,: .. BiotechnologY.
`Techniques
`· ~sl5
`-rz2
`
`Volume 13 No 8 August 1999
`
`ISSN 0951-208X
`
`· The journal of rapid publication and permanent record for methods and
`techniques that are new and generally useful for biotechnology in all its aspects
`
`Published in conjunction with Biotechnology Letters
`
`Colin Ratledge, Editor
`University of Hull} UK
`
`l
`i !
`i .
`
`I
`
`'
`I .,
`I
`
`Kluwer Academic Publishers I Dordrecht, Boston, London
`
`cooEN srEcE6
`
`BEQ 1018
`Page 1
`
`

`
`BIOTECHNOLOGY TECHNIQUES
`The journal of rapid publication and permanent record for methods and techniques that are new and genera ffy
`useful for biotechnology in aft its aspects.
`
`Editor
`Colin Racledge
`Tel: + 44 1482 465 243
`Departmenc of Biolog ical Sciences
`Fax: + 44 1482 465 458
`University of Hull
`Hull HU6 7R X, UK
`E.mai l: c. ratledge@ biosci.hull.ac. uk
`
`Publication programme, 1999 : Volum e 13 (12 issues).
`Subscription prices , per annum: NLG 1589 .00/USD 794.50 including postage and handling.
`Subscrip ti ons should be sent to K luwer Academic Publishers, P.O. Box 322, 3300 AH Dordrecht, The
`Netherlands or at P.O. Box 358, Accord Station, Hingham, MA 02018-0358, U.S.A., or to any subscription
`agent.
`For advertisemeM rates , pri c~s of bac k vofLilnes , and other informatio n, apply to Kluwer Academic Publishers,
`P.O. Box 17, '3300 AA D.ordredn;;;TI:le: Netherlands.
`
`Photocopying; In the U.S.A.: Thi s journ al is registered at the Copyrig ht Clearance Center, In c., 222 Rosewood
`Drive, Danved, MA 019 2~1
`A uth ori zation i:~"ph otocopy iteihs·for ihterilal or personal use, or the internal or personal use of specifi c cli encs,
`is granced by K luwer Academi c Publi shers for use rs reg iste red with the Copyrig ht Clearance Ce m er (CCC)
`Transac tional Reporting Service, provided that th e fee of USD 16.00 per copy is paid direccly to CCC. For those
`organizations that have been g ranced a photocopy li cence by CCC, a separate system of pay ment has bee n arranged .
`T he fee code for users of th e Transactional R eporting Service is 095 1-208X/99 USD 16.00.
`Author izati on does not extend to other kinds of copy ing , such as that for ge neral di stributi on, for advertising or
`pro mot ional purposes, for creating new coll ective works , or for resale.
`In the rest of the world: Permi ssion to photocopy must be obrained from the copy ri g ht owner. Please apply to D ordrec ht
`office .
`
`Consent to publish in this journal entails the author's irrevocab le and excl usive authori sati on of the publisher to
`coll ec t any sum s or co nsiderati ons for copy ing or reprod uct ion payable by third parcies (as mentioned in arricle 17,
`paragraph 2, of the Dutch Copy rig ht Act of 19 12 and in the Royal D ec ree of June 20 , 1974 (S .35 1) pursuant to
`article 16b of t he Dutch Copyrig ht Act of 19 12) and/or to act in or out of court in co nnect ion herewith.
`
`Biotechnology Techniqms is indexed /abst racted in Elsev ier BIOBASE/Current Awareness in Biolog ical Sciences,
`Cambridge Scientifi c Abstracts, Chemi cal Abstracts, Current Contents/Agricu lture, Biology and Env ironmenta l
`Science, Biological Abstracts/RRM , Chem ical Engineering Abstracts, Biotechnology Abstracts, Biotechnology
`Citatio n Ind ex, Sc iSearc h, The ISI A lerting Services, Science Citation Index Expanded and !BIDS.
`
`lviicrofilm and microfiche ed iti ons of this journ al are available fro m Un iversity Microfilms International , 300 North
`Zeeb Road, Ann Arbor, MI 48 106, U.S.A.
`© 1999 K lu wer Academic Publishers. Printed in th e Netherlands.
`
`BEQ 1018
`Page 2
`
`

`
`... Biorecllllologr Techniques 13 : 523-528. 1999.
`' ' © 1999 KhMer Acade111ic Publishers. Prin red in rile Ner!Jer/ands.
`
`523
`
`A probing feeding strategy for Escherichia coli cultures
`
`Mats Akesson 1, *, Per Hagander 1 & Jan Peter Axelsson2
`1 Depa rtment ofAutomatic Control, Lund In stitute of Technology, P 0. Box 11 8, SE-22 1 00 Lund, Sweden
`2?/w rmacia & Upjohn, Process R&D, SE- 1 12 87 Stockholm, Sweden
`*Author fo r co rrespondence (Fax: + 46 46 1381 1 8; E-mail: akesson @contro f.fth.se)
`
`Rece ived 26 April 1999; Revisions req ues ted 28 May 1999; Revisions received 16 June 1999; Accepted 16 Ju ne 1999
`
`Key words: acetate, di sso lved oxygen, Escherichia coli, feedback contro l, glucose feed in g
`
`Abstract
`
`A strain-independent feedin g strategy fo r fed-batch cultures of Escherichia coli is presented. By superimposing
`short pulses in the glucose feed rate, on-line detec ti on of acetate formati on can be made using a standard d issolved
`oxygen sensor. A simple feedback a lgorithm is then used to adju st the feed rate to avo id acetate fo rmation. The
`feas ibility of the strategy is demonstrated by both simulatio n and ex periments.
`
`Introduction
`
`Escherichia coli is a frequentl y used host organism
`fo r productio n of recombinant proteins. It has many
`advantages, such as be ing well charac teri zed and sup(cid:173)
`portin g growth to hi gh cell densities, but also draw(cid:173)
`backs. One of the difficulties encoun tered in E. coli
`cultivati ons is the formati o n of the metabolic by(cid:173)
`prod uct, acetate, in situ ati ons with excess glucose or
`under anaerobi c conditio ns. Acc umul ation of acetate
`has been reported to reduce both cell growth (Luli
`& Strohl 1990) and recombinant protein producti on
`(Bauer eta!. 1990, Bech Jensen & Carl sen 1990).
`The acc umul ation of acetate can be redu ced by
`proper cho ices of strain, medi a, and culti vation condi(cid:173)
`tions. In fed-batch cultures, where additi onal g lu cose
`is feel during the culti vation, the feed rate can be
`manipul ated to restrict formati on of acetate. A num(cid:173)
`ber of feeding strategies have been developed (Lee
`1996), however, the implementation is often co mpli(cid:173)
`cated due to the lack of cheap and reliable on-line
`sensors. Furthermore, most strateg ies also require con(cid:173)
`siderable process knowledge to work well . Typically
`a key process parameter, such as the criti cal growth
`rate above which acetate formati on occ urs, has to be
`know n a priori. This is not always the case, especi ally
`not under ex pression of a recombin ant protein when
`
`signi ficant changes in the process behav ior may occ ur,
`see fo r instance Qiu et al. 1998, Akesson et al. 1999.
`To address thi s problem, thi s paper presents a feed(cid:173)
`ing strategy based on standard instrumentatio n that
`req ui res onl y a minimum of process knowledge. The
`key idea is to ex pl oit a characteri stic change in the
`relati on between oxygen uptake and g lucose uptake at
`the onset of acetate form ati on. This makes it possible
`to obtain on-line detection of acetate formatio n us(cid:173)
`ing a standard di ssolved oxygen probe (Akesson eta!.
`1999). and a simple feedback algorithm is then used to
`adju st the glucose feed rate.
`
`Probing feeding strategy
`
`On- fine detection of acetate fo rmation
`
`Formati on of acetate, when E. coli is grow n under
`full y aerobic conditions, typically occurs when the
`spec ifi c g lucose uptake qg exceeds a criti cal value
`q~rit. The acetate production is thought of as an over(cid:173)
`fl ow phenomenon where flu x of acetyiCoA is d irected
`to acetate, via acetylphosph ate, instead of entering
`the TCA cycle. In batch and continuo us cultivations,
`it was observed that the spec ific oxygen uptake rate
`reached an apparent maximum at the onset of acetate
`fo rmati on (Andersen & von Meyenburg 1980, Reiling
`
`BEQ 1018
`Page 3
`
`

`
`524
`
`List of symbols
`
`Symbol
`
`Va lue/unit
`
`Description
`
`F
`G
`G;n
`H
`Ku1
`ks
`ka
`N
`No
`0 , Op
`ohigh· Otow
`0 ,,.,,
`O *
`c
`Cfa
`c, max
`Cfa
`p
`Cfa
`Cfg
`{j~l~lX
`cnt
`Cfg
`
`{/Ill
`
`(/me
`
`Cfo
`q;}1ax
`
`Tp
`v
`X
`
`Yag
`Yoa
`Yog
`
`Yom
`Y_~s-
`je
`Yxg
`
`Cl
`
`/1
`
`[I h- 1]
`[g ,- I]
`600 g ,- I
`14 X 103 (1%) g-l
`[h - 1]
`0.0 1 g ,- I
`0.05 g ,- I
`
`[rpm]
`150 rpm
`[%]
`[%]
`[%]
`100 %
`[g (gh) - 1]
`0.20 [g (gh)- 1]
`[g (gh)- 1]
`[g (gh)- 1]
`1.0 g (gh)- 1
`0.53 g (gh)- 1
`g (gh) - 1
`0.06 g (gh) -I
`[g (gh)- 1]
`0.30 g (gh)- 1
`20 s
`[I]
`[g ,- I]
`0.55 g g- 1
`0.55 g g- 1
`0.50 g g- 1
`1.07 gg- 1
`0.5 1 g g- 1
`O. ISgg- 1
`1.4 I /(h · rpm)
`[h- 1]
`
`G lucose feed rate
`Glucose concentrati on in reac tor
`Glucose concentration in feed
`Constant
`Volumetri c oxygen transfer coefficient
`Saturation constant for glucose uptake
`Saturati on constant for acetate upta ke
`Stirrer speed
`Constant
`Real and measured di ssolved oxygen tension
`Reacti on leve ls in control algorithm
`Setpoint for disso lved oxygen
`Disso lved oxygen tension in equi librium with ai r bubb les
`Specifi c acetate consumpt ion rate
`Maximum specific acetate co nsumpti on rate
`Specifi c acetate production rate
`Specific glucose uptake rate
`Maximum spec ific glucose uptake rate
`Critical specific glucose uptake rate
`Specific glucose consumpti on rate for maintenance
`Maintenance coeffi cient
`Spec ifi c oxygen uptake rate
`Maximum specific oxygen uptake rate
`Time constant for di ssolved oxygen probe
`Liquid volume
`Biomass concentration
`Acetate/glucose yield
`Ox ygen/acetate yie ld
`Oxyge n/glucose yie ld for grow th
`Oxyge n/g lucose yield fo r ma intenance
`Oxidat ive biomass/g lucose yield
`
`Fermentati ve biomass/g lucose yield
`Constant
`Specific grow th rate
`
`et a/. 1985). Corresponding relations between glu(cid:173)
`cose uptake, oxygen uptake and acetate formation are
`illustrated in Figure 1.
`Akesson et al. ( 1999) have demonstrated, theo(cid:173)
`reticall y and experimentally, how the characteri sti c
`change in the relation between oxygen uptake and
`glucose uptake can be used for on-li ne detection of
`acetate form ati on. Under glucose-limjted conditions
`pulses superimposed to the glucose feed rate give rise
`to changes in the glucose uptake. These changes imply
`variations in the oxygen uptake that can be seen in
`the dissolved oxygen measurement 0 p· For glucose
`
`uptakes above q z rit , the oxygen uptake is saturated
`and the oxygen response to feed pulses will change
`character. Figure l shows ideali zed pulse responses
`for qg below, at, and above q~rit. It is clear that the
`pulse responses reveal if q g is above or below q z rit ,
`and hence if acetate is being produced or not. Note
`that it is a qu alitative change that is detected, and
`conseq uently no values on stoichi ometri c coefficients
`or other process parameters have to be assumed or
`known.
`
`BEQ 1018
`Page 4
`
`

`
`525
`
`Fig. 2. Flow d iagram of the contro l algorithm .
`
`F
`
`...... J .. ':inc ..
`
`T pulse
`
`T control
`
`Fig. 3. Exa mpl e of a cyc le in the algorithm. Durin g the time
`Tcontrol betwee n two pu lses, 0 p is regu Ia ted by a P!D controller
`ma ni pu lating the stirrer speed.
`
`01ow
`
`have a short cyc le time. The cyc le time depends on
`how fast the oxygen set-point can be regained after a
`pul se and thu s hi gh performance is required in the PID
`loop.
`The algori thm has a few parameters that affect the
`performance of the feeding strategy. Simple tuning
`rules have been derived based on a lineari zed process
`model. The only process spec ific information that is
`required is an upper bound of the overall time constant
`in the dynamj cs from glucose feed rate to di sso lved
`oxygen measurement. Thi s quantity is mainl y eq uip(cid:173)
`ment related and can conveni entl y be determined once
`for each reactor setup. It is also ass umed that a well(cid:173)
`tuned controll er for the dissolved oxygen control is
`ava ilable and that an upper bound for the settling time
`in thi s loop is know n.
`
`Consumption. of acetate
`
`In the outline of the detection method, onl y the pro(cid:173)
`du cti o n of acetate has been considered but acetate may
`
`Fig. /. Re lations between glucose uptake, qg. oxygen uptake, {/o .
`and acetate productio n, qf:. Pulses in feed rate F and ideali zed
`responses in measured dissolved oxygen 0 p for q3 be low, at, and
`above qzrit
`
`Feedback algorithm
`
`The possibility to detect acetate formation on line is
`now used to adj ust the g lucose feed rate to ac hieve
`feed ing at q z rit. This choice avo ids acetate fo rmation
`as well as g lucose starvation.
`At each cycle of the algorithm a pulse is given to
`obtain informati on about acetate fo rmati on. The feed
`rate is then adjusted according to a set of ru les that
`can be described by the fl ow diagram in Figure 2. A
`reaction to an up pulse is said to occur if Op gets
`below a level 0 1ow during the pulse. Similarly, a re(cid:173)
`action to a down pulse is defined from O p exceeding
`OhioiJ. An exa mpl e of a cycle in the algorithm is shown
`in Figure 3 where also the algorithm parameters are
`defi ned.
`During the feed pulses, the stirrer speed in the re(cid:173)
`actor is fixed. Between two pulses, the stirrer speed is
`manipulated by a PID controller regulating Op. This
`ensures that 0 P is at the sa me level, O.,"fJ> at the start
`of each pulse. Moreover, thi s w ill avoid that the cell
`. growth becomes limited by oxygen. To achi eve good
`performance in the feed algorithm it is important to
`
`BEQ 1018
`Page 5
`
`

`
`526
`
`also be consumed with a concomjtant consumption of
`oxygen that could affect the oxygen response to a feed
`pulse. One could imagine a situation , for instance af(cid:173)
`ter a batch phase where acetate has accumu lated, with
`a sustained consumption of acetate so that the respi(cid:173)
`ratory system is close to saturation . This could then
`result in a reduced oxygen response to an up pulse in
`the glucose feed , and hence that thi s consumption of
`acetate is mi si nterpreted as production. In the feed(cid:173)
`back situation, however, the control action wou ld be a
`reduced feed rate, which may in fact be the appropriate
`action.
`
`Simulations
`
`The performance of the feeding strategy is first
`demonstrated in a simulation of a laboratory-scale fed(cid:173)
`batch culture of E. coli. The model is largely similar
`to the one presented in Xu et al. (1999) and describes
`cell growth , glucose uptake, oxygen uptake, as well
`as production and consumption of acetate. As the time
`scale of interest is in the order of seconds to minutes it
`is important to consider the dynamics of the disso lved
`oxygen probe. It is here modeled as a linear first or(cid:173)
`der system. The model equations are summarized in
`Appendix I.
`Figure 4 shows a simulation where the initial feed
`rate after a batch phase results in a specific glucose
`uptake qg that is below q~rit . The control algorithm is
`started and increases the feed rate until q~rit is reached.
`Then, the algorithm keeps qg approximately constant
`at q~rit as desired. After 2 h the value of q~rit is
`changed, and as can be seen the algorithm is able to
`detect and adjust the feed accordingly. The example
`illustrates the ability to achieve feeding at q~rit without
`a priori information and in spite of changes that may
`occur in the process.
`
`Laboratory-scale experiments
`
`Experiments with the probing feeding strategy were
`made in fed-batch cultures for production of two dif(cid:173)
`ferent xylanase domains in E. coli BL21 (DE3). Ex(cid:173)
`cept for the feeding strategy the medi a and cultivation
`conditions were as in Akesson et al. (1999). For each
`genetic construct good cultivation condition s were
`achieved at the first attempt. Concentrations of glucose
`and acetate were kept clearly below LOO mg 1- 1 and
`production levels of 5 g recombinant protein per liter
`harvest broth were obtained.
`
`qg [g/(gh}]
`
`OL-----~----~----~----~~----~----~
`2
`0
`0.5
`1.5
`2.5
`3
`
`0.5
`
`1.5
`
`2
`
`2.5
`
`3
`
`F [1/h]
`
`20
`0
`
`0.06
`
`0.04
`
`0.02
`
`1.5
`
`2
`
`2.5
`
`3
`
`0
`0
`
`0.5
`
`N[rpm]
`
`1200
`
`1000
`
`BOO
`
`0.5
`
`1.5
`
`2
`
`2.5
`
`3
`
`Time [h]
`
`Fig. 4. Simulation where the initial feed rate is chosen too low. Af(cid:173)
`ter two q~ri t is changed. The dotted lines indicate the critical glucose
`uptake, q~rit , and the reaction levels, Ohigh, Otow.
`
`The fed-batch part of one experiment is shown in
`Figure 5. Initially, the feed rate apparently gives a
`glucose uptake below the critical value for acetate for(cid:173)
`mation , and the control algorithm increases the feed
`rate. Around 6.2 h, q~rit is approached and the con(cid:173)
`troller makes a slight adjustment of the feed rate to
`avoid acetate formation. When the maximum stirrer
`speed, and hence also the max imum oxygen trans(cid:173)
`fer capacity, is reached after 7 h no further feed
`increments are made in order to guarantee aerobic
`conditions. The probing continues and the controller
`can thus reduce the feed rate if necessary to avoid
`acetate formation . Here it can be concluded from the
`di ssolved oxygen respon ses that no acetate formation
`occurs which is also confirmed by the off-line analy(cid:173)
`sis. In another experiment, the probing indicated that
`
`BEQ 1018
`Page 6
`
`

`
`527
`
`experiments demonstrate the fea sibility of the strategy
`and al so show that process changes can be handled.
`A major ad vantage with the probing strategy is that
`no a priori knowledge of strain- or construct-specific
`parameters are required. It can thu s be a useful tool
`in process development or in other situation s where
`it is important to get operational at first atte mpt. To
`further improve the applicability of the method appro(cid:173)
`priate safety nets will be developed. In particular, the
`handling of situations close to the maximum oxygen
`transfer capacity of the reactor should be investigated.
`
`6
`
`7
`
`8
`
`9
`
`10
`
`Acknowledgements
`
`40
`
`'V X [g/1]
`
`D Xyl (10-4kaUJ]
`
`w
`
`.. . \1 ; . .Y' ..
`D
`
`.. . .D . :
`
`.. : "\/ \1 \1
`
`\1
`
`..
`
`~ 1'i!' H
`
`'8'
`
`6
`
`7
`
`8
`
`9
`
`10
`
`The fin ancial support from Pharmacia & Upjohn is
`gratefully acknowledged . The authors would also like
`to thank Maher Abou-Hachem, Olle Hol st, and Eva
`Nordberg Karl sson at the Department of Biotechnol(cid:173)
`ogy, Lund University for invaluable ass istance during
`the experiments.
`
`200,-----,------,-------,------,------,,--,
`'V G [mg/1] D A[mg/1]
`
`150
`
`References
`
`100
`
`..
`
`&i'
`
`\1 ·
`.\1
`v Y ' <Jl
`rn
`
`[]]
`
`D D
`
`&i'
`
`\1
`D
`
`1'i!'
`
`-- ~-
`
`-~·-·· -~-
`
`6
`
`7
`
`8
`Time [h]
`
`9
`
`recombinant E.
`Fig. 5. Fed-batch part of ex periment with
`coli BL2 1 (DE3) . Xy l indi cates recombinant pro tein detected as
`enzyme activity.
`
`the critical glucose uptake was approached in the end
`of the production phase. The controller then reduced
`the feed rate and no acetate accumulation could be
`detected .
`
`Concluding remarks
`
`An automated feeding strategy for E. coli cultures
`based on standard instrumentation has been presented .
`The key element is that on-line detection of acetate
`formation can be obtained from the di ssolved oxy(cid:173)
`gen response to superimposed pulses in the glucose
`feed rate. This information is then used to adju st the
`feed rate so that both acetate accumulation and glu(cid:173)
`cose starvation are avoided. The simulations and the
`
`Akesso n M, Nordberg Karl sson E, Hagander P, Axe lsson JP, Tocaj A
`( 1999) On-line detectio n of acetate formati on in Escherichia coli
`cultures using di sso lved oxygen responses to feed tran sients.
`Biotechnol. Bioeng. , in press.
`Andersen K, von Meyenburg K ( 1980) Are growth rates of Es(cid:173)
`cherichia coli in batch cultures limited by respiration ? .!. Bac(cid:173)
`teriol144: 11 4- 123 .
`Bauer K, Ben-Bassat A, Dawson M, de Ia Puente V, Neway J
`( 1990) Improved ex press ion of human interl eukin-2 in hi gh(cid:173)
`ce ll -density fermento r cultures of Escherichia coli K- 12 by a
`phosphotran sacetylase mutant. Appl. Environ. Microbial. 56:
`1296- 1302.
`Bech Jensen E, Carlsen S ( 1990) Producti on of recombinant human
`growth hormone in Escherichia coli: ex press ion of different pre(cid:173)
`cursors and phys iological effects of glucose, acetate and salts.
`Biorechn ol. Bioeng. 36: I- ll.
`Lee SY ( 1996) Hi gh cell-density culture of Escherichia coli.
`TIBTECH 14: 98- 105.
`Luli GW, Strohl WR ( 1990) Co mpari son of growth , acetate pro(cid:173)
`duction , and acetate inhibition of Escherichia coli strain s in
`batch and fed-batch fermentati ons. Appl. Envim n. Microbial. 56:
`1004- 1011.
`Qiu J, Swartz JR , Georgiou G ( 1998) Express io n of acti ve hu(cid:173)
`man ti ss ue-type plasminogen acti vator in Escherichia coli. Appl.
`En viron. Microbial. 64: 4891 -4896.
`Reilin g HE , Lauril a H, Fiechter A ( 1985) Mass culture of Es(cid:173)
`cherichia coli: medium development for low and hi gh density
`cultivatio n of Escherichia coli 8/r in minimal and compl ex
`media . .!. Bioteclnw l. 2: 19 1- 206.
`Xu 8 , Jahi c M , En fo rs S-0
`( 1999) Modelin g o f overfl ow
`metaboli sm in batch and feel-batch culLUres of Escherichia coli.
`Biorechnol. Prog. 15: 8 1-90.
`
`BEQ 1018
`Page 7
`
`

`
`528
`
`Appendix I. Model equations
`
`c!V
`-= F
`ell
`
`cl(\1 X)
`- -= t-L·VX
`cit
`
`cl ( \1 G )
`- - = FGin -qg · VX
`ell
`
`ci(\IA )
`-
`-
`cit
`
`p
`c
`= (q0 -q0 )
`
`· VX ,
`
`*
`cl ( \10 )
`- 0 ) - q0
`- - = KLa (N) · V ( O
`cit
`
`· H\IX,
`
`clOp
`Tp-- + Op = 0 ,
`cit
`
`KLa(N) = a · (N - No) , N > No,
`
`G (G) = G max _G_
`k.,. + G,
`lg
`l g
`
`q"' = min (q3 , q""· ).
`
`( I )
`
`(2)
`
`(3)
`
`(4)
`
`(5)
`
`(6)
`
`(7)
`
`(8)
`
`(9)
`
`If G7 < G7cri t then
`'
`g
`g
`qf: = 0,
`
`ma.,
`(/ 0
`
`-
`
`l'
`)Y
`(
`q g -{jm
`og -{jm 0 111
`Yoa
`
`)
`
`·
`
`If G7 > G7crit then
`g- g
`'
`
`(
`
`P
`
`qa = Cfg - q3
`cri t) Y
`ag ,
`q~ = 0,
`
`max
`Cfo=[/0
`
`crit
`= (qg -q", ) Yng+qii,Yo/11 ,
`
`crit
`/.l = (qg
`
`cril
`ox
`-q", ) Y ,g+(qg -qg
`
`fe
`) Y , 8 .
`
`( lOa)
`
`( I I a)
`
`( 12a)
`
`( 13a)
`
`( JOb)
`
`( II b)
`
`( 12b)
`
`( 13b)
`
`BEQ 1018
`Page 8