Hostname: page-component-7b7567c8c-wxvwp Total loading time: 0 Render date: 2026-11-16T07:58:54.229Z Has data issue: false hasContentIssue false

Modelling solute transport in the brain microcirculation: is it really well mixed inside the blood vessels?

Published online by Cambridge University Press:  17 December 2019

Maxime Berg
Affiliation:
Institut de Mécanique des Fluides de Toulouse (IMFT), Université de Toulouse, CNRS,Toulouse, France
Yohan Davit*
Affiliation:
Institut de Mécanique des Fluides de Toulouse (IMFT), Université de Toulouse, CNRS,Toulouse, France
Michel Quintard
Affiliation:
Institut de Mécanique des Fluides de Toulouse (IMFT), Université de Toulouse, CNRS,Toulouse, France
Sylvie Lorthois*
Affiliation:
Institut de Mécanique des Fluides de Toulouse (IMFT), Université de Toulouse, CNRS,Toulouse, France
*
†Email addresses for correspondence: yohan.davit@imft.fr, sylvie.lorthois@imft.fr
†Email addresses for correspondence: yohan.davit@imft.fr, sylvie.lorthois@imft.fr

Abstract

Most network models describing solute transport in the brain microcirculation use the well-mixed hypothesis and assume that radial gradients inside the blood vessels are negligible. Recent experimental data suggest that these gradients, which may result from heterogeneities in the velocity field or consumption in the tissue, may in fact be important. Here, we study the validity of the well-mixed hypothesis in network models of solute transport using theoretical and computational approaches. We focus on regimes of weak coupling where the transport problem inside the vasculature is independent of the concentration field in the tissue. In these regimes, the boundary condition between vessels and tissue can be modelled by a Robin boundary condition. For this boundary condition and for a single cylindrical capillary, we derive a one-dimensional cross-section average transport problem with dispersion and exchange coefficients capturing the effects of radial gradients. We then extend this model to a network of connected tubes and solve the problem in a complex anatomical network. By comparing with results based on the well-mixed hypothesis, we find that dispersive effects are a fundamental component of transport in transient situations with relatively rapid injections, i.e. frequencies above one Hertz. For slowly varying signals and steady states, radial gradients also significantly impact the spatial distribution of vessel/tissue exchange for molecules that easily cross the blood brain barrier. This suggests that radial gradients cannot be systematically neglected and that there is a crucial need to determine the impact of spatio-temporal heterogeneities on transport in the brain microcirculation.

Information

Type
JFM Papers
Copyright
© 2019 Cambridge University Press
Figure 0

Figure 1.Schematics of molecular transport processes at different scales in the brain with (a) scale of the whole brain; (b) scale of the cortex; (c) scale of a microvascular network; (d) scale of a single capillary vessel. This figure is a modified version of figure 1 in Lorthois et al. (2014a).

Figure 1

Figure 2.Flow rates and Péclet numbers within a mouse microvascular network, as computed in Cruz-Hernández et al. (2019). (a) Flow rates in a network that contains more than 15 000 vessels for a volume of approximately 1 cubic millimetre. It was obtained post mortem and imaged by two photon microscopy (Tsai et al.2009; Blinder et al.2013). The topology is described by an adjacency matrix, storing the list of all edges (e.g. $0$, $1$ and $2$) connected to a given vertex (e.g. B), as illustrated for a single bifurcation. (b) Distribution of the radial Péclet numbers, $\langle U\rangle R/D_{V}$, based on the blood flow distribution shown in (a) for a diffusion coefficient $D_{mol}=10^{-9}~\text{m}^{2}~\text{s}^{-1}$. Each dot represents a different vessel and the coloured surface shows the local density of points.

Figure 2

Figure 3.Schematic of solute transport within a single capillary. Here $C_{V}$ is the local concentration of solute within the vessel and $C_{T}$ the local concentration within the tissue. $U(r)$ represents the velocity profile; $D_{V}$ the diffusion coefficient of the solute within the vessel; $K_{m}$ is the membrane permeability; $D_{T}$ denotes the diffusion coefficient within the tissue; and $M(C_{T})$ the consumption rate of solute within the tissue.

Figure 3

Figure 4.Phase diagrams of transport regimes in (a) the $(Pe,Da_{m})$ plane for $Da_{T}=O(\unicode[STIX]{x1D716}^{-3})$ and in (b) the $(Da_{T},Da_{m})$ plane for $Pe=O(1)$. Dotted areas correspond to regimes of weak couplings where the membrane condition becomes a classic Robin condition. Green areas, which highlight the main result of this figure, correspond to parameter regimes for which it is possible to derive an effective transport equation for the cross-section-averaged concentration by multi-scale asymptotics. The associated transport regimes are labelled by roman numerals with: I, molecular diffusion only; II, molecular diffusion with effective reaction; III, advection and molecular diffusion; IV, advection, molecular diffusion with effective reaction; V, generalized Taylor’s dispersion; VI, advection, effective diffusion and effective reaction; VII, effective advection, effective diffusion and effective reaction.

Figure 4

Figure 5.Blood velocity profile as a function of (a) the parameter $n$ that controls the profile bluntness (for no slip velocity, $k=1$), and (b) the parameter $k$ that controls the slip velocity (for a parabolic profile, $n=2$).

Figure 5

Figure 6.Overspeed as a function of the radial membrane Damköhler number for different (a) velocity profile bluntness and (b) slip velocities.

Figure 6

Figure 7.Critical Péclet number as a function of radial membrane Damköhler number for different (a) velocity profile bluntness and (b) slip velocities.

Figure 7

Figure 8.Average stationary axial concentration for different Péclet and membrane Damköhler numbers for $\unicode[STIX]{x1D716}=0.05$. (a) $\unicode[STIX]{x1D716}Pe=0$ and $\unicode[STIX]{x1D716}Da_{m}=0.1$. (b) $\unicode[STIX]{x1D716}Pe=0$ and $\unicode[STIX]{x1D716}Da_{m}=10$. (c) $\unicode[STIX]{x1D716}Pe=100$ and $\unicode[STIX]{x1D716}Da_{m}=0.1$. $(d)$$\unicode[STIX]{x1D716}Pe=100$ and $\unicode[STIX]{x1D716}Da_{m}=10$. Green lines are analytical solutions of the WCA model. Red dashed lines are analytical solutions of the WM model. Blue dash-dotted lines are full 2-D finite volume solutions of the local transport problem. Insets show radial concentration profiles.

Figure 8

Figure 9.Extraction coefficient as a function of the radial membrane Damköhler number for $\unicode[STIX]{x1D716}=0.05$. (a) $\unicode[STIX]{x1D716}Pe=1$. (b) $\unicode[STIX]{x1D716}Pe=100$. Green lines are analytical solutions of the WCA model. Red dashed lines are analytical solutions of the WM model. Blue dash-dotted lines are full 2-D finite volume solutions of the local transport problem.

Figure 9

Figure 10.Transport of an initial square function at different times for $\unicode[STIX]{x1D716}=0.1$. (a) $\unicode[STIX]{x1D716}Pe=5$ and $\unicode[STIX]{x1D716}Da_{m}=0$. (b) $\unicode[STIX]{x1D716}Pe=5$ and $\unicode[STIX]{x1D716}Da_{m}=0.5$. Green lines are analytical solutions of the WCA model. Red dashed lines are analytical solutions of the WM model. Blue dash-dotted lines are full 2-D finite volume solutions of the local transport problem.

Figure 10

Figure 11.Numerical solutions of average concentration of an intravascular tracer ($Da_{m}=0$) with a high diffusion coefficient ($D_{V}=10^{-9}~\text{m}^{2}~\text{s}^{-1}$) resulting from a square input of 0.1 s duration at different times: (a) $T=0.1~\text{s}$; (b) $T=0.2~\text{s}$; (c) $T=0.3~\text{s}$; (d) $T=0.5~\text{s}$; (e) $T=0.6~\text{s}$; (f) $T=0.7~\text{s}$. This solution has been obtained using the WCA model on the network shown in figure 2(a) and using the velocity field computed in Cruz-Hernández et al. (2019). The case presented here corresponds to the plain green curves presented in figure 12(d). For full transient dynamics ($T=0$ s to $T=1$ s), see supplementary movie 1.

Figure 11

Figure 12.Integral output mass flux as a function of time: (a) 0.01 second input $D_{V}=10^{-10}~\text{m}^{2}~\text{s}^{-1}$; (b) 0.01 second input $D_{V}=10^{-9}~\text{m}^{2}~\text{s}^{-1}$; (c) 0.1 second input $D_{V}=10^{-10}~\text{m}^{2}~\text{s}^{-1}$; (d) 0.1 second input $D_{V}=10^{-9}~\text{m}^{2}~\text{s}^{-1}$. Green lines are analytical solutions of the WCA model. Red dashed lines are analytical solutions of the WM model. Thick black lines represent the input duration.

Figure 12

Figure 13.Integral output mass flux for a 5 s input square function $D_{V}=10^{-9}~\text{m}^{2}~\text{s}^{-1}$. Similar results have been obtained for $D_{V}=10^{-10}~\text{m}^{2}~\text{s}^{-1}$. The green line corresponds to the WCA model, the red dashed line to the WM model. The thick black line represents the input duration.

Figure 13

Figure 14.Stationary network extraction coefficient as a function of radial membrane Damköhler number. (a) WCA model and $D_{V}=10^{-10}~\text{m}^{2}~\text{s}^{-1}$. (b) WCA model and $D_{V}=10^{-9}~\text{m}^{2}~\text{s}^{-1}$. (c) WM model and $D_{V}=10^{-10}~\text{m}^{2}~\text{s}^{-1}$. (d) WM model and $D_{V}=10^{-9}~\text{m}^{2}~\text{s}^{-1}$. Tot., total extraction coefficient; Art., arteriolar extraction coefficient; Ven., venular extraction coefficient; Cap., capillary extraction coefficient.

Figure 14

Figure 15.Error estimators for the comparison between reference analytical and numerical solutions of (3.37)–(3.39) in the large anatomical anatomic network displayed in figure 2(a), as a function of the number of degrees of freedom (DoF) for $D_{V}=10^{-9}~\text{m}^{2}~\text{s}^{-1}$ and $\unicode[STIX]{x1D716}Da_{m}=0.1$. Error estimators: $e_{mean}=(1/N_{DoF})\sum |C_{VF}-C_{ref}|$, $e_{max}=\max _{DoF}|C_{VF}-C_{ref}|$, $e_{L2}=\Vert C_{VF}-C_{ref}\Vert$.

Berg et al. supplementary movie

Transient numerical solution (1s) of average concentration of an intravascular tracer with a high diffusion coefficient ($D_V=10^{-9}m^2.s^{-1}$) resulting from a square input of 0.1s duration. This solution has been obtained using the WCA model on the network shown in figure 2a and using the velocity field computed in Cruz-Hernández et al. (2019).

Download Berg et al. supplementary movie(Video)
Video 7.7 MB