1
1
Implantable photonic neural probes for
2
light
-
sheet fluorescence brain imaging
3
4
Wesley D. Sacher
1
,
2
,
3
,
4
,
∗
, Fu
-
Der Chen
3
,
†
, Homeira Moradi
-
Chameh
5
,
†
, Xianshu Luo
6
,
†
,
5
Anton Fomenko
5
, Prajay Shah
5
, Thomas Lordello
3
, Xinyu Liu
1
, Ilan
Felts Almog
3
,
6
John N. Straguzzi
4
, Trevor M. Fowler
1
, Youngho Jung
3
,
4
, Ting Hu
7
, Junho Jeong
3
,
7
Andres M. Lozano
5
,
8
, Patrick Guo
-
Qiang Lo
6
, Taufik A. Valiante
5
,
8
,
9
,
3
,
8
Laurent C. Moreaux
1
, Joyce K. S. Poon
3
,
4
, Michael L. Roukes
1,
2
∗
9
10
1
Division of
Physics, Mathematics, and Astronomy, California Institute of Technology,
11
Pasadena, California 91125, USA
12
2
Kavli Nanoscience Institute, California Institute of Technology, Pasadena, California
13
91125, USA
14
3
Department of Electrical and Computer Engineer
ing, University of Toronto, 10 King’s
15
College Rd., Toronto, Ontario M5S 3G4, Canada
16
4
Max Planck Institute of Microstructure Physics, Weinberg 2, 06120, Halle, Germany
17
5
Krembil Research Institute, Division of Clinical and Computational Neuroscience,
18
Univ
ersity Health Network, Toronto, Ontario, Canada
19
6
Advanced Micro Foundry Pte Ltd, 11 Science Park Road, Singapore Science Park II,
20
117685, Singapore
21
7
Institute of Microelectronics, Agency for Science Technology and Research (A*STAR),
22
2
Fusionopolis Way, #08
-
02, Innovis, 138634, Singapore
23
8
Division of Neurosurgery, Department of Surgery,
Toronto Western Hospital, University
24
of Toronto, Toronto, Ontario, Canada
25
9
Institute of Biomaterials and Biomedical Engineering, University of Toro
nto, Toronto,
26
Ontario, Canada
27
∗
Corresponding authors:
wesley.sacher@mpi
-
halle.mpg.de,
roukes@caltech.edu
28
†
Equal contribution
29
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint
this version posted September 30, 2020.
.
https://doi.org/10.1101/2020.09.30.317214
doi:
bioRxiv preprint
2
ABSTRACT
30
Significance:
Light
-
sheet
fluorescence
microscopy
is
a
powerful technique for high
-
speed
31
volumetric functional
imaging
. However, in typical light
-
sheet microscopes, the illumination and
32
collection optics impose significant constraints upon
the
imaging of n
on
-
transparent brain tissues.
33
Here, we demonstrate that these constraints
can be surmounted using
a new class of implantable
34
photonic neural probes
.
35
Aim:
Mass manufacturable, s
ilicon
-
based light
-
sheet photonic neural probes can generate planar
36
patterned il
lumination at arbitrary depths in brain tissues without any additional micro
-
optic
37
components.
38
Approach:
We
develop implantable photonic ne
ural probes that generate light
sheets in tissue.
39
The probes were fabricated
in a
photonics
foundry on 200 mm diamete
r silicon wafers.
The light
40
sheets were character
ized in fluorescein and in free
space. The probe
-
enabled imaging approach
41
was tested
in fixed and
in vitro
mouse brain tissues. Imaging tests were also performed
using
42
fluoresc
ent beads suspended in agarose
.
43
Results:
The probes had 5 to 10 addressable sheets and average sheet thicknesses
< 16
μm for
44
propagation distances up to 300 μm
in free
space
.
Imaging
areas
were
as large as ≈ 240 μm × 490
45
μm
in brain tissue. Image contrast was enhanced relative to epifluorescence microscopy.
46
Conclusions:
The neural probes can lead to new variants of light
-
sheet fluorescence microscopy
47
for deep brain imaging and
experiments in freely
-
moving animals.
48
49
Keywords:
Neurophotonics, integrated optics,
functional
i
maging, microscopy, biophotonics
50
51
I.
INTRODUCTION
52
New methods in optogenetics
[1
–
3
] and, especially, the advent of fluorescent reporters of
53
neuronal activity, have opened many novel approaches for actuating and recording neural activity
54
en masse
, through
the
use of powerful
free
-
space
single
-
photon and multi
-
photon
microscopy
55
metho
ds [4
–
8
].
However,
e
xisting
approaches to functional imaging
of the brain
have significant
56
limitations.
S
ingle
-
photon (1P) epifluorescence imaging readily lends itself to high frame
-
rate
57
wide
-
field microscopy, but, in its simplest implementations, image contras
t is hampered by out
-
58
of
-
focus background fluorescence
,
and the depth of imaging is restricted by the optical attenuation
59
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint
this version posted September 30, 2020.
.
https://doi.org/10.1101/2020.09.30.317214
doi:
bioRxiv preprint
3
in the tissue. Confocal imaging improves the contrast by optical sectioning, and out
-
of
-
focus light
60
is rejected using a pinhole; howeve
r, a laser beam must be scanned across
each point of
the tissue
61
and this
significantly
slow
s the image acquisition rate [9
]. Multiphoton microscopy is also
62
inherently a point or line scanning method,
but because it uses
infrared excitation
(which provides
63
a longer optical attenuation length [
5
])
, the imaging depth
in brain tissue
can be extended to
∼
1
64
mm and the
focus of the light beam can be rastered in three
-
dimensions t
o achieve volumetric
65
imaging [5, 10
–
12
].
66
Light
-
sheet fluorescence mi
croscopy (LSFM)
, which is also known as selective
-
plane
67
illumination microscopy
,
combines the benefits of fast wide
-
field imaging, volumetric ima
ging,
68
and optical sectioning [13
]. In
conventional
LSFM, a thin sheet of excitation light is generated
69
either b
y cylindr
ical
f
ocusing elements
or
digitally scanni
ng a Gaussian or Bessel beam [14
–
16
].
70
The sheet is translated in one dimension across the sample; the fluorescence images are
then
71
sequentially collected in the direction perpendicular to the illumination
plane to
form a volumetric
72
image [17
].
With d
igitally
scanned two
-
photon (2P) LSFM
, it
i
s also possible to increase the
73
optical penetration depth
[16
]. Non
-
digitally
scanned 1P
-
LSFM is inherently faster than point
-
or
74
line
-
scan methods
;
and since the illumination is restricted to a plane
,
photobleaching,
75
phototoxicity, and out
-
of
-
focus background fluorescence are reduced compared
to
epifluorescence
76
microscopy. However, conventional LSFM requires two orthogonal objective lenses
, and
77
appro
priately
positioning these
largely
limit
s
the imaging
modality to quasi
-
transparent organisms
78
(
e.g.
,
C. elegans
,
Drosophila
embryos
,
larval zebrafish
), chemically cleared m
ammalian brains
79
[
17
]
,
and
brain slices [18
]. A
n
LSFM variant called swept confocally
-
aligned planar excitation
80
(SCAPE) microscopy
, which requires only a single objective,
remove
s
these constraints [6
,
19
].
81
While
in vivo
calcium neural imaging has been demonstrated
using SCAPE
in mice [6
],
82
miniaturization of the system to be compatible wit
h freely
moving animal experiments remains
83
challenging due to the additional optics required.
84
To
make
LSFM compatible with non
-
transparent tissues such as mammalian brains and
,
85
eventually
,
behavioral experiments with
freely
moving animal
s
necessitates
drastic miniaturization
86
of the light
-
sheet generation and fluorescence imaging compared to
today’s
archetypical table
-
top
87
systems. The feasibility of fluorescence microscopy in small and lightweight form factors has
88
already
been established b
y way of head
-
mounted microscopes for 1P and 2P c
alcium imaging in
89
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint
this version posted September 30, 2020.
.
https://doi.org/10.1101/2020.09.30.317214
doi:
bioRxiv preprint
4
mice [4, 20
–
23
], though
the
endoscopic implantation of the
requisite gradient index (GRIN) lenses
,
90
with typical diameters of
0.5
-
2 mm
,
displaces a significant amount of brain tissue.
91
On t
he other hand, i
t
remains a formidable and unsolved challenge
to g
enerat
e
light sheets
by
92
implantable elements at arbitrary brain depths, while minimizing tissue displacement and
93
remaining
compatible with a sheet
-
normal ima
ging system. For example, in [24
]
, to generate a
94
light sheet perpendicular to the imaging GRIN lens
required
implantation
of
a
millimeter
-
scale
95
prism coupled to a
second GRIN
lens. In another example, in [25
], a single light sheet was
96
produced from a microchip using a grating coupler, a g
lass spacer block, and a metallic slit lens.
97
The
overall device was
>
100
μ
m thick
and
>
6
00
μ
m wide
, which would displace
a significant
98
amount of
tissue
upon
implant
ation
.
99
Here,
we solve th
ese
challenge
s
by using
wafer
-
scale nanophotonic technology to realize
100
implantable
,
silicon
-
based,
light
-
sheet photonic neural
probes that require
no add
i
tional
micro
-
101
optics
.
T
hey are
fully compatible with
free
space
fluorescence imaging
(light collection)
outside
102
the brain,
whe
re
the axis of collection is oriented
perpendicular to the light sheet
s
. The
se silicon
103
(Si)
probes synthesize light sheets in tissue using sets of nanophotonic grating couplers (GCs)
104
integrated onto thin, implantable
,
3
mm long
Si shanks
with
50
-
92
μ
m thick
ness,
widths
that
105
taper from 82
-
60
μ
m
along their length,
and
sharp tip
s at the
distal ends
. The
se prototype
photonic
106
neural probes
(
Fig. 1
)
are capable of
generating and sequentially
addressing up to 5
illumination
107
planes with a pitch of
≈
70
μ
m.
Additionally, the
form factor
and illumination geometry of the
108
probes open an avenue toward
their
integration with GRIN lens endoscopes and miniature
109
microscopes, as shown
conceptually
in Fig. 2(b); offering
a singular pathway to rapid, optically
110
sectioned functional imaging at arbitrary depths in the brain.
111
The probes were fabricated on 200 mm Si wafers in a Si photonic
s
foundry
for
manufacturing
112
scalability
and mass
-
producibility
.
Elsewhere
,
we have use
d
this technology to realize p
hotonic
113
neural probes
that
emit dynamically
-
reconfigurable, patterned light with
cellular
-
scale beam
114
widths
[26
]
and
steer
able beams
without moving parts
[27
]
, adding to a growing
number
of
115
photonic neural probe demonstrations
with increasing levels of in
t
egration and sophistication
[28
–
116
30
]
. In
this work,
we employ this integrated nanophotonics technology to
realize implantable,
117
microscale probes that
form
light sheets for imaging over areas as large as
≈
240
μ
m
×
490
μ
m
in
118
brain tissue. Our preliminary results
were
reported in [31
]. Here, we report
in detail
the imaging
119
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint
this version posted September 30, 2020.
.
https://doi.org/10.1101/2020.09.30.317214
doi:
bioRxiv preprint
5
properties of the light
-
sheet neural probes
–
characterizing their performance by means of
120
suspended fluorescent beads in phantom
s as well as
in adult mou
se brain slices. A
first
121
demonstration of
in vivo
calcium imaging is
also reported
in the
Supplementary Material
s
.
122
123
II.
RESULTS
124
A.
Photonic neural probes on 200 mm silicon wafers
125
To ensure that fabrication of our
photonic neural probes
can be scaled up for dissemination
to
126
the neuroscience community, we
have
adapted
from the outset
foundry Si photonic
s
manufacturing
127
processes
. The neural probes
described herein
were fabricated in a 200 mm Si photonic line
;
silicon
128
nitride (SiN) waveguide
s
(135 nm nominal thickness)
with
SiO
2
cladding were
patterned
onto
Si
129
wafers
, deep trenches were etched in the wafers to define the probe shapes, and the wafers
were
130
thinned to
thicknesses of 50
-
92
μ
m.
The shank thickness can
be reduc
ed in future iterat
ions to
18
131
μ
m
,
as
in
[26, 31
].
The f
abrication is
more fully
detailed in
Methods
.
132
The light
-
sheet neural probe design is shown in Figs. 1(a)
-
(c). Light is coupled onto the
probe
133
chip using fiber
-
to
-
chip edge couplers
that taper from
5.2
μ
m in width at the chip facet to single
-
134
mode waveguides
with
widths of
270
–
330
nm
. The
waveguide
-
co
upled optical
power is divided
135
between
four
to
eight
waveguides using a routing network consisting of
1
×
2
multimode
136
interference (MMI) splitters [32
] and in
-
plan
e waveguide crossings [33
]. The light is then guided
137
along the implantable shanks via 1
μ
m wi
de, multimode waveguides, and subsequently emitted
138
near
the
distal end
of
the
probe
by a row of GCs. Light
sheets are synthesized by overlapping the
139
emission from an array of simultaneously
-
fed GCs. Each row of
GCs generates a separate light
140
sheet. The wid
th, period, and duty cycle of the GCs are designed to achieve a large output
141
divergence angle along the width
-
axis of the sheet
,
and
only
a
small divergence along the
142
thickness
-
axis. Nominal lateral GC widths, periods, and duty cycles
are
1.5
μ
m, 440 to 480 nm,
143
and 50%, respectively.
144
The waveguide routing network is detail
ed
in Fig.
S
1
in the Supplementary Material
s
. The
145
photonic components were designed for a wavelength of 488 nm
to enable
excitation of
common
146
fluorophores
such as
green fluore
scent protein (GFP) and green calcium dyes; however, these
147
components can also be designed for green, yellow, and red
wavelengths, as we show in [34
] for
148
excitation of other fluorophores.
The probe
shanks
are
3 mm
in
l
ength
and
separated with
a 141
149
(which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint
this version posted September 30, 2020.
.
https://doi.org/10.1101/2020.09.30.317214
doi:
bioRxiv preprint