secdocument_526Download Mammalian Near Infrared Image Vision Through Injectable And Self Powered Retinal Nanoantennae Yuqian Ma Jin Bao Yuanwei Zhang Zhanjun Li Xiangyu Zhou Changlin Wan Ling Huang Yang Zhao full chapter
secdocument_526Download Mammalian Near Infrared Image Vision Through Injectable And Self Powered Retinal Nanoantennae Yuqian Ma Jin Bao Yuanwei Zhang Zhanjun Li Xiangyu Zhou Changlin Wan Ling Huang Yang Zhao full chapter
secdocument_526Download Mammalian Near Infrared Image Vision Through Injectable And Self Powered Retinal Nanoantennae Yuqian Ma Jin Bao Yuanwei Zhang Zhanjun Li Xiangyu Zhou Changlin Wan Ling Huang Yang Zhao full chapter
Correspondence
baojin@ustc.edu.cn (J.B.),
gang.han@umassmed.edu (G.H.),
xuetian@ustc.edu.cn (T.X.)
In Brief
Injectable photoreceptor-binding
nanoparticles with the ability to convert
photons from low-energy to high-energy
forms allow mice to develop infrared
vision without compromising their normal
vision and associated behavioral
responses.
Highlights
d We designed ocular injectable photoreceptor-binding
upconversion nanoparticles
Article
Brain Function and Disease, School of Life Sciences, Division of Life Sciences and Medicine, University of Science and Technology of China,
Hefei, Anhui 230026, China
2Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, Shanghai 200031, China
3Department of Biochemistry and Molecular Pharmacology, University of Massachusetts Medical School, Worcester, MA 01605, USA
4Institute for Stem Cell and Regeneration, Chinese Academy of Sciences, Beijing 100101, China
5These authors contributed equally
6Lead Contact
SUMMARY lower energy photons, requires opsins (e.g., human red cone op-
sins) to have much lower energy barriers. Consequently, this re-
Mammals cannot see light over 700 nm in wave- sults in unendurable high thermal noise, thus making NIR visual
length. This limitation is due to the physical thermo- pigments impractical (Ala-Laurila et al., 2003; Baylor et al., 1980;
dynamic properties of the photon-detecting opsins. Luo et al., 2011; St George, 1952). This physical limitation means
However, the detection of naturally invisible near- that no mammalian photoreceptor can effectively detect NIR
infrared (NIR) light is a desirable ability. To break light that exceeds 700 nm, and mammals are unable to see
NIR light and to project a NIR image to the brain.
this limitation, we developed ocular injectable
To this end, the successful integration of nanoparticles with
photoreceptor-binding upconversion nanoparticles biological systems has accelerated basic scientific discoveries
(pbUCNPs). These nanoparticles anchored on retinal and their translation into biomedical applications (Desai, 2012;
photoreceptors as miniature NIR light transducers to Mitragotri et al., 2015). To develop abilities that do not exist natu-
create NIR light image vision with negligible side ef- rally, miniature nanoscale devices and sensors designed to inti-
fects. Based on single-photoreceptor recordings, mately interface with mammals including humans are of growing
electroretinograms, cortical recordings, and visual interest. Here, we report on an ocular injectable, self-powered,
behavioral tests, we demonstrated that mice with built-in NIR light nanoantenna that can extend the mammalian vi-
these nanoantennae could not only perceive NIR sual spectrum to the NIR range. These retinal photoreceptor-
light, but also see NIR light patterns. Excitingly, the binding upconversion nanoparticles (pbUCNPs) act as miniature
injected mice were also able to differentiate sophisti- energy transducers that can transform mammalian invisible NIR
light in vivo into short wavelength visible emissions (Liu et al.,
cated NIR shape patterns. Moreover, the NIR light
2017; Wu et al., 2009). As sub-retinal injections are a commonly
pattern vision was ambient-daylight compatible and
used ophthalmological practice in animals and humans (Haus-
existed in parallel with native daylight vision. This wirth et al., 2008; Peng et al., 2017), our pbUCNPs were dis-
new method will provide unmatched opportunities solved in PBS and then injected into the sub-retinal space in
for a wide variety of emerging bio-integrated nanode- the eyes of mice. These nanoparticles were then anchored and
vice designs and applications. bound to the photoreceptors in the mouse retina.
Through in vivo electroretinograms (ERGs) and visually evoked
INTRODUCTION potential (VEP) recordings in the visual cortex, we showed that
the retina and visual cortex of the pbUCNP-injected mice were
Vision is an essential sensory modality for humans. Our visual both activated by NIR light. From animal behavioral tests, we
system detects light between 400 and 700 nm (Dubois, 2009; further demonstrated that the pbUCNP-injected mice acquired
Wyszecki and Stiles, 1982; Schnapf et al., 1988), so called visible NIR light sensation and unique ambient daylight-compatible
light. In mammalian photoreceptor cells, light absorbing pig- NIR light image vision. As a result, the built-in NIR nanoantennae
ments, consisting of opsins and their covalently linked retinals, allowed the mammalian visual spectrum to extend into the NIR
are known as intrinsic photon detectors. However, the detection realm effectively without obvious side effects. Excitingly, we
of longer wavelength light, such as near-infrared (NIR) light, found that pbUCNP-injected animals perceived both NIR and
though a desirable ability, is a formidable challenge for mam- visible light patterns simultaneously. They also differentiated be-
mals. This is because detecting longer wavelength light, with tween sophisticated NIR light shape patterns (such as triangles
and circles). Importantly, this nanoscale device activated the spectrum (Figure S1B). To confirm the glyosidic bonds between
photoreceptors by an exceptionally low power NIR light-emitting ConA and glycoproteins, we added b-cyclodextrin, which pos-
diode (LED) light (1.62 mW/cm2), which was attributed to the sesses a similar glucosyl unit as that found on the photoreceptor
proximity between the nanoantennae and photoreceptors in outer segment, to the pbUCNP solution. Characteristic ConA-
the eye. Moreover, we comprehensively examined the biocom- b-cyclodextrin aggregation thus occurred, as seen in the trans-
patibility of the pbUCNPs and found negligible side effects. mission electron microscope (TEM) images (Figure 1G) and
Therefore, these novel photoreceptor-binding NIR light nano- dynamic light scattering (DLS) spectrum (Figure S1C). This result
antennae provide an injectable, self-powered, biocompatible, suggests that the pbUCNPs can bind to glycoproteins on
and NIR-visible light compatible solution to extend the mamma- the photoreceptor outer segment. In contrast, the paaUCNPs
lian visual spectrum into the NIR range. This concept-proving without ConA remained monodispersed when b-cyclodextrin
research should guide future studies with respect to extending was added (Figures 1H and S1D). After injecting these pbUCNPs
human and non-human vision without the need for any external into the mouse sub-retinal spaces (Figures 1F and S1E), we
device or genetic manipulation. Endowing mammals with NIR observed that, through the glyosidic bond, these pbUCNPs
vision capacity could also pave the way for critical civilian and self-anchored and remained tightly bound to the inner and outer
military applications. segments of both rods and cones (Figures 1J–1L) forming a layer
of built-in nanoantennae with the characteristic upconversion
RESULTS spectrum (Figures 1I, left, S1F, and S1G). In contrast, the in-
jected paaUCNPs were loosely bound and easily removed
The Design of pbUCNPs from the photoreceptors with gentle washing (Figure 1I, right).
The human eye is most sensitive to visible light at an electromag- We then evaluated the biocompatibility and potential side ef-
netic wavelength of 550 nm under photopic conditions (Bieber fects of the pbUCNPs in vivo. We found that the pbUCNP injec-
et al., 1995; Boynton, 1996). To convert NIR light to this tion did not cause a higher rate of adverse reactions compared
wavelength, we generated core-shell-structured upconversion with the control PBS injection. All common minor or transient
nanoparticles (UCNPs) (i.e., 38 ± 2 nm b-NaYF4:20%Yb, 2% side effects (e.g., cataracts, corneal opacity) generally associ-
Er@b-NaYF4) (Figures 1A and 1B), which exhibited an excitation ated with sub-retinal injection (Qi et al., 2015; Zhao et al., 2011)
spectrum peak at 980 nm and emission peak at 535 nm upon disappeared completely 2 weeks after the injections (Table
980-nm light irradiation (Mai et al., 2006; Wu et al., 2015) (Figures S1). In addition, we evaluated possible retinal degeneration by
1C and 1D). To design water-soluble pbUCNPs, we further con- counting the number of photoreceptors in the retinal outer nu-
jugated concanavalin A protein (ConA) with poly acrylic acid- clear layer (ONL), a standard and widely used method in the field
coated UCNPs (paaUCNPs) (Figure 1E; STAR Methods). ConA of retinal research, as photoreceptors are sensitive and prone to
can bind to sugar residue and derivatives of the photoreceptor degenerate upon stress (Chen et al., 2006; Namekata et al.,
outer segment, forming glyosidic bonds (Bridges, 1981; Bridges 2013; Wang et al., 2013). As a result, we observed that the retinal
and Fong, 1980; Rutishauser and Sachs, 1975). Successful layer structure and the number of photoreceptor layers in the
ConA conjugation on the surface of the UCNPs was suggested retinal ONL were not changed, even with 50 mg of pbUCNPs in-
by the appearance of N-H bending peaks in the Fourier transform jected into each eye, up to 2 months after the injections (Figures
infrared (FT-IR) spectrum (Figure S1A) and by the 285 nm pro- 2A and 2B). This result clearly suggested that there was no
tein absorption on the ultraviolet-visible spectroscopy (UV/Vis) obvious retinal degeneration using this standard measure.
Moreover, we examined potential inflammation in the retinal injection did not cause obvious acute or long-term side effects.
through microglia marker Iba1 staining that is a widely used indi- In addition, the excitation and emission spectra of the pbUCNPs
cator of microglia accumulation (Krady et al., 2005). From this, in either fixed or fresh retinae were in good agreement with those
we observed negligible retinal inflammation at 3 days or 1, 2, 4, measured from pbUCNP solution, indicating that binding with
and 10 weeks after pbUCNP injection (Figures 2C, 2E, and the photoreceptors did not change the characteristics of the
S2A). We further examined retinal cell apoptosis after injection pbUCNPs (Figures S2C and S2D).
via terminal deoxynucleotidyl transferase deoxyuridine triphos-
phate (dUTP) nick-end labeling (TUNEL). We only found sparse NIR Light-Mediated Photoreceptor Activation
TUNEL signals 3 days after injection in both the PBS and Based on the biocompatibility noted above, we tested if the pho-
pbUCNP-injected retinae (Figure 2D), with the TUNEL signals toreceptors could be activated by NIR light with the help
being undetectable 1, 2, 4, and 10 weeks after pbUCNP injection of pbUCNPs. We performed single rod suction pipette record-
(Figures 2E and S2B). These results suggest that the pbUCNP ings on acutely dissected retinae from pbUCNP-injected and
non-injected mice (Figure 3A). The action spectra of rods from agreed well with the experimental NIR light PLR response (Fig-
pbUCNP-injected and non-injected mice were identical in the ure S5C). Therefore, the non-linearity shown in the NIR light-
visible light range, with differences only appearing after induced behavior was attributed to the non-linearity of the
900 nm, where the action spectra of rods from pbUCNP-in- upconversion process.
jected mice matched the excitation spectra of pbUCNPs (Fig- In addition to the above sub-conscious light sensation PLR
ure S3A). The rods from pbUCNP-injected mice had normal behavior, we also explored whether pbUCNP-injected mice
visible light (535 nm)-elicited photocurrents compared with could consciously perceive NIR light. In this regard, we
that of non-injected mice (Figures 3B and 3D). The 980-nm light performed light-dark box experiments with visible and NIR
flash elicited rod photocurrents from pbUCNP-injected mice light (Figures 4C and 4D) as well as light-induced fear-condition-
(Figure 3E), whereas the rods from non-injected mice exhibited ing experiments (Figures 4E and 4F). In the conventional light-
no responses (Figure 3C). The amplitude and kinetics of the dark box experiments with visible light, mice instinctively
980-nm light-elicited photocurrents were identical to those preferred the dark box to the light box illuminated with visible
activated by 535-nm visible light (Figures 3F–3H). The similar light. In our study, we replaced conventionally used visible
time-to-peak values suggest that, compared to the visible light with 980-nm LED light, which delivered 8.1 3 107
light stimulation, there was no delay in the activation of the photons 3 mm2 3 s1 at the center of the light box, equal to
rods by NIR light. Furthermore, the pbUCNPs did not alter a power density of 1.62 mW/cm2. The pbUCNP-injected mice
the light adaptation or dark noise characteristics of the rods, exhibited a significant preference for the dark box, whereas
and rods adapted to visible and NIR light in the same manner the non-injected control mice could not distinguish between
following the Weber-Fechner relationship (Baylor et al., 1980; the NIR light (980 nm)-illuminated and dark boxes (Figure 4D;
Morshedian et al., 2018; Fu et al., 2008) (Figure S4). To Video S1). This suggests that mice with injected nanoantennae
determine whether the pbUCNPs can serve as NIR nanosen- perceived NIR light and exhibited innate light-sensing behavior.
sors in vivo, we recorded the population response of photore- To exclude the possibility of any visible light emission from the
ceptors activated by light via ERGs (Dalke et al., 2004) NIR LEDs, the emission spectra of the 980-nm LEDs were
(Figure 3I). Upon 980-nm NIR light illumination to the eye, the measured and no light emission below 900 nm was detected
ERG from pbUCNP-injected mice resembled that of visible (Figures S5D and S5E).
light-induced responses, whereas no such signal could be de- We then tested whether such NIR light perception can serve
tected from the non-injected control mice. Furthermore, we as a visual cue for learned behavior. Mice were trained to pair
performed ERG recordings on pbUCNP-injected rod-function- a 20-s 535-nm light pulse to a 2-s foot shock (Figure 4E) in order
less mice (Gnat1/) and demonstrated that, through the to acquire a conditioned freezing behavior. After acquisition of
pbUCNPs, 980-nm NIR light indeed activated cones in vivo such conditioning, mice received either NIR light at 980 nm
(Figure S3B). or visible light at 535 nm as conditional stimuli (CS) in the test tri-
als. The pbUCNP-injected mice showed significant freezing
NIR Light Sensation of pbUCNP-Injected Mice behavior in response to both wavelengths, whereas the non-in-
To reveal whether pbUCNP-injected mice could see NIR light, jected control mice exhibited freezing behavior only to visible
we first performed pupillary light reflex (PLR) experiments (Xue light stimuli (Figure 4F; Video S2). These results clearly demon-
et al., 2011). The pupils of the pbUCNP-injected mice showed strated that mice acquired NIR light sensation and were able to
strong constrictions upon 980-nm light illumination, whereas ‘‘see’’ NIR light with our ocular injectable photoreceptor target-
the non-injected control mice did not exhibit PLR with the ing nanoantennae.
same NIR illumination (Figure 4A). Moreover, we discovered
that the PLR of the pbUCNP-injected mice was two orders of NIR Light-Activated Imaging Visual Pathway of
magnitude more sensitive to NIR light than that of the non-photo- pbUCNP-Injected Mice
receptor-binding paaUCNP-injected mice (Figure 4B). This was In addition to the NIR light sensation, we were curious whether
attributed to the proximity between the pbUCNPs and the bound pbUCNP-injected mice had acquired NIR light image visual
photoreceptors. Photon upconversion was measured (Fig- ability. In general, visual image perception is associated with
ure S5A) and showed a non-linear light intensity relationship activation of the visual cortex. In order to record visually evoked
plotted at the log-log scale (Figure S5B). We fitted the power potential (VEPs), we placed recording electrodes in six different
relationship between emitted 535-nm light and 980-nm excita- locations of the visual cortex (No. 1, 2, 3, and 5 in the monocular
tion light and determined the power to be 1.6. Interestingly, we areas and No. 4 and 6 in the binocular areas) during contralateral
found that this non-linearity was also shown in the NIR light- eye illumination (Cooke et al., 2015; Smith and Trachtenberg,
induced behavior. The light dose-response curves of the PLR 2007) (Figure 5A). When the visible 535-nm light pulse was
(normalized pupil area versus light intensity) were fitted to the applied, VEPs were detected at all locations in both the non-in-
Hill function. The NIR light-induced PLR dose-response curve jected controls and pbUCNP-injected mice (Figures 5B and
was steeper than that of visible light, and the Hill coefficients 5D). In contrast, under 980-nm NIR light illumination, no VEPs
for the NIR and visible light PLR dose-responses were 1.10 were observed in the control mice, but were detected from the
and 0.78, respectively (Figure S5C). To obtain the theoretical binocular visual cortical areas in pbUCNP-injected mice (Figures
NIR light PLR dose-response curve, the fitted upconversion 5C and 5E). This is topologically consistent with the pbUCNP in-
function was applied to the visible light PLR dose-response Hill jection site (temporal side, binocular projection area) in the
function. This theoretical NIR light PLR dose-response curve retina.
NIR Light Pattern Vision discriminate between different light patterns (Prusky et al.,
We next examined whether mice obtained NIR light pattern 2000) (Figure 6A). The mice were trained to find a hidden platform
vision. Accordingly, Y-shaped water maze behavioral experi- that was associated with one of two patterns. We designed
ments were conducted to determine whether mice could five different tasks to examine their NIR pattern vision ability
Figure 5. NIR Light Activated the Imaging Visual Pathway of pbUCNP-Injected Mice
(A) Diagram of six recording sites for visually evoked potentials (VEPs) in the mouse visual cortex.
(B and C) VEPs of non-injected control (black traces) and pbUCNP-injected mice (gray traces) under 535-nm (B) and 980-nm (C) light illumination. Intensities of the
535-nm and 980-nm lights were 3.37 3 103 and 7.07 3 108 photons 3 mm2 3 s1, respectively. Recording sites 1, 2, 3, and 5 were monocular areas; 4 and 6
were binocular areas. Traces were averaged from six sweeps and presented as mean ± SD (shaded area).
(D and E) Peak VEPs triggered by 535-nm (D) or 980-nm (E) light at each recording site (mean ± SD, n = 4 for both, two-sided t test, **p < 0.01, ***p < 0.001).
regarding different pattern stimuli and various background light that the sub-retinal injection of pbUCNPs did not interfere with
conditions. Task 1 used light gratings as pattern stimuli (Figures visible light vision. With respect to NIR light gratings, the
6B and S6A). After training with 980-nm light gratings, the pbUCNP-injected mice detected a maximum of 0.14 ± 0.06 cy-
pbUCNP-injected mice were able to discriminate between the cles/degree. This decrease in spatial resolution in NIR light vision
two orientations (vertical or horizontal) of the NIR light gratings, may be due to the isotropic radiation and scattering of the in situ
whereas the non-injected control mice made such choices in a transduced visible light from the NIR light-excited pbUCNPs
random manner (Figure 6C; Video S3). In the parallel control (Figures 6D and S6D).
testing, when the mice were trained and tested with visible light In addition, to confirm if visible light background interfered
gratings, both the pbUCNP-injected and non-injected mice were with the NIR light pattern perception, we designed Task 2 using
able to find the associated platform (Figures 6C, S6B, and S6C). two LED boards with visible (535 nm) and NIR (980 nm) LED
We then measured the spatial resolution of the NIR image arrays arranged in a perpendicular manner on each board. These
perception. The pbUCNP-injected mice detected the visible light two boards appeared identical under an ambient visible light
gratings with a maximum spatial frequency of 0.31 ± 0.04 cycles/ background when all LEDs (visible and NIR) were turned off.
degree, which did not significantly differ from that of the non-in- The orientations for the 535-nm and 980-nm LED stripes be-
jected control mice (0.35 ± 0.02 cycles/degree). This indicates tween the two boards were 90 rotated respectively (Figure 6E).
Trainings were carried out under visible room light (196 lux) and the test trials, we presented one visible (535 nm) and one NIR
with only the 980-nm LEDs on. In the tests, only pbUCNP-in- (980 nm) light in a triangular-circular pattern at the left-right
jected mice learnt to locate the platform (Figure 6F), indicating ends of the water maze, shuffled in a random sequence (Fig-
that NIR light pattern vision persisted in the visible light-illumi- ure 6I). Only the pbUCNP-injected mice were able to discrimi-
nated environment. Interestingly, we subsequently tested these nate between the two patterns with different shapes and
mice with the 535-nm LEDs on and 980-nm LEDs off. Both wavelengths (Figures 6J, left, and S6I). To exclude the possibility
pbUCNP-injected and control mice could discriminate the visible that mice simply used either visible or NIR light patterns to guide
light gratings, again indicating that pbUCNP injection did not decision-making rather than seeing them simultaneously, we
affect normal visible light vision. Additionally, pbUCNP-injected calculated the correct choice rates separately for the visible
mice could discriminate the visible light gratings from the begin- and NIR light triangle patterns. In the subset of stimuli where
ning of the test, suggesting that pbUCNP-injected mice were the triangular patterns was in visible light (Figure 6J, middle),
able to implement the rule learnt from the NIR light pattern to control mice selected both sides randomly, indicating they did
visible light pattern discrimination, indicating that NIR light pat- not simply use the visible triangle to make decisions. When the
terns did not differ perceptually from visible light patterns for circular pattern was in visible light, control mice still picked the
pbUCNP-injected mice (Figure 6F; Video S4). side randomly, indicating that the mice did not use the strategy
To test more sophisticated pattern vision, we further prompted of avoiding circles to make decisions (Figure 6J, right). In
animals to discriminate triangular and circular patterns in Task 3 contrast, pbUCNP-injected mice made correct choices in both
(Figure 6G). We found that pbUCNP-injected mice were able to cases (Video S5), suggesting they used visible and NIR light pat-
discriminate NIR and visible light patterns in the dark environ- terns together to guide behavior. These results clearly indicate
ment, whereas non-injected control mice could only detect the that the built-in nanoantennae enabled mice to see visible and
visible light pattern (Figures 6H, S6E, and S6F; Video S5), indi- NIR light patterns simultaneously.
cating that pbUCNP-injected mice could perceive sophisticated
NIR light patterns. We subsequently speculated whether back- DISCUSSION
ground NIR light would interfere with the visible light pattern
vision of pbUCNP-injected mice. Thus, in Task 4, mice were In this study, we demonstrated the successful application of
tested to discriminate between visible light triangles and circles UCNPs as ocular injectable NIR light transducers, which
under a visible or NIR light background (Figure 6G). Same as extended mammalian vision into the NIR realm. These implanted
control mice, the pbUCNP-injected mice did not behave differ- nanoantennae were proven to be biocompatible and did not
ently regarding their ability to discriminate visible light patterns interfere with normal visible light vision. Importantly, animals
under dark, visible, or NIR light backgrounds (Figures 6H and were able to detect NIR and visible light images simultaneously.
S6F–S6H; Video S5). These results clearly suggest that back-
ground NIR light does not interfere with visible light pattern Extension of the Visual Spectrum into the NIR Range
perception. One way to obtain NIR light vision is to implement new machinery
Task 5 was designed to test whether pbUCNP-injected mice for NIR photon transduction, such as the thermal detection of
could see NIR and visible light patterns simultaneously. In gen- snakes (Gracheva et al., 2010). However, a more plausible
eral, saturation by visible light is a common problem for conven- method to achieve such NIR photon detection is the use of the
tionally used tools, such as optoelectronic night vision devices or endogenous visual system. The method we developed here uti-
IR cameras, as it prevents smooth detection between visible and lized the very first step of the visual image perception process
NIR light objects. To test if our built-in NIR light vision could over- through photoreceptor outer-segment binding NIR nanoanten-
come this problem and coexist with visible light vision, we de- nae. The NIR light image was projected to the retina through
signed the following experiments. Mice were first trained in a the optical part of the eyes, cornea, and lens, after which the
Y-shaped water maze with visible light triangles and circles to pbUCNPs upconverted NIR light into visible light and then acti-
learn that the platform was associated with triangles only. During vated the bound photoreceptors. Subsequently, the retinal
circuit and cortical visual system generated perception of the not cause any separation between the photoreceptors and
NIR image. It is important to note that these injected nanoanten- retinal pigment epithelium, the latter of which is the supporting
nae did not interfere with natural visible light vision. The ability to layer for photoreceptors. As a result, neither inflammation nor
simultaneously detect visible and NIR light patterns suggests apoptosis occurred, which is in line with that of another reported
enhanced mammalian visual performance by extending the retinal application of rare earth nanoparticles (Chen et al., 2006).
native visual spectrum without genetic modification and avoiding The stability and compatibility of the pbUCNPs were also
the need for bulky external devices. This approach offers several demonstrated by successful detection of NIR light images,
advantages over the currently used optoelectronic devices, such even after 10 weeks without any repeated injections.
as no need for any external energy supply, and is compatible
with other human activities. Further Development of pbUCNPs
In the present study, we created NIR light vision while over-
Improved Efficiency through ConA Modification coming several key drawbacks that yet exist in currently used
of UCNPs man-made systems. It may also be possible to design NIR color
Regarding the practical applications of UCNPs, higher visual vision through multicolor NIR light-sensitive UCNPs that have
sensitivity and resolution are desirable. We modified the UCNPs multiple NIR light absorption peak wavelengths and correspond-
and generated photoreceptor-binding nanoparticles to increase ing multicolor visible light emissions. Further applications using
the proximity between the nanoparticles and photoreceptors. our strategy for visual repair and enhancement could also be
Thus, sensitivity to NIR light with respect to generating light- achieved by similar nanoparticles with tailored light absorptions.
induced behaviors was improved by two orders of magnitude. In addition, combined with a drug delivery system, these photo-
Therefore, it is now possible to use biocompatible low-power receptor-binding nanoparticles could be modified to release
NIR LEDs to elicit visual behavior in animals, rather than the small molecules locally upon light stimulation.
more invasive high-power NIR lasers inevitably used in conven- In summary, these nanoparticles not only provide the potential
tional UCNP biomedical applications (Chen et al., 2018; He et al., for close integration within the human body to extend the visual
2015). In the Y-shaped water maze experiment, we estimated spectrum, but also open new opportunities to explore a wide
that the 980-nm LED light was transduced to 535-nm light by variety of animal vision-related behaviors. Furthermore, they
the pbUCNPs at 293 photons 3 mm2 3 s1 intensity at the exhibit considerable potential with respect to the development
retina. The rod and cone-mediated visual behavior thresholds of bio-integrated nanodevices in civilian encryption, security,
of mice are 0.012 and 200 photons 3 mm2 3 s1 at the cornea, military operations, and human-machine interfaces, which
respectively (Sampath et al., 2005), equal to 0.003 and 50 pho- require NIR light image detection that goes beyond the normal
tons 3 mm2 3 s1 at the retina (Do et al., 2009). Therefore, in functions of mammals, including human beings. Moreover, in
our system, the 293 photons 3 mm2 3 s1 at the retina was addition to visual ability enhancement, this nanodevice can serve
adequate to activate both rod and cone photoreceptors, and in as an integrated and light-controlled system in medicine, which
practice, this NIR visual system was able to detect NIR light could be useful in the repair of visual function as well as in drug
that was of several magnitudes lower intensity than currently delivery for ocular diseases.
applied. Compared to rods, cones encode several orders of
magnitude higher intensity light and are more important for hu- STAR+METHODS
man high acuity vision. Thus, pbUCNP-bound cones may
mediate high-resolution NIR image pattern vision. Retinae also Detailed methods are provided in the online version of this paper
possess intrinsic photosensitive retinal ganglion cells (ipRGCs), and include the following:
which mediate non-image forming visual functions, such as pho-
toentrainment of the circadian rhythm (Do and Yau, 2010). Under d KEY RESOURCES TABLE
the intensity used in our behavioral experiment, NIR light did not d CONTACT FOR REAGENT AND RESOURCE SHARING
activate ipRGCs (Figure S5F), which was likely due to the longer d EXPERIMENTAL MODEL AND SUBJECT DETAILS
distance of ipRGCs to pbUCNPs and their low sensitivity (Do B Mice
et al., 2009). With respect to NIR image spatial resolution, d METHOD DETAILS
pbUCNP-injected mice had good NIR eye sight (0.14 ± 0.06 cy- B Synthesis of pbUCNPs
cles/degree, half of the visible image resolution), allowing them B Sub-retinal injection
to see sophisticated NIR light patterns. B Distribution and spectrum analysis
B Retinal histology
Biocompatible NIR Nanoantennae B TUNEL Apoptosis detection
Sub-retinal injection in humans is a common practice in ophthal- B Microglia staining
mological treatment (Hauswirth et al., 2008; Peng et al., 2017). B Single cell electrophysiology
The implantation of microscale sub-retinal devices is a potential B Electroretinography
method of repairing vision following retinal photoreceptor B Pupillary light reflex
degeneration, though current devices can lead to biocompati- B Light-dark box
bility issues, such as retinal detachment, fibrosis, and inflamma- B Light induced fear conditioning
tion (Zrenner, 2013). Yet, this did not occur in our system, as the B Visually evoked potential
intimate contact between the pbUCNPs and photoreceptors did B Y shaped water maze
d QUANTIFICATION AND STATISTICAL ANALYSIS Boynton, R.M. (1996). Frederic Ives Medal paper. History and current status of
B Imaging quantifications a physiologically based system of photometry and colorimetry. J. Opt. Soc.
B Statistics Am. A Opt. Image Sci. Vis. 13, 1609–1621.
B Fitting procedures Bridges, C.D.B. (1981). Lectin receptors of rods and cones. Visualization by
fluorescent label. Invest. Ophthalmol. Vis. Sci. 20, 8–16.
SUPPLEMENTAL INFORMATION Bridges, C.D.B., and Fong, S.L. (1980). Lectin receptors on rod and cone
membranes. Photochem. Photobiol. 32, 481–486.
Supplemental Information can be found with this article online at https://doi. Burstyn-Cohen, T., Lew, E.D., Través, P.G., Burrola, P.G., Hash, J.C., and
org/10.1016/j.cell.2019.01.038. Lemke, G. (2012). Genetic dissection of TAM receptor-ligand interaction in
A video abstract is available at https://doi.org/10.1016/j.cell.2019.01. retinal pigment epithelial cell phagocytosis. Neuron 76, 1123–1132.
038#mmc7.
Calvert, P.D., Krasnoperova, N.V., Lyubarsky, A.L., Isayama, T., Nicoló, M.,
Kosaras, B., Wong, G., Gannon, K.S., Margolskee, R.F., Sidman, R.L., et al.
ACKNOWLEDGMENTS
(2000). Phototransduction in transgenic mice after targeted deletion of the
rod transducin alpha -subunit. Proc. Natl. Acad. Sci. USA 97, 13913–13918.
We thank members of the Neuroscience Pioneer Club for valuable discus-
sions. We thank Drs. Dangsheng Li and Zilong Qiu for reading this manuscript Chen, J., Patil, S., Seal, S., and McGinnis, J.F. (2006). Rare earth nanoparticles
and providing helpful comments. We thank Dr. Yang Xiang from University of prevent retinal degeneration induced by intracellular peroxides. Nat. Nano-
Massachusetts Medical School for helpful discussion. We thank Min Wei and technol. 1, 142–150.
Shouzhen Li for helpful discussions and Jiawei Shen and Dr. Huan Zhao for Chen, S., Weitemier, A.Z., Zeng, X., He, L., Wang, X., Tao, Y., Huang, A.J.Y.,
helping illustration drawing and videos clipping. We thank Dr. Kai Huang and Hashimotodani, Y., Kano, M., Iwasaki, H., et al. (2018). Near-infrared deep
Mr. Nuo Yu for helping characterizing nanoparticles. We thank Dr. Qiuping brain stimulation via upconversion nanoparticle-mediated optogenetics. Sci-
Wang and Xiaokang Ding from National Synchrotron Radiation Laboratory ence 359, 679–684.
for helping with the measurement of various spectra of light sources. We thank
Cooke, S.F., and Bear, M.F. (2013). How the mechanisms of long-term synap-
Dr. Yuen Wu and Xing Wang from School of Chemistry and Materials Science
tic potentiation and depression serve experience-dependent plasticity in pri-
(USTC) for helping with the measurement of the absorption spectrum of
pbUCNPs. We acknowledge support from the National Key Research and mary visual cortex. Philos. Trans. R. Soc. Lond. B Biol. Sci. 369, 20130284.
Development Program of China (2016YFA0400900), the Strategic Priority Cooke, S.F., Komorowski, R.W., Kaplan, E.S., Gavornik, J.P., and Bear, M.F.
Research Program of the Chinese Academy of Science (XDA16020603, (2015). Erratum: Visual recognition memory, manifested as long-term habitua-
XDPB10, XDB02010000), the National Young Scientists 973 Program of China tion, requires synaptic plasticity in V1. Nat. Neurosci. 18, 926.
(2013CB967700), the National Natural Science Foundation of China
Cui, Z.Z., Feng, R.B., Jacobs, S., Duan, Y.H., Wang, H.M., Cao, X.H., and
(81790644, 61890953, 31322024, 81371066, 91432104, 31571073,
Tsien, J.Z. (2013). Increased NR2A: NR2B ratio compresses long-term depres-
81401025, 61727811, 91748212), the NIH (R01MH103133 to G.H.), a UMass
sion range and constrains long-term memory. Sci. Rep. 3, 1036.
OTCV award, a Worcester Foundation Mel Cutler Award to G.H., and the
Human Frontier Science Program (RGY-0090/2014). Dalke, C., Löster, J., Fuchs, H., Gailus-Durner, V., Soewarto, D., Favor, J.,
Neuhäuser-Klaus, A., Pretsch, W., Gekeler, F., Shinoda, K., et al. (2004). Elec-
troretinography as a screening method for mutations causing retinal dysfunc-
AUTHOR CONTRIBUTIONS
tion in mice. Invest. Ophthalmol. Vis. Sci. 45, 601–609.
Conceptualization, T.X. and G.H.; Methodology, Y.M., J.B., G.H., and T.X.; Desai, N. (2012). Challenges in development of nanoparticle-based therapeu-
Investigation, Y.M., J.B., Y. Zhang, Z.L., L.H., Y. Zhao, X.Z., C.W., G.H., and tics. AAPS J. 14, 282–295.
T.X.; Validation, J.B., G.H., and T.X.; Formal Analysis, Y.M. and J.B.; Writing – Do, M.T.H., and Yau, K.W. (2010). Intrinsically photosensitive retinal ganglion
Original Draft, J.B., Y.M., Y. Zhang, G.H., and T.X.; Writing – Review & Editing, cells. Physiol. Rev. 90, 1547–1581.
J.B., Y.M., G.H., and T.X.; Funding Acquisition, J.B., G.H., and T.X.; Supervi-
sion, J.B., G.H., and T.X. Do, M.T., Kang, S.H., Xue, T., Zhong, H., Liao, H.W., Bergles, D.E., and Yau,
K.W. (2009). Photon capture and signalling by melanopsin retinal ganglion
cells. Nature 457, 281–287.
DECLARATION OF INTERESTS
Dong, A., Ye, X., Chen, J., Kang, Y., Gordon, T., Kikkawa, J.M., and Murray,
T.X. and G.H. have a patent application related to this work. C.B. (2011). A generalized ligand-exchange strategy enabling sequential sur-
face functionalization of colloidal nanocrystals. J. Am. Chem. Soc. 133,
Received: June 6, 2018 998–1006.
Revised: November 9, 2018 Dubois, E. (2009). The Structure and Properties of Color Spaces and the Rep-
Accepted: January 24, 2019 resentation of Color Images (Morgan & Claypool Publishers).
Published: February 28, 2019
Fu, Y., Kefalov, V., Luo, D.G., Xue, T., and Yau, K.W. (2008). Quantal noise from
human red cone pigment. Nat. Neurosci. 11, 565–571.
REFERENCES
Govardovskii, V.I., Fyhrquist, N., Reuter, T., Kuzmin, D.G., and Donner, K.
Ala-Laurila, P., Albert, R.J., Saarinen, P., Koskelainen, A., and Donner, K. (2000). In search of the visual pigment template. Vis. Neurosci. 17, 509–528.
(2003). The thermal contribution to photoactivation in A2 visual pigments stud- Gracheva, E.O., Ingolia, N.T., Kelly, Y.M., Cordero-Morales, J.F., Hollopeter,
ied by temperature effects on spectral properties. Vis. Neurosci. 20, 411–419. G., Chesler, A.T., Sánchez, E.E., Perez, J.C., Weissman, J.S., and Julius, D.
Baylor, D.A., Matthews, G., and Yau, K.W. (1980). Two components of electri- (2010). Molecular basis of infrared detection by snakes. Nature 464,
cal dark noise in toad retinal rod outer segments. J. Physiol. 309, 591–621. 1006–1011.
Bieber, M.L., Volbrecht, V.J., and Werner, J.S. (1995). Spectral efficiency Hauswirth, W.W., Aleman, T.S., Kaushal, S., Cideciyan, A.V., Schwartz, S.B.,
measured by heterochromatic flicker photometry is similar in human infants Wang, L., Conlon, T.J., Boye, S.L., Flotte, T.R., Byrne, B.J., and Jacobson,
and adults. Vision Res. 35, 1385–1392. S.G. (2008). Treatment of leber congenital amaurosis due to RPE65 mutations
Bourin, M., and Hascoët, M. (2003). The mouse light/dark box test. Eur. J. by ocular subretinal injection of adeno-associated virus gene vector: short-
Pharmacol. 463, 55–65. term results of a phase I trial. Hum. Gene Ther. 19, 979–990.
He, S., Krippes, K., Ritz, S., Chen, Z., Best, A., Butt, H.J., Mailänder, V., and Roberts, J.E. (1961). Lanthanum and neodymium salts of trifluoroacetic acid.
Wu, S. (2015). Ultralow-intensity near-infrared light induces drug delivery by J. Am. Chem. Soc. 83, 1087–1088.
upconverting nanoparticles. Chem. Commun. (Camb.) 51, 431–434. Rutishauser, U., and Sachs, L. (1975). Cell-to-cell binding induced by different
Karnas, D., Hicks, D., Mordel, J., Pévet, P., and Meissl, H. (2013). Intrinsic lectins. J. Cell Biol. 65, 247–257.
photosensitive retinal ganglion cells in the diurnal rodent, Arvicanthis ansorgei. Sampath, A.P., Strissel, K.J., Elias, R., Arshavsky, V.Y., McGinnis, J.F., Chen,
PLoS ONE 8, e73343. J., Kawamura, S., Rieke, F., and Hurley, J.B. (2005). Recoverin improves rod-
Krady, J.K., Basu, A., Allen, C.M., Xu, Y., LaNoue, K.F., Gardner, T.W., and mediated vision by enhancing signal transmission in the mouse retina. Neuron
Levison, S.W. (2005). Minocycline reduces proinflammatory cytokine expres- 46, 413–420.
sion, microglial activation, and caspase-3 activation in a rodent model of dia- Schmidt, T.M., and Kofuji, P. (2010). Differential cone pathway influence on
betic retinopathy. Diabetes 54, 1559–1565. intrinsically photosensitive retinal ganglion cell subtypes. J. Neurosci. 30,
Le, Y.Z., Ash, J.D., Al-Ubaidi, M.R., Chen, Y., Ma, J.X., and Anderson, R.E. 16262–16271.
(2004). Targeted expression of Cre recombinase to cone photoreceptors in Schnapf, J.L., Kraft, T.W., Nunn, B.J., and Baylor, D.A. (1988). Spectral sensi-
transgenic mice. Mol. Vis. 10, 1011–1018. tivity of primate photoreceptors. Vis. Neurosci. 1, 255–261.
Liu, Y., Lu, Y., Yang, X., Zheng, X., Wen, S., Wang, F., Vidal, X., Zhao, J., Liu, D., Smith, S.L., and Trachtenberg, J.T. (2007). Experience-dependent binocular
Zhou, Z., et al. (2017). Amplified stimulated emission in upconversion nanopar- competition in the visual cortex begins at eye opening. Nat. Neurosci. 10,
ticles for super-resolution nanoscopy. Nature 543, 229–233. 370–375.
Luo, D.G., Yue, W.W.S., Ala-Laurila, P., and Yau, K.W. (2011). Activation of vi- Soucy, E., Wang, Y., Nirenberg, S., Nathans, J., and Meister, M. (1998). A novel
sual pigments by light and heat. Science 332, 1307–1312. signaling pathway from rod photoreceptors to ganglion cells in mammalian
retina. Neuron 21, 481–493.
Mai, H.X., Zhang, Y.W., Si, R., Yan, Z.G., Sun, L.D., You, L.P., and Yan, C.H.
(2006). High-quality sodium rare-earth fluoride nanocrystals: controlled syn- St George, R.C.C. (1952). The interplay o light and heat in bleaching rhodopsin.
thesis and optical properties. J. Am. Chem. Soc. 128, 6426–6436. J. Gen. Physiol. 35, 495–517.
Wang, D., Li, Y., Wang, Z., Sun, G.Y., and Zhang, Q.H. (2013). Nimodipine res-
Mitragotri, S., Anderson, D.G., Chen, X., Chow, E.K., Ho, D., Kabanov, A.V.,
cues N-methyl-N-nitrosourea-induced retinal degeneration in rats. Pharma-
Karp, J.M., Kataoka, K., Mirkin, C.A., Petrosko, S.H., et al. (2015). Accelerating
cogn. Mag. 9, 149–154.
the translation of nanomaterials in biomedicine. ACS Nano 9, 6644–6654.
Wu, S., Han, G., Milliron, D.J., Aloni, S., Altoe, V., Talapin, D.V., Cohen, B.E.,
Morshedian, A., Woodruff, M.L., and Fain, G.L. (2018). Role of recoverin in rod
and Schuck, P.J. (2009). Non-blinking and photostable upconverted lumines-
photoreceptor light adaptation. J Physiol 596, 1513–1526.
cence from single lanthanide-doped nanocrystals. Proc. Natl. Acad. Sci. USA
Namekata, K., Kimura, A., Kawamura, K., Guo, X., Harada, C., Tanaka, K., and 106, 10917–10921.
Harada, T. (2013). Dock3 attenuates neural cell death due to NMDA neurotox-
Wu, X., Chen, G., Shen, J., Li, Z., Zhang, Y., and Han, G. (2015). Upconversion
icity and oxidative stress in a mouse model of normal tension glaucoma. Cell
nanoparticles: a versatile solution to multiscale biological imaging. Bioconjug.
Death Differ. 20, 1250–1256.
Chem. 26, 166–175.
Peng, Y., Tang, L., and Zhou, Y. (2017). Subretinal injection: a review on the Wyszecki, G., and Stiles, W.S. (1982). Color Science: Concepts and Methods,
novel route of therapeutic delivery for vitreoretinal diseases. Ophthalmic Quantitative Data and Formulae, Second edition (John Wiley and Sons).
Res. 58, 217–226.
Xue, T., Do, M.T.H., Riccio, A., Jiang, Z., Hsieh, J., Wang, H.C., Merbs, S.L.,
Pittler, S.J., and Baehr, W. (1991). Identification of a nonsense mutation in the Welsbie, D.S., Yoshioka, T., Weissgerber, P., et al. (2011). Melanopsin signal-
rod photoreceptor cGMP phosphodiesterase beta-subunit gene of the rd ling in mammalian iris and retina. Nature 479, 67–73.
mouse. Proc. Natl. Acad. Sci. USA 88, 8322–8326.
Zhao, L., Ma, W., Fariss, R.N., and Wong, W.T. (2011). Minocycline attenuates
Prusky, G.T., West, P.W.R., and Douglas, R.M. (2000). Behavioral assessment photoreceptor degeneration in a mouse model of subretinal hemorrhage mi-
of visual acuity in mice and rats. Vision Res. 40, 2201–2209. croglial: inhibition as a potential therapeutic strategy. Am. J. Pathol. 179,
Qi, Y., Dai, X., Zhang, H., He, Y., Zhang, Y., Han, J., Zhu, P., Zhang, Y., Zheng, 1265–1277.
Q., Li, X., et al. (2015). Trans-corneal subretinal injection in mice and its effect Zrenner, E. (2013). Fighting blindness with microelectronics. Sci. Transl. Med.
on the function and morphology of the retina. PLoS ONE 10, e0136523. 5, 210ps16.
STAR+METHODS
Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Other
65 RN 5uL SYR W/O NEEDLE Hamilton, Switzerland Cat# 7633-01
RN Needle (34/8 mm/3) Hamilton, Switzerland Cat# 207343
535 nm LED Starsealand, China Cat# XL001WP01WBGC/535
980 nm LED Starsealand, China Cat# XL001WP01IRC/980
Photometer Newport 1936-R
Spectrometer Avantes ULS2048
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Tian Xue
(xuetian@ustc.edu.cn).
Mice
The experimental procedures on animals followed the guidelines of the Animal Care and Use committee of University of Science and
Technology of China. Mice were kept under Specific pathogen Free (SPF) housing facilities, with lighting period of 12h:12h (L:D), con-
stant temperature at 20-24 C and humidity around 40%–70%. The mouse lines in the Key Resources Table and their crossings have
been used in this study at the age of 2-3 months. Confirming non-injected mice and PBS-injected mice are identical in retinal histol-
ogy, we used non-injected mice as control mice in most other experiments. Both male and female mice were used in all experiments.
METHOD DETAILS
Synthesis of pbUCNPs
General chemicals
Y2O3 (99.9%), Yb2O3 (99.9%), Er2O3 (99.9%), CF3COONa (99.9%), CF3COOH, 1-octadecene, oleic acid, oleylamine, and other
organic solvents were purchased from Sigma-Aldrich and used directly without further purification. Lanthanide trifluoroactates,
Ln(CF3COO)3 were prepared according to literature method (Roberts, 1961).
Synthesis of b-NaYF4:20%Yb, 2%Er core
The b-NaYF4:20%Yb,2%Er core UCNPs were prepared by a modified two-step thermolysis method (Mai et al., 2006). In the first step,
CF3COONa (0.5 mmol) and Ln (CF3COO)3 ((Y+Yb+Er) 0.5 mmol in total, Y:Yb:Er = 78%:20%:2%) precursors were mixed with oleic
acid (5 mmol), oleyamine (5 mmol) and 1-octadecene (10 mmol) in a two-neck reaction flask. The slurry mixture was heated to 110 C
to form a transparent solution followed by 10 minutes of degassing. Then the flask was heated to 300 C with a rate of 15 C/min under
dry argon flow, and it maintained at 300 C for 30 minutes. The b-NaYF4: Ln intermediate UCNPs were gathered from the cooled re-
action solution by centrifugal washing with excessive ethanol (7500 RCF, 30 min). In the second step, the b-NaYF4: Ln intermediate
UCNPs were re-dispersed into oleic acid (10 mmol) and 1-octadecene (10 mmol) together with CF3COONa (0.5 mmol) in a new two-
neck flask. After degassing at 110 C for 10 minutes, this flask was heated to 325 C with a rate of 15 C/min under dry argon flow, and
remained at 325 C for 30 minutes. Then, b-NaYF4: Ln UCNPs were centrifugally separated from the cooled reaction media and pre-
served in hexane (10 mL) as stock solution.
Synthesis of b-NaYF4:20%Yb,2%Er@b-NaYF4 core/shell UCNPs
In this thermolysis reaction, as-synthesized-NaYF4:20%Yb, 2%Er UCNPs served as cores for the epitaxial growth of undoped-
NaYF4 shells. Typically, a stock solution of b-NaYF4: 20%Yb, 2%Er UCNPs (5 mL, ca. 1 mmol/L core UCNPs) was transferred into
a two-neck flask and hexane was sequentially removed by heating. CF3COONa (0.5 mmol) and Y(CF3COO)3 (0.5 mmol) were added
along with oleic acid (10 mmol) and 1-octadecene (10 mmol). After 10 minutes of degassing at 110 C, the flask was heated to 325 C
at a rate of 15 C /min under dry argon flow and was kept at 325 C for 30 minutes. The products were precipitated by adding 20 mL
ethanol to the cooled reaction flask. After centrifugal washing with hexane/ethanol (7500 RCF, 30 min), the core/shell UCNPs were
collected and re-dispersed in 10 mL of hexane.
Synthesis of pbUCNPs
As synthesized b-NaYF4:20%Yb,2%Er@b-NaYF4 UCNPs were first treated by surface ligand exchange using a modified literature
method (Dong et al., 2011). Generally, nitrosonium tetrafluoroborte/DMF solution (0.2 g NOBF4, 5 mL DMF) was added into 1 mL
UCNPs hexane stock solution, followed by 4 mL hexane and 3 hours of stirring at room temperature. Then oleic acid-free UCNPs
were precipitated by adding 5 mL isopropanol and purified by centrifugal wash with DMF. UCNPs solids were re-dispersed in poly
(acrylic acid)/DMF (10 mg/mL, 5 mL) solution to coated UCNPs surface with PAA. After overnight stirring, PAA coated UCNPs
(paaUCNPs) were purified by centrifugal and wash with DI-water. Then ConA proteins were conjugated to paaUCNPs surface by
tranditional EDC/NHS coupling. Generally, 10 mg paaUCNPs in 1 mL DI-water were treated with 1 mL EDC/NHS water solution
(1 g/L). After stirring at room temperature for 1 hour, 30 mL ConA solution was introduced (5 g/L) and the mixture was further stirred
overnight. The pbUCNPs were purified by washing with deionized water, centrifugation and dispersed in water for further use.
Sub-retinal injection
For sub-retinal injection, pupils were dilated with atropine (100 mg/mL, Sigma-Aldrich), and animals were anesthetized by Avertin
(450 mg/kg, Sigma-Aldrich). A 33 Gauge needle was inserted through the cornea to release the intra-ocular pressure. Nanoparticles
dissolved in 2 mL sterile PBS to reach 25 mg/ml was injected into the sub-retinal space through a beveled, 34-gauge hypodermic
needle (Hamilton, Switzerland). During and after the injection the animal was kept on a warming blanket and eyes were kept wet
to avoid cataract.
Retinal histology
To analyze whether nanoparticles are potentially toxic to retina, we injected nanoparticles in different concentrations and then per-
formed hematoxylin-eosin (HE) staining on fixed retinal slices (Burstyn-Cohen et al., 2012). Cell bodies of photoreceptors were
located in the outer nucleus layer (ONL) and we counted the number of cell layers as a parameter to evaluate the damage. The number
of cell layers were counted at 5 different locations of injection sites and repeated at 5 randomly selected different slices of each retina,
then averaged.
Microglia staining
To detect immune reactions in nanoparticles injected retinas, we implemented Iba1 (ionized calcium binding adaptor molecule 1, one
marker protein of microglia) staining assay. Retinal slices were washed twice by PBS solution and then blocked in 1% Triton X-100
(Sangon-A110694) and 5% goat serum solution (blocking solution) for 2-3 hours. Then slices were incubated in blocking solution with
rabbit anti-Iba1 antibody (Wako-019-19741, 1:1000) at 4 C overnight. Afterward, slices were washed 3 times by PBS solution,and
then incubated in Alexa Fluor 568 goat anti-rabbit IgG (H+L) secondary antibody (Thermo Fisher Scientific-1832035, 1:800) at
room temperature for 2-3 hours. Finally, retinal slices were incubated in DAPI-PBS solution for 5 min and then washed 3 times by
PBS solution. All retinal slices were scanned by Leica two-photon microscope to analyze retinas immune activities. The number
of positive cells were counted at 6 different locations of injection site from each retina and then averaged.
535-nm light was from a filter block in front of a white light LED. Infrared light was generated by 980-nm laser. Flash light intensities in
Figures 3B–3E, S4A, S4B, S4D, S4E, S4G, and S4H: 535-nm - 0.64, 2.57, 12.55, 49.38, 162.5, 1102.21; 980-nm - 1.50 3 105, 6.31 3
105, 1.85 3 106, 5.81 3 106, 2.26 3 107, 1.17 3 108; unit - photons$mm-2.
To measure the action spectrum of rods, light flashes in a series of wavelengths were delivered to elicit flash photoresponses. All
the recorded rods were given a strong light test pulse to check if a normal saturating photocurrent can be elicited, before the family of
light pulses with different wavelengths was given. Flash light intensities were selected to generate photocurrents within the linear
range. Light pulses were from LEDs or lasers. The sensitivity was calculated as photoresponse / (flash intensity 3 area) and then
normalized to the peak sensitivity.
To quantify light adaptation, we applied background light and allowed at least 2-min adaptation for the recorded rod (Karnas et al.,
2013; Schmidt and Kofuji, 2010). Then on top of the background light, a light flash was applied to elicit a transient photocurrent. Flash
sensitivity (SF) was calculated by dividing the peak amplitude of the transient photocurrent with the intensity of the flash light. For a
series of applied background light with different intensities, flash sensitivities were measured and normalized to the flash sensitivity in
the dark (SD). The background light intensities are: 535-nm - dark, 12.9, 49.20, 113.24, 266, 542.79, 2700.06, 5965.32; 980-nm - dark,
4.08 3 106, 1.03 3 107, 2.02 3 107, 4.60 3 107, 7.44 3 107, 1.94 3 108, 4.36 3 108; unit - photons$mm-2$s-1). The intensity-response
curves except under dark background were measured under three different conditions: visible light flashes on top of a visible or NIR
light background (535-nm - 776 photons$mm-2$s-1 or 980-nm - 1.3 3 108 photons$mm-2$s-1) and NIR light flashes on top of a visible
light background (535-nm - 776 photons$mm-2$s-1).
Data were lowpass filtered at 50 Hz and sampled at 25 kHz by Axon 700B Amplifier and Digital 1440A interface. Data were analyzed
with custom routines in Origin 8.0 and presented as mean ± SD. Single cell electrophysiology was carried out 5-6 weeks after
pbUCNPs injection. Data for noise analysis of rods were filtered with low-pass Bessel filter (cutoff frequency - 5Hz) (Baylor et al.,
1980; Fu et al., 2008). Power spectra density was calculated with function ‘periodogram()’ in MATLAB (Mathworks, USA).
Electroretinography
Mice were anesthetized by Avertin (450 mg/kg, Sigma-Aldrich) after their pupils were dilated with atropine (100 mg/mL, Sigma-Al-
drich). During the experiment, the anesthetized animal was kept on a warming blanket and eyes were kept wet to avoid cataract
(Dalke et al., 2004). Mice were placed into a Faraday cage and a glass recording electrode with a tip diameter of 10 mm was put tightly
against the center of the cornea. A ground electrode was inserted into subcutaneous space of the tail and a reference electrode was
inserted into subcutaneous space of the head. A 535-nm LED light (8.26 3 103 photons$mm-2) and a 980-nm laser beam with a spot
diameter of 1.8 mm (9.83 3 108 photons$mm-2) was placed in front of the pupil for stimulation. Data acquisition was carried out by a
differential amplifier (AM-SYSTEM INC) and Digital 1440A (Axon CNS). Data were analyzed with custom routines in Origin 8.0 (Origin
Lab Corp). ERG was carried out 4 weeks after pbUCNPs injection.
Light-dark box
Mice were placed in a 59 cm 3 28.5 cm 3 28.5 cm custom-made light and dark double box (Bourin and Hascoët, 2003). On the
three sides of the light box, 20 980-nm LEDs (1 Watt) and twenty 535-nm LEDs (1 Watt) were evenly placed for light stimulation.
Intensity of 980 nm light at the center of the light box was 8.1 3 107 photons$mm-2$s-1, and intensity of 535-nm light was 9.1 3
102 photons$mm-2$s-1. Animals were introduced to the box and allowed for 5-min adaptation. A series of light stimulation in the order
of 5 min in dark, 5 min in 980-nm light, and 5 mins in 535-nm light was programed. All these experiments were carried out in the dark
environment and videos were acquired by an infrared camera and custom-made software. Experiments were performed 4-5 weeks
after injection.
Imaging quantifications
Fluorescence analysis was based on careful match of the confocal imaging parameters. The intensities of pixels or the number of
cells were then quantified in ImageJ (NIH).
Statistics
Statistical analysis was performed using Microsoft Excel or Matllab softwares. Unpaired two-tailed Student’s test was used to deter-
mine statistical significance. The ‘‘n’’ numbers for each experiment are provided in the text and figure legends. Data are all presented
as mean ± SD. For immunocytochemistry and toxicological detection experiments were repeated on at least 3 animals.
Fitting procedures
Fittings were conducted in MATLAB (Mathworks, USA) with an unconstrained nonlinear optimization routine. The upconversion rela-
tionship was fitted by a linear function in the log-log scale. The dose response curve of PLR - the normalized pupil area versus light
intensity - was fitted by the Hill function. The action spectrum of rods was fitted to absorption-spectrum template (Fu et al., 2008;
Govardovskii et al., 2000). The relative flash sensitivity of background light adapted rods was fitted to Weber-Fechner equation (Bay-
lor et al., 1980; Morshedian et al., 2018; Fu et al., 2008).
Figure S1. Properties of UCNPs and Distributions in Subretinal Space, Related to Figure 1
(A) Fourier Transform-Infrared (FT-IR) spectra of UCNPs before (black) and after (red) ConA surface modifications. After ConA conjugation, new peaks at 1681 and
1456 cm-1 emerged, which are attributed to amide bond formation.
(B) Absorption spectrum of UCNPs before (black) and after (red) ConA conjugation.
A new absorption peak at 285 nm is assigned to ConA. NIR range absorption is shown in inset, and the absorption peak at 980 nm is attributed to the absorption
for upconversion.
(C) Dynamic light scattering (DLS) spectrum of pbUCNPs (0.2 mg/mL) upon the introduction of different concentrations of b-cyclodextrins (0, 50, 100 and 200 nM)
showing aggregation of pbUCNPs.
(D) Dynamic light scattering spectrum of paaUCNPs (0.2 mg/mL) upon the introduction of different concentrations of b-cyclodextrins (0, 50, 100 and 200 nM).
(E) Illustration of subretinal injection of pbUCNPs.
(F) Distribution of pbUCNPs (green) in the subretinal space after a single injection. Scale bar: 250 mm.
(G) Quantified green light intensity distribution along the spread of pbUCNPs in the subretinal space (n = 4). Data are mean ± SD. X axis is the distance from the
injection site and y axis is the total green pbUCNPs emission intensity from a 50 3 50 mm2 measuring window on a retina slice along the distribution of pbUCNPs
from the injection site.
(legend on next page)
Figure S2. Toxicity and Biocompatibility Evaluation of Retina Slices in a Time Series up to 10 Weeks after Injection, Related to Figure 2
(A) Microglia marker Iba1 staining of retina slices. Retinal injection of 6 mM H2O2 was used as the positive control. Red: Iba1; Green: pbUCNPs emission upon
excitation by NIR light; Blue: DAPI (4’, 6-diamidino-2-phenylindole) signal indicating cell nucleuses. Scale bar: 50 mm.
(B) TUNEL staining of retina slices. Retinal injection of 6 mM H2O2 was used as the positive control. Red: TUNEL staining; Green: pbUCNPs emission upon
excitation by NIR light; Blue: DAPI. Scale bar: 50 mm.
(C) and (D) Excitation (C) and emission (D) spectra of pbUCNPs in solution (black), fixed retina (gray) and fresh retina (violet).
Figure S3. Action Spectra of Rods and ERG of Gnat1/ Mice, Related to Figure 3
(A) Left: Action spectra of rods from non-injected (gray) and pbUCNP-injected (red) retina. Fit of action spectrum of rods from non-injected retina to absorption-
spectrum template (Fu et al., 2008; Govardovskii et al., 2000) is shown in dashed blue line. Open squares are theoretical predicted values from the fit under
detection threshold. The predicted sensitivity of non-injected rods to the NIR light (980 nm) is extremely small. With the help of pbUCNPs, rods gained 23 orders of
magnitude higher sensitivity to 980-nm NIR light. Right: Same action spectra were plotted with excitation spectrum of pbUCNPs (cyan) normalized to the action
spectrum value at 980 nm of rods from pbUCNP-injected mice. All data are mean ± SD.
(B) ERG recordings of Gnat1/ mice. Light intensities are 7.26 3 103 photons$mm-2 for 535-nm light and 8.01 3 107 photons$mm-2 for 980-nm light.
Figure S4. Adaptation and Noise Properties of Rods from pbUCNP-Injected Retina, Related to Figure 3
(A) Visible light flash responses of rod photoreceptors under dark and visible light background. Background and flash light stimulations are illustrated on top of the
recording traces (A, D and G).