Docstoc

Optimization of Gene Regulatory Networks

Document Sample
Optimization of Gene Regulatory Networks Powered By Docstoc
					“God could cause us considerable
embarrassment by revealing
all the secrets of nature to us:

           we should not know what to do for
                 sheer apathy and boredom.”


                          -- Johann Wolfgang
                                  von Goethe
 Systems Biology of Osmotic
Shock in Antibody Producing
                  Cell Lines

                           Candidacy Proposal
                              Thomas R. Kiehl
               NSF Graduate Research Fellow,
      Multidisciplinary Science Ph.D. Program
      What is an Antibody?
• Antibodies are an important
  component of the body’s
  natural defenses.
• These glycoproteins recognize
  foreign substances and tag
  them for remediation by other
  parts of the immune system.
• mAb’s are an effective part of
  a growing number of medical
  treatments, lab techniques,
  diagnostics and imaging.



                                   Image source: Wikipedia
           Roche buys antibody technology
          company for $56.6 mln, Apr 2,2007
       ZURICH (MarketWatch) -- Swiss drugmaker Roche Holding AG (RHHBY) Monday said it has acquired
            privately-held Therapeutic Human Polyclonals Inc, an emerging biotechnology company focused
            on research in human antibody technologies, for $56.6 million in cash.
       Roche, based in Basel, said it plans to fully integrate THP, which is based in Germany and the U.S.,
  Monday (Roche) said it has acquired
            into its protein research center in Penzberg, Germany.
       "We are delighted about this acquisition as it builds on our strength in therapeutic antibodies," said
    privately-held Therapeutic Human
            Jonathan Knowles, head of global research at Roche.
       The development of therapeutic proteins and antibodies is an important area of research for the
       Polyclonals Inc, an emerging
            company, Roche said.
                                     on
   biotechnology company focused Roche itself paid $181 million
       At 0826 GMT, Roche shares were CHF1.80, or 0.8% higher, at CHF216.80, in a slightly lower broader
            market.
       research in human antibody last year to acquire GlycArt
       THP focuses on research in the field of human antibody technologies, where drugs made out of
            antibodies fight infectious agents, including bacteria and viruses, by seeking them out and
 technologies, for $56.6 million in cash.
            helping the body to destroy them.
                                     Biotechnology AG of Zurich,
       THP says it has developed a unique transgenic mammalian platform to create human antibodies. The
                                     which also had a crop of early-
            technology will enable the generation of both monoclonal and polyclonal antibody drugs with
            enhanced efficacy, Roche said.
                                            stage antibodies.
       Monoclonal antibodies are identical because they were produced by one type of immune cell and are
            all clones of a single parent cell.
In May, Merck & Co. (MRK) agreed to
       "Improved monoclonal antibody companies are hot commodities," said Denise Anderson,
            pharmaceutical analyst in Zurich with broker Kepler Equities, who has a buy rating on the
    pay a combined $480 million to
            stock, pointing to a string of deals over the past twelve months.
       Roche itself paid $181 million last year to acquire GlycArt Biotechnology AG of Zurich, which also
  acquire Abmaxis and GlycoFi, two
            had a crop of early-stage antibodies.
       In May, Merck & Co. (MRK) agreed to pay a combined $480 million to acquire Abmaxis and GlycoFi,
 biotechnology firms that brought the
            two biotechnology firms that brought the drug maker new methods to discover and produce
                                     Also
drug maker new methods to discover in the second quarter of 2006,
            drugs. Merck, based in Whitehouse Station, N.J., isn't affiliated with its German namesake.
       Also in the second quarter of 2006, Pfizer inc. (PFE) acquired Bioren, a small specialist in the
                                     Pfizer inc. (PFE) acquired Bioren,
            discovery of monoclonal antibodies.
           and produce drugs.
       "We think the deal makes good strategic sense for Roche, where top drugs Rituxan, Herceptin and
                                      a small specialist in the discovery
            Avastin are all antibodies, Anderson said.
       At a time when many traditional drugs made from small molecules are facing the loss of patent
                                          of monoclonal antibodies                                      .
            protection, medicines made out of large proteins are still protected from this threat not only
            because they've only entered the market over the past decade but also because they are more
            complex to imitate.
   2005 Market, $13 Billion
• ½ of that from just two drugs
   – Rituxan ($3.3Bn) – non-Hodgkin’s Lymphoma (CD20)
   – Remicade ($3.4Bn) - rheumatoid arthritis (TNF-α)
• 17 therapeutic monoclonal antibodies have
  received FDA approval and are on the market in
  the U.S.
• Several antibodies have been approved for use in
  diagnostic imaging applications.
• Report does not mention BMS’ Abatacept which
  is a fusion protein composed of an
  immunoglobulin fused to the extracellular
  domain of CTLA-4 (Sales for the second quarter
  of 2006 were $18 million, sales could reach US$
  1 billion by 2009/2010, )
        Market Report:   Monoclonal Antibodies: From Magic Bullets to Successful Drugs
             Abatacept: Nature Reviews Drug Discovery 5, 185-186 (March 2006)
   Herceptin, A prototypical
    Antibody Therapeutic
• This mAb targets a receptor which is
  over expressed in certain breast cancers
  (Bange 2001, Sliwkowski 1999).
• Herceptin targets the epidermal growth
  factor receptor, HER2, which is part of
  the ErbB family of tyrosine kinases.
• This targeting results in cell cycle arrest
  and suppression of tumor growth.
 Systems Biology of Osmotic
Shock in Antibody Producing
                  Cell Lines

                           Candidacy Proposal
                              Thomas R. Kiehl
               NSF Graduate Research Fellow,
      Multidisciplinary Science Ph.D. Program
 How do you make mAb’s?
• In 1975 Köhler and Milstein first
  developed cell lines which could reliably
  produce monoclonal antibodies
• These cell lines, known as hybridomas,
  were a fusion of an antibody-secreting
  murine lymphocyte cell with an murine
  myleoma cell.
• From this process emerges an
  immortalized cell line which secretes
  identical antibodies that have been
  raised against a specific antigen.
 Systems Biology of Osmotic
Shock in Antibody Producing
                  Cell Lines

                           Candidacy Proposal
                              Thomas R. Kiehl
               NSF Graduate Research Fellow,
      Multidisciplinary Science Ph.D. Program
                                                                   Why Osmotic Shock?
                                                  • Osmotic stress as well as a number
                                                    of other stresses can increase the
                                                    antibody production rates of a
                                                    culture
                                                  • Just add NaCl.
                          2.0×10 -06
Antibody Secretion Rate




                          1.5×10 -06
      (g/cell/hr)




                          1.0×10 -06


                          5.0×10 -07          control
                                              osmotic
                          0.0×10 -00
                                    12   24      36     48   60   72   84   96
                                                  Culture Time (hrs)
                                                                            Sun, Z., Zhou, R., Liang, S., McNeeley, K.M., Sharfstein, S.T. (2004) Biotechnology Progress. 20, 576-589
                                                                                     Ozturk, S.S., Palsson, B.O. (1991) Biotechnology and Bioengineering, Vol. 37, Pp. 989-993
    Is it really that easy?
• Higher osmolarities negatively
  impact viable cell concentration.




       Sun, Z., Zhou, R., Liang, S., McNeeley, K.M., Sharfstein, S.T. (2004) Biotechnology Progress. 20, 576-589
                Ozturk, S.S., Palsson, B.O. (1991) Biotechnology and Bioengineering, Vol. 37, Pp. 989-993
  So, just shock them a little.
             Right?
• In fed-batch cultures osmolarity becomes
  problematic both due to the addition of
  nutrients as well as the production of waste
  products, primarily lactic acid.
• Lactic acid acidifies the culture, necessitating
  the addition of base to control the pH.
• Over the course of a fed-batch culture the
  osmolarity can increase from ~290mOsm/kg to
  500mOsm/kg (Zhu 2005).
• Viability can be reduced by as much as 50%
  (Kurano 1990).
 Systems Biology of Osmotic
Shock in Antibody Producing
                  Cell Lines

                           Candidacy Proposal
                              Thomas R. Kiehl
               NSF Graduate Research Fellow,
      Multidisciplinary Science Ph.D. Program
         Systems Biology


“I am a Biologist, and I work on systems.
 I guess that makes me a Systems Biologist.”
                     -Howard Berg, ICSB 2005
         Systems Biology
“To understand biology at the system
  level, we must examine the structure
  and dynamics of cellular and organismal
  function, rather than the characteristics
  of isolated parts of a cell or organism.
  Properties of systems, such as
  robustness, emerge as central issues,
  and understanding these properties may
  have an impact on the future of
  medicine.” – Hiroaki Kitano

          Kitano, H. (2002), Systems Biology: a brief overview, Science, 295:1662-1664
3 C’s of Systems Biology


 • Complexity
 • Computation
 • Cross-Disciplinary Cooperation
Systems Biology

Lab Experiment(s)




                          Refine model




In-Silico Experiment(s)
 Systems Biology of Osmotic
Shock in Antibody Producing
                  Cell Lines

                           Candidacy Proposal
                              Thomas R. Kiehl
               NSF Graduate Research Fellow,
      Multidisciplinary Science Ph.D. Program
             Objective
• Engineer mammalian cells for
  optimal recombinant protein
  production.
  – To build a model of the cellular
    response to osmotic shock.
    • Characterize the response in terms of
      some specific components.
                    Overview
•   Mammalian Pathway
•   Yeast Model
•   Model Scope
•   Sample Model
•   TonEBP/NFAT5/OREBP

• Experimental Plan & Preliminary Results

• Related Efforts
    –   Batch Culture Model
    –   Microarrays
    –   CoEPrA
    –   Evolutionary Computing
          Mammalian Pathway




Dmitrieva, N. I., M. B. Burg, et al. (2005). "DNA damage and osmotic regulation in the kidney" Am J Physiol Renal
                                                 Physiol 289(1): F2-7.
Yeast Osmostress Signalling
   Simulating Yeast Response to
          Osmotic Shock




Klipp, E., B. Nordlander, et al. (2005). "Integrative model of the response of yeast to osmotic shock." Nature Biotechnology 23(8):
                                                                975-982.
              Yeast Model
                       r
                     nij v j (i  1,.., m)
        dci
              dt      j 1


• The ODEs in Klipp’s model generally take
  the form of equation 4. In this
  formulation m is the number of
  biochemical species, r is the number of
  reactions each with a rate v and
  stoichiometry n. This equation governs
  how the concentration of each species
  evolves over time.
Yeast Output
            Yeast Model
• Klipp showed that the pathway can be
  activated again by an additional shock.
• They also showed that this reactivation
  would not be possible if the pathway
  were structured such that the
  phosphatases provided the primary
  feedback control.
• They demonstrated that the gene
  transcripts for phosphatases should not
  increase by more than two-fold.
          Mammalian Pathway




Dmitrieva, N. I., M. B. Burg, et al. (2005). "DNA damage and osmotic regulation in the kidney" Am J Physiol Renal
                                                 Physiol 289(1): F2-7.
             Model Scope
An initial model will capture three main
  concepts.
• The insult of osmolarity within the
  context of the cell culture life-cycle
• The dependence of TonEBP activation on
  osmolarity
• TonEBP-dependant osmolyte
  accumulation.

                             Osmolyte
      Osmolarity   TonEBP
                            Accumulation
      Refined Objective
• Experimentally demonstrate the
  central role of NFAT5 in our cell
  lines the cellular osmotic response.
• Build a model to characterize that
  role
  – What portion of the osmotic response
    can be accounted for solely by TonEBP?
  – Are other factors or feedback loops
    required to explain observed dynamics?
Toward a simplified model
                                                                               Osmolyte
                Osmolarity                         TonEBP
                                                                              Accumulation




Dmitrieva, N. I., M. B. Burg, et al. (2005). "DNA damage and osmotic regulation in the kidney" Am J Physiol Renal
                                                 Physiol 289(1): F2-7.
               Osmolarity
• This is the primary independent
  variable in the system
  – Could be modeled in terms of a
    rapidly decreasing osmotic gradient
  – Could be kept at a constant
  – Could be modeled as a slowly
    increasing quantity.

                             Osmolyte
      Osmolarity   TonEBP
                            Accumulation
                   TonEBP
• First dependant variable, primarily
  dependant on the osmolarity
  – Goal is to fit this quantity to
    experimental data




                              Osmolyte
      Osmolarity    TonEBP
                             Accumulation
  Osmolyte Accumulation
• We presume that osmolyte
  accumulation is dependant on
  TonEBP activation
• We’ll use a proxy of cell volume
  initially.


                            Osmolyte
     Osmolarity   TonEBP
                           Accumulation
             Basic Model
          dO                          (a)
               k 0 Ot 1
          dt
          dN
              k1Ot 1  k 2 N t 1   (b)
          dt
          dP
              k 3 N t 1             (c)
          dt
• O, the osmotic gradient. The kinetic constant,
  kO, governs the rapid equilibration of this
  gradient immediately after the osmotic shock.
• N, amount of activated transcription factor
• P, the amount of accumulated osmoprotectants.
• k1 relates the activation of TonEBP to the
  osmolarity (O).
• k2 is a decay rate for activated TonEBP
• k3 relates TonEBP activation to osmolyte
  accumulation
                                    Model Output
– Osmotic gradient, blue.
– Level of activated NFAT5, red.
– Accumulation of osmolytes, green
                                                     Time course
                         1.2


                           1


                         0.8
   Relative Quantities




                         0.6


                         0.4


                         0.2


                           0


                         -0.2
                                0   20   40   60   80     100   120   140   160   180   200
                                                    Unitless Time
                 Iterate on the model
             • Generally fits with what we expect
             • Missing some important features
             • Must relate the model to actual
               data.
                                                                                         Time course
                                                             1.2


                                                               1


                                                             0.8




                                       Relative Quantities
                         Osmolyte                            0.6

Osmolarity     TonEBP
                        Accumulation                         0.4


                                                             0.2


                                                               0


                                                             -0.2
                                                                    0   20   40   60   80     100   120   140   160   180   200
                                                                                        Unitless Time
        Experimental Plans,
            Initial Data
•   Osmotic stress protocol
•   Quantify TonEBP
•   Quantify Cell Volume
•   Other experimental possibilities
  Osmostress Experiment
• Stress cells with 100mOsm increase
• Sample Cells at
  – Pre-stress Control
  – Post-stress 5, 10, 15, 30, 60 & 120 min
• For western blot:
  – Lyse in SDS and shear DNA
  – Use lysate in chemoluminescent or
    fluorescent western blot.
                   NFAT5 DNA Binding
• Consensus Sequence
  – TGGAAANN(C/T)N(C/T) [1]
    • N = any nucleotide
    • C/T = any pyrimidine
• NFAT Family, but similar
  to an NF-kB




       1) Miyakawa H, Woo S K, Dahl S C, Handler J S, Kwon H M. Proc Natl Acad Sci USA. 1999;96:2538–2542. [PubMed]
                        2) <image> James C. Stroud et al Nature Structural Biology 9, 90 - 94 (2002)
         About TonEBP
• Western blot of TonEBP after 18
  hours of incubation in isotonic (I)
  and hypertonic (H) medium
  (Miyakawa 1999)
         About TonEBP
• Localization of TonEBP under
  different mutations of the nuclear
  location signal (Tong 2006).
         About TonEBP
• Ratio of TonEBP
  localization after
  200, 300 or 500
  mosmol solution
  for 30 minutes
     (Zhang 2005)
              TonEBP
• We intend to use a chemiluminescent
  EMSA to watch TonEBP activation over
  time
• Previous work (Stroud 2002, Kojima
  2004)
                         Cell Size
• Intend to quantify with the FACS
  machine using forward light
  scattering techniques




         Ozturk, S.S., Palsson, B.O. (1991) Biotechnology and Bioengineering, Vol. 37, Pp. 989-993
    Other measurements
• As time allows
  – Upstream signaling components
  – Specific osmolyte accumulation
  – Lactic acid production
        GPC & Lactate
• Glycerophosphocholine and
  Lactate can both be quantitated by
  YSI




                               Lactic acid
             Betaine
• Near IR spectroscopy
     Sorbitol and Inositol
• Observe dehydrogenase activity by
  spectrophotometry
  – Sorbitol Dehydrogenase and Inositol
    dehydrogenase respectively
 Aldose Reductase Activity
• Spectrophotometry, absorbtion at
  340 (Bagnasco et al., PNAS
  84:1718) (JBC 1965 page 877)
            PKA & Fyn
• PKA by ELISA, from Stressgen
  Bioreagents (already attempted with a
  kit from Omnia, need to further
  optimize)
• Fyn immuniprecipitation following Ko
  et. al from JBC vol 273 pp 46083
           P38 MAPK
• Chemiluminescent Western from
  Cell Signaling Technologies
                                         MAPK
OKT3                    High                  Low
 30’ C   5’   10’ 15’          30’ 60’    120’ 30’
      SAPK/JNK, HSP27
• Chemiluminescent Western from
  Cell Signaling Technologies


          sapk/jnk
  SAPK/JNK Initial Results
• Initial Experiment




• Currently replicating this work to see if
  we can get better resolution
      Refined Objective
• Experimentally demonstrate the
  central role of TonEBP in our cell
  lines the cellular osmotic response.
  – EMSA for TonEBP, FACS for size
  – Westerns, ELISA & Spectrophotometry
    as time and resources allow
• Build a model to characterize that
  role, informed by experimental data
                              Osmolyte
       Osmolarity   TonEBP
                             Accumulation
           Other Efforts
•   Microarray Analysis
•   Batch Culture Model
•   CoEPrA
•   Evolving Bifurcating Networks
     Microarray Analysis
• Looking at network component
  analysis (NCA)
• Conceptualized some other SVM
  related approaches with Prof.
  Embrects (DSES)
Batch Culture Model




                      (Gao 2007)
                  CoEPrA
Comparative Evaluation of Prediction Algorithms

• “Primitive” Linear Algebra
  approach Placed 8th out of 16
  participants on a classification
  task.
• Paper submission invited.
• Task was to classify short peptides
  (8-9 amino acids) so as to predict
  activity.

                http://www.coepra.org
                 Method
• Our method utilized a simple mechanism of
  computing distances between LOGO’s
  generated for each sequence and each class of
  sequences (Crooks 2004).
• We used a random search algorithm to identify
  active nonapeptides in the prediction set.
• Random subsets of the joint calibration-
  prediction superset were compared with the
  active calibration subset. The retained loss
  function is the Frobenius matrix norm of the
  difference between the logos.
• One thousand runs were completed and results
  were pooled together to make the final
  prediction.
                                                                         Logos
Figure 1. Logo for whole calibration data set.
                                                                Shown in figures 1-4 are visual
                                                                  representations of the
                                                                  Logos in question. The
                                                                  search algorithm seeks out
                                                                  a partitioning of the
Figure 2. Logo for negative examples in calibration data set.     prediction data set (4). An
                                                                  optimal partitioning would
                                                                  yield a positive and
                                                                  negative subset of the
                                                                  prediction data set such
Figure 3. Logo for positive examples in calibration data set.     that their logos would show
                                                                  a minimal distance to the
                                                                  respective calibration logo
                                                                  (2 or 3).

Figure 4. Logo for prediction data set
Evolving Bifurcating Networks
• A good body of literature has
  started to form in the area of
  Evolving Biochemical Reaction
  Networks.
• Looking to build on previous work
  to create networks with specific
  distributions of outputs
Evolving Bifurcating Networks




    1   2   3   4   5   … 34 35
    1   2   3   4   5   … 34 35


    1   2   3   4   5   … 34 35
"Evolving Synthetic Biochemical Reaction Networks: First Steps" , ICSB St. Louis, MO, 2003, Kiehl T.R., Bonissone P.P.
Bioinformatics. 2004 Feb 12; Kiehl et al. 20(3):316-22
          Thanks.
•   NSF GRF
•   Susan Sharfstein
•   Lealon Martin
•   Sam Wait
•   David Isaacson
•   Joyce Diwan
•   Mark Embrechts
•   Numerous folks @ GE
•   Charles Bergeron
•   Duan Shen
•   Family & Friends
Ongoing work




               the end
the end
Osmotic Shock
    TonEBP Quantitation
• Chemiluminescent EMSA
• Can’t use generic NFAT kits, since
  TonEBP (NFAT5) is very different
  from other NFAT’s. More like some
  NFKappa’s.
     Antibody Production
• How does one stimulate production
  and maintain cell viability, thereby
  increasing specific productivity?
• Various types of stress are used to
  stimulate production, including
  Osmotic stress.
• What mechanisms are responsible
  for this response?
                                                               Batch Culture Timeline
                                                               days                    minutes                         hours                 days
                                                         Exponential                                                                       Stationary
                                                           Growth                      Osmotic                    “Adaptation”              Phase &
                                                           Phase                        Shock                                              Cell Death



                          2.0×10 -06
Antibody Secretion Rate




                          1.5×10 -06
      (g/cell/hr)




                          1.0×10 -06


                          5.0×10 -07           control
                                               osmotic
                                   -00
                          0.0×10
                                     12   24      36      48   60   72   84   96
                                                   Culture Time (hrs)



                                                                                   Ozturk and Palsson Biotech. Bioeng. 37:989-993 (1991)
     Modelling Response to
        Osmotic Shock
• Incorporate the acquired data,
  along with data from literature to
  into a computational model
• Following Klipp et al in their yeast
  model
                              LP & NCA
                              Ê = Â Pbar
                                                           Pbar(:,j) This
                                                          column is our
   Ê(i,j) : Our target                                   set of variables
    value and error
                                  =
   tolerance define
      constraints                                    Â(i) : This row
                                                     held constant

 Ê(r,j) : Used to define      Minimize Â(i,:)P(j)
“secondary” constraints       s.t.
                                      Â(i,:)P(j) ≤ target + ε1
Ê(i) : Each element in this
                                      Â(i,:)P(j) ≥ target – ε1
    row presents an LP        For r != i , 1 ≤ r ≤ length(Â)
independent of the other              Â(r,:)P(j) ≤ initial value + ε2
   elements in the row                Â(r,:)P(j) ≥ initial value – ε2
          PCA → NCA
With Prof. Martin:
• Relative acid concentrations in
  grape varieties.
• Can NCA be applied to get more
  information out of the data?
        Questions asked
• Where to publish?
  – Sys bio journals, bioinformatics, ieee
  – Probably multiple, some more bio
    focused, some more computationally
    focused.
• Have you thought about the model?
  – Two main pieces, the structure and
    the numbers.
                        Afterthoughts…
                continuing to work, towards next step,
                  all of this added post presentation.
osmolarity

                                                       waste

                                                                         Aquaporin?




                                                                 osmolytes
                                                               Endogenous production,
                                                                   Vs. transport.




               output osmotic       output protein
             pressure over time   products over time
      Calculation of osmotic
             pressure
• Osmotic pressure in atmospheres
   – π = MRT
      • M is the molar concentration of dissolved species (units
        of mol/L).
      • R is the ideal gas constant (0.08206 L atm mol-1 K-1, or
        other values depending on the pressure units).
      • T is the temperature on the Kelvin scale. molarity * R *
        temp(kelvin)
• Quantitating amounts of osmolytes prior to
  stress should give us an idea of:
   – Baseline pressure
   – Initial maximum pressure
• Quantitating during stress should give us a time
  course of osmotic pressure.
      Random questions
• Osmolarity in our cultures vs.
  industry relevent cultures vs. in
  vivo renal medullary conditions
  – Qi Cai et al 2004 cite different
    responses for linear increases in
    osmolarity vs. step incresases
       Market reports…
… are numerous.
• This is a very simplistic measure of
  the importance of this market
• Google: monoclonal antibody
  market
• See how much money you could
  spend just buying reports on the
  market.
    (ROUGH) Candidacy Practice
                  Experiments and blah blah blah
       Biological blablah blah blah Computational
           blah blah Elucidation of Signaling and
 Analysis Toward theblah blah blah Signaling blah
   Gene Expression blah blah Osmotic Shock and
Gene Expression Responses to blah Osmotic Shock
    Resultant Osmolyte Accumulation in
 blah blah blah blah blah blah blah blah Antibody
                               blah blah blah blah
                              Producing Cell Lines
                                       Spring 2007
                                        Tom Kiehl
                YSI Analyzer
Can provide quick measurements of the following
  analytes:
   –   D-Glucose (Dextrose)
   –   L-Lactate
   –   Sucrose
   –   Lactose
   –   Ethanol
   –   L-Glutamate
   –   Choline
   –   L-Glutamine
   –   Methanol
   –   Galactose*
   –   Hydrogen Peroxide*

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:33
posted:3/31/2008
language:English
pages:84