critique of a nanoparticle cancer therapy model
DESCRIPTION
A scientific critique of a computational fluid dynamics and mass transport model describing nanoparticle flux into a tumor. Methods and results are briefly described, followed by a point by point critique of the author's methods, conclusions, and the paper's significance.TRANSCRIPT
Paper CritiqueCationic Nanoparticles Have Superior Transvascular Flux into Solid
Tumors: Insights from a Mathematical Model
T. Stylianopoulos, K. Soteriou, D. Fukumura, R. Jain
Presented by: Nils Persson
2
Challenge
• Deliver nanoparticles into the tumor interstitium uniformly and selectively
• Resistances to this goal:– Flow through capillaries– Flow through capillary pores– Diffusion through interstitium– Drainage to normal tissue
3
Structure of the Model
Capillary – Hagen Poiseuille flowConvection only
PoresHindered
stokesflow
TumorPorous flow and diffusion
No drainageNormal tissueP = 0
DiffusionConvection
Grid 1
Grid 2
4
Structure of the Model
Capillary – Hagen Poiseuille flowConvection only
TumorPorous flow and diffusion
No drainageNormal tissueP = 0
DiffusionConvection
Grid 1
Grid 2
PoresHindered
stokesflow
5
Structure of the Model
Capillary – Hagen Poiseuille flowConvection only
TumorPorous flow and diffusion
No drainageNormal tissueP = 0
DiffusionConvection
Grid 1
Grid 2
Effect of charge,Ionic strength,Pore size
PoresHindered
stokesflow
6
Two steps: Pressure, then Concentration
• Steady state pressure profile
7
Capillary – Hagen Poiseuille flowConvection only
TumorPorous flow and diffusion
No drainageNormal tissueP = 0
DiffusionConvection
Grid 1
Grid 2
Two steps: Pressure, then Concentration
PoresHindered
stokesflow
8
Two steps: Pressure, then Concentration
• RK4 time stepping with centered finite differences to solve C(r,t)– Not reported, most interesting result
• Figure of merit: Vascular permeability or transvascular flux, Peff
– Similar to a mass transfer coefficient into tumor:
Averaged concentration, C
9
Representative Results
Fluid in tumor stagnatesPore diameter is too restrictive
Peff
10
Critiques• Blood viscosity/compressibility• Capillary Compliance• Dilute particle assumption• Concentration BCs unspecified and unrealistic• Transvascular flux is over-fit• Deen’s theory is shaky for attractive charges• Centerline approximation invalid for attractive charges• “lacking information” on glycocalyx charge, the crux of the entire
paper• Neglected axial diffusion even in areas of zero flow• 2-dimensional percolation network was unnecessary to obtain results• Lead author did not apply results of a previous modeling study
11
Critiques• Blood viscosity/compressibility• Capillary Compliance• Dilute particle assumption• Concentration BCs unspecified and unrealistic• Transvascular flux is over-fit• Deen’s theory is shaky for attractive charges• Centerline approximation invalid for attractive charges• “lacking information” on glycocalyx charge, the crux of the entire
paper• Neglected axial diffusion even in areas of zero flow• 2-dimensional percolation network was unnecessary to obtain results• Lead author did not apply results of a previous modeling study
12
Blood Flow
• Modeled with Hagen-Poiseuille equation– Assumes laminar flow, smooth pipe, constant density and viscosity
(3 mmHg-s = 4 mPa-s)
• Properties of blood:– Viscoelastic: erythrocyte deformability and aggregation– Shear thinning– Viscosity varies with hematocrit level
• Plasma:– Newtonian– µ = 1.2 mPa-s (water = 0.65 mPa-s @ 37 °C)
13
Dilute Particle Assumption
• Deen’s theory assumes no solute-solute interactions (continuous solvent)
• Nanoparticles are in pores with plasma, concentration of albumin (~10 nm) is about 0.5 wt%
• Deen predicts ~10% higher partitioning into pores at this concentration
14
Where does Deen’s theory break down?
• “Only for attractive interactions (E<0) is there likely to be a problem.” – Deen, 1987
• Hindrance integrals blow up
15
The Centerline Approximation
• Says that hindrance coefficients are those of r=0– Very good for repulsive charges– Very bad for attractive charges
- - - - - -
- - - - - -
Deen, 1987
16
The glycocalyx – source of charge• Layer of carbohydrates on
inner wall of capillary– But, in pores?
• Negatively charged, ~-0.022 C/m21
-used -0.05 C/m2
• 0.5 µm thick – too thick for capillary pores
• Acts as molecular sieve and shear stress transducer
1Donath et al., 1996, image Reitsma et al., 2007
17
Previous Studies
• Campbell and Jain et al., 2002– Tumor uptake not affected by charge, however,
accumulation in tumor vessels doubled for cationic particles
• Stylianopoulos, et al., 2010– Positive charge hinders
diffusion in the interstitium– Not applied to this model
18
Conclusions
• Very powerful model used to study known charge phenomena– Many bad approximations, missing data
• Much greater insight is provided on the transport phenomena and pore size distributions
• Slight modifications could yield very realistic results
19
Dimensionless Quantities, 400nm pores, 60nm particles
DiffusionConvection
Grid 1
Grid 2
D ~ 10-13 m2/s= 10-9 cm2/s
V:4E-5 m/sPe,r = 9Pe,L = 220Re = 4E-6∆P = 1.25 mmHg
V:2.5E-4 m/sPe,r = 53Pe,L = 1300Re = 2.5E-5∆P = 7.5 mmHg
5µm
15µm
V:2.3E-4 m/sPe,L = 2.3E6Re = 8.6E-4∆P = 20 mmHg (max)
V:2E-6 cm/sPe,L = 100Re << 1∆P = 1.25 mmHg“L” = 0.5cm
Assumes Ac is void surface area
V:2.4E-5 cm/sPe,L = 600Re << 1∆P = 7.5 mmHg“L” = 0.25cm
20
Reynolds Number in Pores… <<1∆P = 0.05 = 1.25 mmHg ∆P = 0.3 = 7.5 mmHg
166.6 Pa 999.7 PaR (nm) R (m) v (m/s) Re v (m/s) Re
10 1.00E-08 1.0E-07 5.2E-10 6.2E-07 3.1E-0950 5.00E-08 2.6E-06 6.5E-08 1.6E-05 3.9E-07
100 1.00E-07 1.0E-05 5.2E-07 6.2E-05 3.1E-06150 1.50E-07 2.3E-05 1.8E-06 1.4E-04 1.1E-05200 2.00E-07 4.2E-05 4.2E-06 2.5E-04 2.5E-05250 2.50E-07 6.5E-05 8.1E-06 3.9E-04 4.9E-05300 3.00E-07 9.4E-05 1.4E-05 5.6E-04 8.4E-05350 3.50E-07 1.3E-04 2.2E-05 7.7E-04 1.3E-04400 4.00E-07 1.7E-04 3.3E-05 1.0E-03 2.0E-04450 4.50E-07 2.1E-04 4.7E-05 1.3E-03 2.8E-04500 5.00E-07 2.6E-04 6.5E-05 1.6E-03 3.9E-04
21
Peclet Numbers in Pores… >>1
∆P = 1.25 mmHg ∆P = 7.5 mmHg166.61 Pa 999.67 Pa
R (nm) R (m) v (m/s) D(m^2/s) Pe, r Pe,L v (m/s) D(m^2/s) Pe, r Pe,L10 1.00E-08 1.0E-07 1.89E-11 5.5E-05 2.8E-02 6.2E-07 1.89E-11 3.3E-04 1.7E-0150 5.00E-08 2.6E-06 3.78E-12 3.4E-02 3.4E+00 1.6E-05 3.78E-12 2.1E-01 2.1E+01
100 1.00E-07 1.0E-05 1.89E-12 5.5E-01 2.8E+01 6.2E-05 1.89E-12 3.3E+00 1.7E+02150 1.50E-07 2.3E-05 1.26E-12 2.8E+00 9.3E+01 1.4E-04 1.26E-12 1.7E+01 5.6E+02200 2.00E-07 4.2E-05 9.46E-13 8.8E+00 2.2E+02 2.5E-04 9.46E-13 5.3E+01 1.3E+03250 2.50E-07 6.5E-05 7.57E-13 2.2E+01 4.3E+02 3.9E-04 7.57E-13 1.3E+02 2.6E+03300 3.00E-07 9.4E-05 6.30E-13 4.5E+01 7.4E+02 5.6E-04 6.30E-13 2.7E+02 4.5E+03350 3.50E-07 1.3E-04 5.40E-13 8.3E+01 1.2E+03 7.7E-04 5.40E-13 5.0E+02 7.1E+03400 4.00E-07 1.7E-04 4.73E-13 1.4E+02 1.8E+03 1.0E-03 4.73E-13 8.5E+02 1.1E+04450 4.50E-07 2.1E-04 4.20E-13 2.3E+02 2.5E+03 1.3E-03 4.20E-13 1.4E+03 1.5E+04500 5.00E-07 2.6E-04 3.78E-13 3.4E+02 3.4E+03 1.6E-03 3.78E-13 2.1E+03 2.1E+04
Assumes γ=0.3 (similar to paper)Assumes Stokes-Einstein diffusivityAssumes Hagen-Poiseuille velocity in pores
22
Data
23
Ionic Strength
• I (mol/kg) = ½ Σmizi2
– Seawater: 0.72 mol/kg– In paper: varied from 0 to 0.15M
• Debye length κ-1 ~ I-1/2
– 1 nm• ln γ ~ κ ~ I1/2
Increased concentration = decreased Debye length = lower potential
Debye LengthsI K-1 (m) e0 8.85E-12
0.005 4.38E-09 eR 780.01 3.10E-09 R 8.3140.06 1.26E-09 T 3100.15 7.99E-10 dS 1000
Na 6.02E+23e 1.60E-19
24
H and W for negative particles
Lambda = 0.3I = 0.15
3
25
Flux of negatively charged particles
Q = -.05
Lambda = 0.3
Both 400nm/60nm
4
26
More negative charge fluxes
positive
NegativeI = .15
5
400nm/60nm
27
Varying pore size
28
Fitting the Vascular Permeability
• C(t) linear – why?
• They fit an exponential0
100200
300400
500600
700800
9001000
0.00
0.05
0.10
0.15
0.20
0.25
f(x) = 0.0002199202 x + 0.00040044141R² = 0.999948648166217
Time (sec)
Dim
ensi
onle
ss C
on-
cent
ratio
n
Source: Data Thief
Large Kd yields vastly simpler eqn.Kd ~ 80,000
29
No flowpredicted
30
No flow to the tumor
Wu et al. 2008
- Also predicted by a paper they cite
- Nanoparticles reach tumor’s edge, then spill out
- This paper also treats viscosity as a function of hematocrit
31
Transport Properties of a Tumor
Chauhan et al. 2011
32
Starling’s Law
• Jv = Kf([Pc-Pi]-σ[πc-πi])• σ = reflection coef.– More likely varies with pore size– Human serum albumin: ~10nm in diameter
• Assumes Hagen-Poiseuille flow in pore• R = 200 nm, µ = 4 Pa-s, ∆P varies, L = 5 µm• ∆P is normalized to 25 mmHg
33
Vascular Equations
Hagen - Poiseuille
Mass balance, missing transvascular loss
- ϕ
34
Interstitial Equations
+ ϕTumor mass balane:
Pe ~ 800 in the interstitium
Transvascular“generation” term
Darcy’s Law, Ac = void surface area Ac = pi*d/Sv
35
Transvascular Equations
S = inner surface area of vessel
L = thickness of capillary wall
For steady state
R (nm) Lp (m^3/s/Pa/m^2) and in cm/mmHg v (m/s)10 1.88E-10 2.50E-06 6.25E-0750 4.69E-09 6.25E-05 1.56E-05
100 1.88E-08 2.50E-04 6.25E-05150 4.22E-08 5.62E-04 1.41E-04200 7.50E-08 1.00E-03 2.50E-04250 1.17E-07 1.56E-03 3.91E-04300 1.69E-07 2.25E-03 5.63E-04350 2.30E-07 3.06E-03 7.66E-04400 3.00E-07 4.00E-03 1.00E-03450 3.80E-07 5.06E-03 1.27E-03500 4.69E-07 6.25E-03 1.56E-03
36
Transvascular EquationsQt
FluxMass/time*area
Flowmass/time
W = convective hindranceH = diffusive hindrance
1-σ=WDerived from…
37
Hindrance Factors
diffusive
convective pore
Centerline:Pull out K-1 and GAs constants
τ=κ
38
Derivation of Partition Coefficients
Drag forceDiffusive force
• N = U*C• Poiseuille velocity profile• Average flux over pore area• Assume C(z,β) separable• Neglect Taylor dispersion• Radial Pe > 2• Axial Pe >29• Both are satisfied (See table)
39
Runge-Kutta 4th Order
Industry standard, O(h5) error
40
Transport through pores
• Assumptions:– low Re <1, Per>>1 (samples all radii)– dilute in fluid continuum (?)– significant radius relative to pore– no Taylor dispersion (Per>2, PeL>29)– fully developed flow
• Diffusion and hydrodynamic drag forces
Smith & Deen, 1982
41
Length Scales
• RBC: D = 8 µm• Human serum albumin: 10 nm• Nanomedicines: 10 – 100 nm– In paper: 6 – 200 nm, usually 0.3*Dp
• Capillaries: 10 µm (15 µm in paper)• Vascular pores: 1-1000 nm– In paper: 400 nm average, distribution
42
Vascular Network
• 200x200 two-layer grid, each point is defined by the neighbors to which it connects
• Volumetric flow conserved through each point• Vasculature only exists at pre-defined points• Interstitium occupies all grid points inside radius
43
Overview of Tumor Modeling
Wu et al. 2007