, when 6 , E,,
ventional A/O process (called A/O model). This model was
and matp, are replaced by 6, E, and inatp.
modified for the denitrifying dephosphatation process
In the same way, the stoichiometric nitrate conversion
(called A, model).
rate under anoxic conditions, rno3,can be expressed by using
three nitrate yield coefficients ( 1/Yno3-x,1/Yno3-pp,l/Yno3-gl)
Anaerobic Metabolism and one maintenance coefficient (mno,):
The anaerobic metabolism is identical in both processes. -rnoj = (1/Yno3-x) . rx + (1/Yno,-pp> rpp + (1/Yno3-gJ
Three basic reactions are involved in the anaerobic metabo- . rgl+ mno, . Cx (8)
lism (Smolders et al., 1994a): (i) the transport and storage of
substrate (acetic acid, HAc) as poly-P-hydroxy-butyrate with
(PHB), (ii) polyp degradation, and (iii) glycogen degrada-
l/Yno,-x = (0.123 + 0.9~)/(2.256,+ 0.5) (9)
tion. These three processes are stoichiometrically coupled,
and only one rate equation is needed. l/Yno3-pp = (0.96,/~, + 0.901,)/(2.256,+ 0.5) (10)
l/Yno3-gl = 0.95/(2.256, + 0.5) (11)
Electron Transport Phosphorylation with Nitrate
m,,, - 0.9ma,/(2.256, + 0.5) (12)
Five basic reactions are involved in the aerobic metabolism
(Smolders et al., 1994b): (i) PHB catabolism, (ii) electron The formulas of these yield and maintenance coefficients
transport phosphorylation, (iii) production of biomass, (iv) are similar between aerobic (Smolders et al., 1994b) and
phosphate transport and polyp synthesis, and (v) glycogen anoxic conditions, but the aerobic coefficients are 5/4 times
synthesis. The difference between the aerobic and anoxic higher than the anoxic coefficients corresponding to the
metabolism is only electron transport phosphorylation with difference of available electrons per mole of electron ac-
oxygen or nitrate [Eqs. (1) and (2)]. The P/NADH, ratio ceptor for oxygen (four electrons) and nitrate (five elec-
under anoxic conditions, however, will be different from the trons).

686 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 52, NO. 6,DECEMBER 20, 1996
Kinetics constant in biomass synthesis (K), and the maintenance en-
ergy (mat,,,). In this anoxic metabolic model, the parameters
The conversion rates of HAc, phosphate, nitrate, PHB, and ag and K can be assumed equal to the aerobic parameters.
glycogen under A, conditions were simulated using the Smolders et al. (1994b) reported a3 = 1 mol ATP/mol P
same kinetic equations as proposed by Smolders et al. and K = 1.6 mol ATP/mol C(biomass). The other param-
(1995a,b). The kinetic equations are summarized in Table I. eters (E, matp,.S,) are determined by the following proce-
Under anaerobic conditions, the HAc uptake rate can be dure.
described by a Monod-type kinetic equation. From the stoi-
chiometry (Smolders et al., 1994a) (Table V), the PHB, Determination of the Phosphate Transport
glycogen, and phosphate conversion rate under the anaero- Coefficient
bic condition can be calculated from the HAc uptake rate. It Under aerobic or anoxic conditions, phosphate is taken up
was assumed that maintenance energy under anaerobic con- inside the cell and polyp is formed. The phosphate transport
dition is obtained from polyp degradation. The anaerobic into the cell is an energy-consuming process. The energy
maintenance coefficient (man) was obtained from the sec- required for the phosphate transport is generated by the
ondary phosphorus release rate (Kuba et al., 1993) when import of protons which are subsequently exported over the
both HAc and nitrate are absent. cell membrane in the oxidation of NADH, [Eq. (1) or (2)]
Under anoxic conditions, three kinetic equations for bio- (Mitchell, 1986). The ratio of energy consumption of the
mass synthesis, phosphate uptake, and glycogen synthesis phosphate transport under anoxic conditions (E,) to aerobic
and a maintenance coefficient were defined. From the stoi- conditions will be identical to the ratio of the PNADH,
chiometry [Eqs. (3) and (S)], the PHB and nitrate conver- value under both conditions:
sion rate under anoxic conditions can be calculated by using
these three kinetic equations.
E, = E SJS - (134
Smolders et al. (1994b) obtained S = 1.85 mol ATP/mol
In the simulation these kinetic equations were switched
NADH, and E = 7 mol P/mol NADH,, leading to
on and off with switching functions, depending on anaero-
bic or anoxic conditions. The switching functions were de- E, = 7.SJ1.85 ( 13b)
fined to be 1 when the concentration of a component is not
Determination of the Maintenance Coefficient
0; otherwise these switches are 0. For instance, under aero- under the Anoxic Condition (matpJ
bic conditions, the anoxic kinetic equations are inactive due
Smolders et al. (1994b) determined the ATP maintenance
to a switching function of nitrate which is equal to 0.
coefficient (mat,) from the endogenous oxygen consumption
rate. They found mat,, = 0.019 mol ATP/mol C * h. In this
DETERMINATION OF THE STOlCHlOMETRlC -
research, a value of 0.01 mol ATP/mol C h was chosen for
COEFFICIENTS OF THE A2 METABOLIC MODEL the A, system, because the anaerobic phosphate release for
maintenance (m,) for the A, system was found to be 50%
The yield coefficients for biomass synthesis (1/Y,hb-, of the value in the A/O system too.
1/Yno3J, polyp formation (l/Y,,hb.pp, l/Yno,-pp),and glyco-
gen synthesis (l/Yphb.gl, l/Yno3-gl) are now expressed as a Determination of the P/NADH2 Ratio under the
function of the PNADH, ratio (S,), the coefficient for the Anoxic Condition (6,)
NADH, requirement of phosphate transport (E,), the re- Determination of the P/NADH, ratio under anoxic condi-
quired ATP for polyp synthesis (a,),the polymerization tions (6,) can be achieved by combining measured conver-

Table I. Kinetic equations for anaerobic and anoxic reactions of biological phosphorus-removing organisms.

Reaction Equation Parameter Value Unit A/O value"

Anaerobic
HAc uptake 0.2 mol C h o l C.h 0.4
1.o mmol C/L 1.0
0.05 mmOVL -
Maintenance 2.5x 10-3 mol P/mol C.h 4 x lo-3
Anoxic
Biomass synthesis 0.05 mol C/mol C.h 0.14

Phosphate uptake 0.1 mol P h o l C.h 0.2

0.1 mmol P L 0.1
0.3 mol P h o l C 0.3
Glycogen formation 0.8 rnol C h o l C.h 0.8
0.3 rnol C h o l C 0.21
Maintenance 3.6 x 10-3 mol C/mol C.h 4x

"Reported values for organisms enriched in an anaerobic-aerobic SBR (Smolders et al., 1994b, 1995b)

KUBA ET AL.: METABOLIC MODEL OF DPB 687

sion rates of nitrate (rno,), biomass (rx),PHB (Ip+ and phorus source (15 mg PL), 0.227 g NH4Cl (60 mg NL),
polyp (rpp) in batch tests with the enriched denitrifying 0.6 g MgSO, * 7H,O, 0.07 g CaCl, * 2H20, 0.01 g ethylene
dephosphatation sludge from the following equation, which diamine tetraacetic acid (EDTA), and 2 mL nutrient solu-
is obtained from Eqs. (3) and (8) by elimination of the tion. The nutrient solution contained, per liter, 1.5 g
glycogen conversion rate (rgl): FeCl, * 6H,O, 0.15 g H,BO,, 0.03 g CuSO, * 5H,O, 0.03 g
= rnoj + (1/Yno3-J . r x + (1/Yno3-pp) . rpp
-
KI, 0.12 g MnCl, 4H,O, 0.06 g Na,MoO, * 2H20, 0.12 g
0 ZnSO, * 7H,O, and 0.15 g CoC1, 6H,O. -
- {rphb + ( yph&x) ' rx + ( Yphb-pp)
'rpp -tmphb * cx} . (Yphb-gl/Yno,-gl) + %o, ' cx (14)
Batch Tests for Determination of the P/NADH,
Because all conversion yield (Y)values and E, are linked to Ratio under Anoxic Conditions
6, [Eqs. (4)-(7) and (9)-(12) and Eq. (13b)l and because K,
a, and matp, are known, Eq. (14) only contains 6, and In order to determine the P/NADH, ratio in electron trans-
measured conversion rates (r). This equation then allows us port phosphorylation with nitrate @,) batch tests with vari-
to evaluate the best estimate for .6, able initial phosphate concentrations were conducted using
the enriched denitrifying dephosphatation sludge (SRT 8
MATERIALS AND METHODS days). The conversion rates of phosphate, nitrate, PHB,
glycogen, and NH,+ and the active biomass concentration
were obtained from five batch tests. The formula
Continuous Operation of the Anaerobic-Anoxic CH,,,Oo,~,No,,$,,o,, was assumed as the active biomass
Sequencing Batch Reactor formula (Smolders et al., 1995b). It was assumed that the
The study has been carried out in a laboratory-scale SBR measured MLVSS (mixed liquor volatile suspended solids
with a working volume of 3.5-3.6 L at room temperature concentration) is the sum of PHB, glycogen, and the active
(20-25°C). The SBR was operated in a cycle of 6 h. The biomass (C,) (Smolders et al., 1995b). The active biomass
cycle consisted of three phases: anaerobic (2.0 h), anoxic conversion rate (rx)was calculated from the measured NH,+
(3.5 h), and a settling period (0.5 h). Inoculum for the an- conversion rate (rnh,):
aerobic-anoxic (A2) SBR was obtained from a phosphorus
rx = -r nh4 /0.2 (15)
and nitrogen removal process operating on municipal waste-
water. Of the synthetic wastewater containing HAc and
phosphate, 1.75 L was pumped into the reactor during the For each batch test, 400 mL sludge was taken from the A,
first 10 min of the anaerobic phase. In the anoxic phase, 100 SBR at the end of the anaerobic phase. Thus, much PHB
was stored inside denitrifying phosphorus-removing bacte-
mL NaNO, solution (117 mmol/L) was pumped into the
ria (DPB), and little polyphosphate was stored due to an-
SBR with a constant flow rate during the first 110 min. At
the end of the anoxic phase, 64 or 112 mL of excess sludge aerobic phosphorus release with HAc consumption. After
was removed, resulting in a biomass retention time (SRT) of centrifugation (3500 rpm) of the sludge for 15 min, the
supernatant was replaced by a medium without HAc and
14 or 8 days, respectively. After settlement (15 min), 1.85 L
supernatant as treated effluent was pumped out from the phosphate, and the resuspended sludge was transferred into
a 1-L laboratory fermentor. After addition of proper
SBR. The overall hydraulic retention time was 0.5 days. The
amounts of phosphate and nitrate to the pretreated sludge,
SBR was continuously mixed by impellers at 450 rpm, ex-
the batch test was started. The initial phosphate and nitrate
cept for the settling period. Dinitrogen gas continuously
concentrations in the batch tests are summarized in Table 11.
flowed through the SBR head space with a flow rate of 60
All batch tests were conducted in the 1-L laboratory fer-
L h , to prevent introduction of oxygen in the system. The
mentor at 25°C using the pretreated DPB sludge. The pH
pH was strictly controlled at 7.0 f 0.5 by addition of 1 M
was controlled at 7.0-7.1 by addition of 0.5 M HC1 or
HCl or NaOH to avoid chemical phosphate precipitation.
NaOH. Dinitrogen gas was flushed through the head space
For the batch experiments, enriched sludge from the SBR
was used. of the fermentor with a flow rate of 30 L/h to prevent
oxygen contamination.
The A, SBR was operated under 14 days SRT for 10
months; thereafter the SRT was decreased to 8 days. After
steady-state conditions were achieved at each SRT, HAc, Table XI. Initial phosphate and nitrate concentrations in the batch tests.
phosphate, nitrate, NH,+, PHB, and glycogen concentra-
tions were regularly measured in a cycle. Initial P concentration Initial NO,- concentration
Run No. (=OW (rnrnollL)

Media 0.0 6.0

1.1 6.3
Synthetic medium was used containing, per liter, 0.375 g 2.1 6.4
1.1 6.2
HAc [400 mg chemical oxygen demand (COD)L] as car- 2.1 3.0
bon source, 0.049 g K,HP04 and 0.028 g KH2P04 as phos-

688 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 52, NO. 6, DECEMBER 20, 1996
Analyses fying conditions in comparison with aerobic conditions is in
the expected range (John and Whatley, 1970).
Orthophosphate was determined by the ascorbic acid The obtained value of 6, ( = 1.0) was used in the meta-
method. HAc was determined on a glass chromatograph bolic model. The conversion of each component was simu-
(GC) with a Hayesep Q 80-100 mesh column (185°C) and lated to verify the reliability and validity of the anaerobic-
a flame ionization detector (FID) (185°C). Nitrate was ana- anoxic model (A2 model).
lyzed on a liquid chromatography (LC) column (Skalar In-
struments) by Skalar methods. The NH,+ concentrations
were measured with an ammonia-selective gas electrode Stoichiometry under Anaerobic and Anoxic
(METROHM, 605060 10). Conditions for the A2 Model
For dry weight determination, a 20-mL sample of the From the obtained 6, ( = 1.O), all yield coefficients, Y [Eqs.
sludge was filtered on a Whatman glass microfiber (GF/C) (4)-(7) and (9)-(12)], can be calculated. The overall rela-
filter. The filter was dried for 24 h at 105°C and weighed on tions for PHB and nitrate conversion in the anoxic phase are
a microbalance. The organic content (MLVSS) was deter-
mined by incinerating the dry filters in an oven at 550°C +
-rphb = 1.628rx + 0.460rpp 1 . 2 7 3 + ~ ~ x 10-3Cx
~ 3.64
for 1 h. (16)
The extraction and analyses of intracellular PHB and gly- -r,,, +
= 0.568rX 0.414rpp + 0.346rg, + 3.27 x 10-3Cx
cogen were carried out as described by Smolders et al.
(17)
(1994a,b).
The yield coefficients for the metabolic reactions under an-
oxic conditions are given in Table V.
RESULTS

Simulation of the Conversion in the Batch Tests

P/NADH, Ratio under Anoxic Conditions
Simulations were performed with the software package
In order to determine the P/NADH, ratio (aN),the conver- AQUASIM (version 1.Ob) (Reichert, 1994). AQUASIM al-
sion of each component and the active biomass concentra- lows the stoichiometry and kinetic equations to be defined
tions were obtained from five batch tests with the enriched in relation to biological conversions, hydraulic effects, and
DPB sludge under anoxic conditions. Two data sets for the reactor configurations. Also parameter estimation and sen-
determination of 6, were prepared from the results in each sitivity analyses can be made.
batch test. One data set was the converted amounts of each Kinetic parameters under anoxic conditions were esti-
component which were calculated from the difference be- mated with AQUASIM (weighted least-squares estimation)
tween the initial and the final concentration. Another data by using the experimental results of the batch tests. The best
set was the initial conversion rates of each component. The estimated kinetic parameters under anoxic conditions are
data sets are shown in Tables I11 and IV. shown in Table I.
According to Eq. (14), 6, was calculated in each batch By using the kinetic (Table I) and stoichiometric equa-
test. As shown in Tables 111 and IV, the average of obtained tions [Eqs. (16) and (17)] and the obtained optimal kinetic
6, values was approximately 1.0 (mol ATP/mol NADH,) parameters under anoxic conditions, it is possible to calcu-
with a standard deviation of 0.19. Smolders et al. (1994b) late the conversion rate of each component in the batch
reported 1.85 (mol ATP/mol NADH,) under aerobic con- tests. Figure 1 shows the measured and calculated phos-
ditions. Therefore, the energy (ATP) production in electron phate, nitrate, NH,+, PHB, and glycogen concentrations in
transport phosphorylation with nitrate is approximately 40% the batch tests. The simulation results indicated that the
lower than with oxygen. The lower efficiency for denitri- model could properly simulate the measured conversion of

Table 111. The converted amounts of each component in the batch tests.

Run polyp NO,- PHB Biomass” Cxb 6, (mmol ATFV

No. (mmol/L) (mmol/L) (mmol C L ) (mmol CIL) (mmol C L ) mol NADH,)

0 0.00 -4.40 -1 1.89 2.15 35.77 0.87

1 0.95 -1.63 -4.92 2.15 32.25 1.41
2 2.51 -4.69 -12.34 2.10 30.20 1.00
3 3.42 -4.85 -1 1.65 1.80 24.10 1.03
4 1.19 -2.20 -5.47 1.50 3 1.40 0.97

Average 1.06

aThe active biomass conversion was calculated from the NH; conversion (rx = rnb/0.20).
bBiomass excluding polyp and carbon reserves.

KUBA ET AL.: METABOLIC MODEL OF DPB 689

 Table IV. The initial conversion rates of each component in the batch tests.

Run ‘PP ‘w rphb r, C, 6,(molATFV

No. (mm0VL.h) (mmol/L.h) (mmol C/L.h) (mmol C L h ) (mmol C/L) mol NADH,)

0 0.00 -1.79 -4.60 0.90 35.77 0.74

1 0.98 -1.74 -5.27 1.15 32.25 1.26
2 1.12 -2.17 -5.62 0.83 30.20 0.94
3 1.01 -1.45 -4.21 0.45 24.10 1.16
4 1.23 -2.27 -5.70 1.03 31.40 0.93

Average 1.01

each component at various initial concentrations of phos- DISCUSSION

phate.
P/NADH2 Ratio under Denitrifying Conditions
Application of the A, Model to Cycle Behavior in The P/NADH, ratio under anoxic conditions (S,,,) was in-
the A2 SBR dependently obtained from the batch tests with enriched
cultures of DPB. The 6, was 1.0 mol ATP/mol NADH,.
Phosphate, nitrate, NH,+,HAc, PHB, and glycogen concen- Smolders et al. (1994b) reported 1.85 under aerobic condi-
trations were measured in a steady-state cycle at 8 and tions. Therefore, the energy (ATP) production in electron
14 days SRT. The cycle behavior of each component was transport phosphorylationwith nitrate is approximately40%
simulated by the above-mentioned model, and the simula- lower than oxygen. This is one of the exceptional cases in
tion results were compared with the measured data sets of which the in vivo efficiency of denitrification is measured.
each component conversion under 8 and 14 days SRT. Un- John and Whatley (1970) measured the P/NADH, ratio in
der anoxic conditions, the calculation was done by using vitro with nitrate or oxygen by using membrane prepara-
the same kinetic equations, stoichiometric relations, and tions of Micrococcus denitrificans. They reported that the
kinetic parameters as derived from the batch tests. The ki- P/NADH, ratio was 0.9 with nitrate, while it was 1.5 with
netic parameters for the anaerobic conditions were obtained oxygen. The derived value for the S,,, of 1.0 and the 40%
from the cycle behavior by parameter estimation with lower efficiency compared with the PI0 ratio in this re-
AQUASIM. These are included in Table I. search are within the same range.
Now, all kinetic equations, stoichiometric relations, and
kinetic parameters under anaerobic and anoxic conditions
are specified. Identical kinetic and stoichiometric param- Sensitivities of the HAc Loading Rate and the
P/NADH, Ratio on the PHB Content
eters were used for the simulation under 8 and 14 days SRT.
The calculation was done until steady states were obtained In this research the power of metabolic modeling is shown.
(25 A, cycles = 150 h). The measured and simulated cycle The same model structure could be used under aerobic and
behaviors at each SRT are shown in Figures 2 and 3. denitrifying conditions, since the metabolism under both
The A, model could predict the concentration of all dis- conditions is similar. The A, model, which was modified
solved components (HAc, phosphate, nitrate, NH,+). How- from the N O model by using the measured 6, and with a
ever, a large difference between the measured and simulated proper set of kinetic parameters (Table I), could predict the
organic intracellular polymer (PHB and glycogen) concen- concentration of all dissolved components in the A, SBR at
tration was observed. different SRT conditions. A problem in the simulations is

Table V. Metabolic reactions under anaerobic and anoxic conditions.

Anaerobic condition
HAc uptake -CHzO - 0.5 CH1,y6O5/6- 0.36 HPO3 + 1.33 CH1.500.5 + 0.17 COZ + 0.36 H,PO4
+ 0.059 H,O = 0
Maintenance -HPO3 - HZO + H3PO4 = 0
Anoxic condition
Biomass synthesis -1.63 CH1.500.5 - 0.57 HNO3 - 0.2 NH3 - 0.015 H,PO4
+ CH,~osOo,54No~zJ’o,o15
+ 0.28 N, + 0.63 CO, + 0.78 H,O = 0
Phosphorus uptake -0.46 CHl.500.5 - 0.41 HNO3 - H3PO4 + HPO, + 0.21 N, + 0.46 CO,
+ 1.55 H,O = 0
Glycogen synthesis -1.27 CHl,500,5 - 0.35 HNO3 + CH,o/605/6 + 0.17 N z + 0.27 CO, + 0.29 HZO = 0
Maintenance -CHl.500,5 - 0.9 HNO3 + 0.45 N, + COZ + 1.2 HZO = 0

690 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 52, NO. 6, DECEMBER 20, 1996
 -- I
20-

".
- ------I-

O
0
L
1
- .
2
. . - . . 5.
3 4
. I
6
0 4 . .
0 1
I s
2
.
3
.
4
- . 5- .
*

Time [hours] (a) Time [hours]

04
0
. 1. . 2. . 3. . 4. . 5. . 6I
Time [hours] (b) Time [hours]

--------..----
,

04
0
. 1. . 2. . 3. . 4. . 5 . .
Time [hours] (c) Time [hours]

a
O J . .
0 1
. 2. . 3. . 4. . 5. . 6I 0
0 1 2 3 4 5
1
Time [hours] (d) Time [hours]

Figure 1. Phosphate (U),nitrate (O), NH,+ (A),PHB (+), and glycogen (x) concentrations in batch test: (a) run 0, (b) run 1, (c) run 2, and (d) run 3.
Lines are simulation results.

KUBA ET AL.: METABOLIC MODEL OF DPB 69 1

 = 5
>

Y
f
€ 4
21 3
m-
0
= 2
t
c
8'
c
144 145 146 147 148 149 150
Time [hours] Time [hours]
(a) (a)

r 3 5
3
530
E
E 25
U

g20
0
0
2 15
a
6 10
B
o " 5
g o144 145 146 147 148 149 150
Time [hours]
(b) (b)
Figure 2. (a) Phosphate (W), nitrate (A),and NH4+ (0) concentrations Figure 3. (a) Phosphate (B), nitrate (A),and NH4+ (0) concentrations
during anaerobic and anoxic phases at 8 days SRT.(b) HAc (W), PHB (A), during anaerobic and anoxic phases at 14 days SRT.(b) HAc (W), PHB
and glycogen (0) concentrations during anaerobic and anoxic phases at 8 (A),and glycogen (0) concentrations during anaerobic and anoxic phases
days SRT.Lines are simulation results. at 14 days SRT. Lines are simulations results.

that a large difference between the measured and simulated influent HAc concentration brought about approximately
PHB and glycogen concentration was observed. 15% and 35% differences on the PHB concentration under
The underestimation of PHB concentration on the model both SRT conditions (Fig. 4). The 2%-5% higher HAc input
can be due to two possible reasons: a high sensitivity for (i) can be caused due to an error in preparing the media and
deviating influent HAc or nitrate concentrations and (ii) the inaccurate flow rates of the influent pumps. The changed
6, obtained from independent batch tests. The main prob- influent HAc concentration had a large effect on the PHB
lem here is that a small error in the predicted PHB conver- concentration and slight effect also on the glycogen con-
sion in one cycle (or batch tests) accumulates to a large error centration but little effect on the concentration of dissolved
in the steady-state calculation. In fact, the error is enhanced components. The 5% higher influent HAc concentration re-
by a factor equivalent to the number of cycles in one sludge sulted in very good agreement on PHB between simulation
retention time (in these cases 32 or 56 times at 8 or 14 days and experiment in both SRT conditions. This effect can be
SRT), leading to large accumulation of such errors in the understood as follows. According to the kinetic model of the
estimation of storage components. This effect can be shown anoxic phase, the rate of growth, polyp synthesis, and gly-
by using slightly higher HAc loading rates in the simulation. cogen production stop if nitrate concentration is zero (Table
The results indicated that 2% and 5% differences of the I). This means that the PHB consumption also stops [Eq. (3)

692 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 52, NO. 6, DECEMBER 20, 1996
 35 I
Anaerobic I Anoxlc et
were recalculated in the batch tests and in the cycle behav-
ior. The model with 10% lower or higher 6, still predicted

/I
Nitrate
the measured conversion of each component in the batch
~
tests, which indicates that 6, has little influence on the
L estimation of kinetic parameters. The model with 10%
lower 6, (0.9) resulted in very good agreement of the cycle
behavior of PHB and also glycogen between the simulation
and the experiment in both SRT conditions. Figure 5 shows
the results of PHB and glycogen in 14 days SRT.
It is clear that the model structure can simulate the be-
5- havior of soluble components very well. The prediction of
the steady-state PHB content is, however, extremely sensi-
04 I I I ! I I , I I I I ! I tive to a small change in HAc and nitrate loading into the
144 145 Time147
146 [hours] '41 150 system or the value of 6,. Since the measurements of HAc
and nitrate loading always have a certain margin of uncer-
(a)

35
30

5"
-
-
E
m
15
2 10
5
0
144 145 146 147 148 149 150
Time [hours]

144 145 146 147 148 149 1

Time [hours]
(b)
Figure 4. Comparison of measured (a) PHB and (b) glycogen concen-
trations during anaerobic and anoxic phases at 14 days SRT, with calcu-
lation using three different HAc loading rates. HAc concentration in media 35
was increased by 0% (12.5 mmol C L ) , 2% (12.75 mmol C L ) , and 5%
(13.13 mmol C L ) , on calculation.
30
T25

or (16)]. If there is a slight imbalance, within one cycle, Yi 2 0

between the PHB produced in the anaerobic phase and con- 15
sumed in the anoxic phase, this imbalance will accumulate
over many cycles. A small change in the effect on this 8 10
imbalance of PHB will lead to a significant effect on the B 5
ultimate steady-state PHB level. It should be noted that the
model is in the same way also very sensitive to small errors 0
144 145 146 147 148 149 1
in the nitrate loading rate.
Time [hours]
The second possible reason for the underestimation of
PHB concentration can be sensitivity for the 6,. In this (b)
research, 6, was independently found from the batch tests
Figure 5. Comparison of measured (a) PHB and (b) glycogen concen-
with the enriched DPB sludge, not by parameter estimation. trations during anaerobic and anoxic phases at 14 days SRT, with calcu-
By using the same kinetic parameters and somewhat lower lation using three different P/NADH, ratios (8,): 0.9, 1.0, and 1.1 mol
or higher 6, ( = 0.9 or 1.1, instead of 1.O), the conversions ATP/mol NADH,.

KUBA ET AL.: METABOLIC MODEL OF DPB 693

tainty, the accurate prediction of PHB contents will always Overall Conversions during Anaerobic and
be cumbersome in the A, process. Anoxic Phases
The overall stoichiometry of the A, phosphorus removal
Comparing A2 and A/O Model process at 8 days SRT is shown in Figure 6. With 1 mol C
HAc, 1.33 mol C PHB is synthesized and 0.5 mol C gly-
The simulation results showed that the metabolic model cogen is anaerobically converted, while 0.36 mol P phos-
structure previously derived for the aerobic phosphorus re- phate is released. In the anoxic phase, 0.18 mol C PHB is
moval process can be used also for the denitrifying dephos- used for the synthesis of polyp and 0.70 mol C PHB is used
phatation process. However, several model parameters in for the synthesis of glycogen. Including maintenance, 0.40
each process obtained a strongly different value; see Table mol C PHB is used for biomass synthesis. In each cycle 0.16
I (qsmaX,k, kpp). From the aspect of modeling the waste- mol C active biomass, 0.05 mol C glycogen, and 0.05 mol
water treatment process, this makes the model complex, C PHB are removed as excess organic matter. The con-
because two sets of kinetic parameters are required for the sumed amount of nitrate for the synthesis of polyp is 0.16
phosphorus removal under anaerobic-aerobic and anaero- mol, for glycogen 0.19 mol, for biomass 0.09 mol, and for
bic-anoxic processes. From a physiological point of view, maintenance 0.13 mol.
the kinetic parameters could be expected to be similar in The reported output values of the conventional A/O phos-
both processes, since the metabolism is essentially the same. phorus removal process (Smolders et al., 1994b, 1995b) are
Microbiologically, it could be that the kinetic parameters also shown in Figure 6. The A/O process had the same input
deviate because the A, sludge contained a population dif- and SRT as the A, process. The biomass (MLVSS; the sum
ferent from the A/O sludge. In that case indeed the stoichi- of active biomass, PHB, and glycogen) produced in a cycle
ometry and kinetic structure should not be affected, but the is 0.26 mol C in the A, process, while it is 0.41 mol C in the
kinetic parameters can change strongly between different A/O process. Thus, in the A, process the overall biomass
enriched cultures. yield is much lower than the A/O process, indicating less
At present it is difficult to state whether the differences energy requirement for the excess sludge treatment process.
are due to different microbial populations present or an Since the A, process does not require oxygen for phospho-
intrinsic problem in the model structure. Future research is rus removal, clearly, energy for aeration can be saved in the
needed to resolve this question. A, process (nitrification then has to occur in a separate

Anaerobic In
1.OO mol-C HAc
0.04 mot-P PO4

II
\

out NO
0.16 blomass 0.34
lywgen degradation ? 0.04polyp 0.04
0.05 glycogen 0.06
0.05 PHB 0.01
0.75 C02 0.59
0.58 NO3 NO3
tf 0.29 N2 ------ N2
--I--

---- 02 0.55 0 2
/

0.15 maintenance

Anoxic
Figure 6. Conversions, expressed in moles, during the anaerobic and anoxic phase of the A, phosphorus removal process (8 days SRT) after addition
of 1 mol C HAc. Reported output values of the anaerobic-aerobic ( N O ) process (Smolders et al., 1994b, 1995b) (8 days SRT) are shown in parentheses.

694 BIOTECHNOLOGY AND BIOENGINEERING, VOL. 52, NO. 6, DECEMBER 20, 1996
phase). Moreover, in the A, process, 0.58 mol nitrogedmol p phosphate
pp polyphosphate
C HAc can be simultaneously removed from wastewater
s HAC
due to denitrification. x active biomass

CONCLUSIONS References
A metabolic model has been established for denitrifying Arun, V., Mino, T., Matsuo, T. 1989. Ietabolism of carboxylic acids
dephosphatation, based on the previous aerobic phosphorus located in and around the glycolytic pathway and the TCA cycle in the
removal model of Smolders et al. The form of the kinetic biological phosphorus removal process. Wat. Sci. Tech. 21: 363-374.
equations used was the same, but a major difference is the Gujer, W., Henze, M., Mino, T., Matsuo, T., Wentzel, M. C., Marais,
P/NADH, ratio. The measured PNADH, ratio in electron G. V. R. 1995. The activated sludge model no. 2: Biological phospho-
rus removal. Wat. Sci. Tech. 31: 1-1 1.
transport phosphorylation with nitrate was approximately John, P., Whatley, F. R. 1970. Oxidative phosphorylation coupled to oxy-
1.0 mol ATP/mol NADH,. This value indicates that the gen uptake and nitrate reduction in Micrococcus denitrificuns. Bio-
ATP production efficiency with nitrate is approximately chim. Biophys. Acta 216: 342-352.
40% lower than with oxygen. This is one of the first times Kerrn-Jespersen, J. P., Henze, M. 1993. Biological phosphorus uptake un-
der anoxic and aerobic conditions. Wat. Res. 27: 617424.
the relative efficiency of the oxidative phosphorylation with
Kuba, T., Smolders, G. J. F., van Loosdrecht, M. C. M., Heijnen, J. J.
different electron acceptors could be established in vivo. 1993. Biological phosphorus removal from wastewater by anaero-
Results of the simulation indicate that this model can bic-anoxic sequencing batch reactor. Wat. Sci. Tech. 27: 241-252.
predict the conversion of each dissolved as well as poly- Kuba, T., Wachtmeister, A,, van Loosdrecht, M. C. M., Heijnen, J. J. 1994.
meric component in the anaerobic-anoxic SBR operation Effect of nitrate on phosphorus release in biological phosphorus re-
moval systems. Wat. Sci. Tech. 30:263-269.
under different SRT conditions using a proper set of kinetic
Mino, T., Arun, V., Tsuzuki, Y., Matsuo, T. 1987. Effect of phosphorus
parameters. However, partly different sets of kinetic param- accumulation on acetate metabolism in the biological phosphorus re-
eters for denitrifying and aerobic phosphorus removal pro- moval process. Biological phosphate removal from wastewaters, pp.
cesses are obtained. 27-38. In: R. Ramadon (ed.), Advances in water pollution control, vol.
4. Pergamon, Oxford.
The authors thank S. van Hateren, C. Ras, D. C. Reuvers, G. van Mitchell, P. 1986. Chemiosmotic coupling and energy transduction. Glynn
der Steen, and M. Zomerdijk for the analytical work and P. Research, Bodmin, Cornwall, United Kingdom.
Kroon and S. Lispet for technical assistance in the experimental Reichert, P. 1994. Concepts underlying a computer program for the iden-
setup. This research was financially supported by NOVEM tification and simulation of aquatic systems. Schriftenreihe der
(351230/1110) and STOWA (RWZI 2000-3234/5). One of us EAWAG Nr.7, Swiss Federal Institute for Environmental Science and
(E.M., University of Agriculture Wien) participated in the re- Technology.
search due to a grant from the EC-ERASMUS program. Smolders, G. J. F., van der Meij, J., van Loosdrecht, M. C. M., Heijnen,
J. J. 1994a. Model of the anaerobic metabolism of the biological phos-
phorus removal process: Stoichiometry and pH influence. Biotechnol.
NOMENCLATURE Bioeng. 43: 461470.
Smolders, G. J. F., van der Meij, J., van Loosdrecht, M. C. M., Heijnen,
ATP required for uptake of HAc (mol ATP/mol C) J. J. 1994b. Stoichiometric model of the aerobic metabolism of the
ATP required for polyphosphate synthesis (mol ATPhol P) biological phosphorus removal process. Biotechnol. Bioeng. 44:
P/NADH, ratio (mol ATPhol NADH,) 837-848.
energy for transport of phosphate (mol P h o l NADH,) Smolders, G. J. F., van der Meij, J., van Loosdrecht, M. C. M., Heijnen,
biomass formation and polymerizationconstant (mol ATPhol C) J. J. 1995a. Structured metabolic model for anaerobic and aerobic
concentration of component i (mourn3) stoichiometry and kinetics of the biological phosphorus removal pro-
intracellular fraction of component i (mol/mol) cess. Biotechnol. Bioeng. 47: 277-287.
saturation constant of component i (moum3) Smolders, G. J. F., van Loosdrecht, M. C. M., Heijnen, J. J. 1995b.A meta-
rate constant of component i (h-') bolic model for the biological phosphorus removal process. Wat. Sci.
maintenance (mol/mol C h)
specific conversion rate of HAc (mol C h o l C * h) ., Deinema, M. H., Zehnder, A. J. B. 1987. ATP pro-
conversion rate of component i (mol/m3 * h) duction from polyphosphate in Acinetobucter strain 210A. Arch. Mi-
yield coefficient (moumol) crobiol. 148: 14-19.
van Groenestijn, J. W., Bentvelsen, M. M. A,, Deinema, M. H., Zehnder,
Subscripts and superscripts A. J. B. 1989. Polyphosphate-degrading enzymes in Acinerobucter
spp. and activated sludge. Appl. Environ. Microbiol. 55: 219-223.
anaerobic van Heen, H. W., Abee, T., Kortstee, G. J. J., Konings, W. N., Zehnder,
ATP A. J. B. 1993. Characterization of two phosphate transport systems in
poly-P-hydroxybutyrate Acinerobucter juhnsonii 210A. J. Bacteriol. 175: 2W206.
glycogen Vlekke, G. J. F. M., Comeau, Y., Oldham, W. K. 1988. Biological phos-
maximum phorus removal from wastewater with oxygen or nitrate in sequencing
anoxic batch reactors. Environ. Technol. Lett. 9: 791-796.
ammonia Wanner, J., Cech, J. S., Kos, M. 1992. New process design for biological
nitrate nutrient removal. Wat. Sci. Tech. 25: 445-8.

KUBA ET AL.: METABOLIC MODEL OF DPB 695