By using cryo-EM, we next investigated whether
branching occurs at the tip or on the main body of
the A-microtubule and found that B-microtubules
assembled mainly on the body of the A-microtubule
(fig. S2A), corroborating previous in vivo find-
ings ( 10 ). Cryo-ET of these reconstituted MTDs
showed structural similarity to the ciliary MTDs
(Fig. 2, A and B; movie S1; and fig. S2B), with
23% of multiple MTDs surrounding a single
A-microtubule (Fig. 2C and fig. S2C). This in-
dicates that B-microtubule nucleation in vitro
was not restricted to one protofilament of the
A-microtubule and that, in vivo, additional pro-
teins may be needed to provide positional in-
formation. By using subtomogram averaging to
improve the resolution of the OJ to 17.2 Å (fig.
S2D), we confirmed the typical triangular junc-
tion formed among protofilaments A10 and A11
of the A-microtubule and protofilament B1 of
the B-microtubule in the reconstituted MTDs
(Fig. 2, D to G, and fig. S2E) ( 9 ). Additionally,
inspecting the curvature of the B-microtubule
junction from individual MTDs in cryo-ET (fig.
S3A) revealed a curvature similar to that ob-
served in in vivo MTDs (Fig. 2F). This suggested
that the surface-to-side tubulin-tubulin interac-
tion at the OJ is sufficient to drive the correct
angleforMTDassembly.Wenoticedanim-
portant mobility of the B-microtubule at the
MTD inner junction, possibly because of the
lack of MIPs or because of a protein such as
FAP20, which closes the MTDs in cilia (Figs.
1A and 2, I to J) ( 16 ).
By using an in silico approach, we explored
how the C-terminal tails of the A-microtubule
hinder MTD assembly. In our simulations, the
A-microtubule was composed of 13 protofila-
ments of threeab-tubulin dimers each, where
all atoms were taken into account. Because
tubulin tails are unstructured, they may adopt
random conformations. To obtain a represent-
ative sample of these conformations, we used
molecular dynamics simulations (see materials
and methods). The first protofilament from the
B-microtubule, B1, was added to the A-microtubule
between A10 and A11 according to the method
of ( 9 )(fig.S3,BtoD).Tocapitalizeonallthe
A-microtubule sampled tails, every two suc-
cessive protofilaments of the A-microtubule
(A1-A2, A2-A3, A3-A4, ... A12-A13) were super-
imposed on A10-A11 of the same microtubule in
order to obtain a variety of tail positions at the
OJ (Fig. 2K and fig. S3C). Then, for each of these
couples of protofilaments, the tails were relaxed
and their interaction energy with the entire
protofilament B1 was calculated. For the A10 tails,this energy was distributed around 0 kcal/mol,
indicating that these tails did not play a role in
the insertion of B1 (Fig. 2M). By contrast, for A11
tails, the interaction energy was highly repul-
sive in 11% of the cases, with an energy value of
several thousands of kilocalories per mole, which
is sufficient to strongly hinder the insertion of B1
(Fig. 2N). Visual inspection of the constructed
junctions confirmed that A11 tails did inter-
penetrate the core of B1, whereas A10 tails did
not (Fig. 2L). This provides an explanation for
the results of the in vitro experiments but not
for the formation of MTDs in vivo despite the
presence of these tails. Observation of the in
vivo MTD structure isolated from flagella ( 9 )
showed the presence of an unidentified MIP
(MIP7) at the junction between A11 and B1, at
thesamelocationasthetailsofA11inourmodel.
This MIP7 has been proposed to stabilize the
interaction between B1 and A10-A11 ( 9 ). We hy-
pothesized that MIP7 action is not in stabilizing
the interaction but rather in binding A11 proto-
filament tubulin tails to enable the B1 insertion.
Next, we monitored the assembly dynamics
of MTDs. We immobilized subtilisin-treated
Alexa 488–labeled A-microtubules on a glass
slide. Rhodamine-labeled free tubulin was added
to the reaction mixture to trigger B-microtubuleSchmidt-Cernohorskaet al.,Science 363 , 285–288 (2019) 18 January 2019 2of4
Fig. 2. Cryo-EM reconstruction of in vitro MTDs.(A) Representative
image of a cryo-ET section. Scale bar, 25 nm. (B)zxview of a cryo-ET
section. Scale bar, 25 nm. (C)zxview of a cryo-ET section showing an MTD
flower. Scale bar, 25 nm. Arrowheads in (A) to (C) indicate B-microtubules.
(DandF) Subtomogram averaging of in vitro MTDs at 17-Å resolution
(D) and ofTetrahymenaciliary MTDs at 5.7 Å (EMD-8528 map from the
Electron Microscopy Data Bank) (F). Scale bars, 25 nm. (EandG) Closer
view of the OJ for the in vitro MTD (E) and the OJ of the ciliary MTD (G).
Arrowheads indicate the triangular shape of the A10, A11, and B1 protofila-
ments. (H) Traces of the B-microtubules starting at the OJ and highlighting
the curvatures of the B-microtubules in vitro compared with those of
the in vivo ciliary MTDs (n=44microtubules).(IandJ)Plotprofilesatthe
positions indicated by the arrows in (H) showing that the curvature of the OJ
is stable (I) whereas the end of the B-microtubule is more flexible (J). The
black line indicates the position of the B-microtubule in vivo. A.U., arbitrary
units. (K) Side and top views of an MTD model with the A-microtubule in
green and gray (a-tubulin, gray;b-tubulin, green), the tubulin C-terminal tails
of the A-microtubule in red, and the protofilament B1 in blue and beige
(a-tubulin, blue;b-tubulin, beige). Atoms are represented as spheres.
(L) Closer view of the OJ highlighting the interactions of the tubulin
C-terminal tails of the protofilaments A10 and A11 with B1. The arrowhead
indicates the conflict between the C-terminal tails of A11 and B1.
(MandN) Plots of the interaction energy between tubulin C-terminal tails of
A10 and the protofilament B1 (M) or A11 and the protofilament B1 (N).RESEARCH | REPORT
on January 17, 2019^http://science.sciencemag.org/Downloaded from