Article

Magnetic sensitivity of cryptochrome 4 from

a migratory songbird

https://doi.org/10.1038/s41586-021-03618-9 Jingjing Xu1, Lauren E. Jarocha2, Tilo Zollitsch2, Marcin Konowalczyk3, Kevin B. Henbest2,3,

Sabine Richert4, Matthew J. Golesworthy3, Jessica Schmidt1, Victoire Déjean3,
Received: 17 July 2019
Daniel J. C. Sowood2, Marco Bassetto1,2, Jiate Luo2, Jessica R. Walton2, Jessica Fleming2,
Accepted: 6 May 2021 Yujing Wei2, Tommy L. Pitcher3, Gabriel Moise3, Maike Herrmann1, Hang Yin5, Haijia Wu6,
Rabea Bartölke1, Stefanie J. Käsehagen1, Simon Horst1, Glen Dautaj1, Patrick D. F. Murton2,
Published online: 23 June 2021
Angela S. Gehrckens2, Yogarany Chelliah7,8, Joseph S. Takahashi7,8, Karl-Wilhelm Koch6,9,
Check for updates Stefan Weber4, Ilia A. Solov’yov9,10 ✉, Can Xie11,12 ✉, Stuart R. Mackenzie2 ✉,
Christiane R. Timmel3,13 ✉, Henrik Mouritsen1,9 ✉ & P. J. Hore2 ✉

Night-migratory songbirds are remarkably proficient navigators1. Flying alone and

often over great distances, they use various directional cues including, crucially, a
light-dependent magnetic compass2,3. The mechanism of this compass has been
suggested to rely on the quantum spin dynamics of photoinduced radical pairs in
cryptochrome flavoproteins located in the retinas of the birds4–7. Here we show that
the photochemistry of cryptochrome 4 (CRY4) from the night-migratory European
robin (Erithacus rubecula) is magnetically sensitive in vitro, and more so than CRY4
from two non-migratory bird species, chicken (Gallus gallus) and pigeon (Columba
livia). Site-specific mutations of ErCRY4 reveal the roles of four successive flavin–
tryptophan radical pairs in generating magnetic field effects and in stabilizing
potential signalling states in a way that could enable sensing and signalling functions
to be independently optimized in night-migratory birds.

The radical-pair mechanism is a well-established source of magnetic field passerine birds. So far, no cryptochrome from a migratory animal that
effects on the rates and yields of chemical reactions8. Proof-of-principle unequivocally uses a light-dependent magnetic compass has been
experiments have demonstrated the sensitivity of a model radical-pair shown to exhibit magnetic sensitivity.
system to the direction of an Earth-strength magnetic field (around Magnetic field effects on the coherent spin dynamics of light-induced
50 μT) via magnetic interactions that are a million times smaller than radical pairs in cryptochromes are manifested as changes in the quan-
the thermal energy9,10, kBT (in which kB is the Boltzmann constant and tum yields of stabilized states of the protein that could initiate magnetic
T is temperature). [FAD•− TrpH•+] radical pairs in cryptochromes are signalling, most probably through a change in the conformation of
formed by the sequential hopping of an electron along a chain of three the C-terminal tail24. Here we present a detailed in vitro analysis of the
or four tryptophans to the photoexcited, non-covalently bound flavin photochemistry and magnetic sensitivity of ErCRY4, a protein that
adenine dinucleotide (FAD) chromophore11–15. Magnetically sensitive, is expressed in double-cone and long-wavelength single-cone pho-
light-induced radical pairs have been reported in vitro for a few members toreceptor cells in the eyes of night-migratory European robins25. In
of the cryptochrome/photolyase family of proteins: AtCRY1 (Arabi- addition to the light-induced radical pairs studied here, it has been
dopsis thaliana), DmCry (Drosophila melanogaster) and Escherichia suggested that radical pairs formed by ‘dark’ back-reactions could also
coli photolyase16,17. Indeed, a generic sensitivity of cryptochromes to be responsible for the magnetic sensitivity of cryptochromes in vivo
magnetic fields would be required to provide evolution with scope for and that CRY1a could be more suitable as a magnetoreceptor than
species-specific optimization in animals in which an awareness of the CRY421,26–30 (but see ref. 31). However, the identity of any ‘dark’ radical
Earth’s magnetic field would be advantageous for survival and fitness. pairs is unknown, and vertebrate CRY1a proteins seem not to bind FAD
The detailed properties of cryptochromes from non-migratory spe- strongly in vitro29,32. We have therefore focused our investigation on
cies such as plants and fruit flies18–23 might therefore not be relevant whether light-induced electron-transfer reactions in ErCRY4 in vitro
when considering magnetoreception in long-distance, night-migratory are fit for purpose for magnetic sensing.

1
AG Neurosensory Sciences/Animal Navigation, Institut für Biologie und Umweltwissenschaften, Carl-von-Ossietzky Universität Oldenburg, Oldenburg, Germany. 2Physical and Theoretical
Chemistry Laboratory, Department of Chemistry, University of Oxford, Oxford, UK. 3Inorganic Chemistry Laboratory, Department of Chemistry, University of Oxford, Oxford, UK. 4Institut für
Physikalische Chemie, Albert-Ludwigs-Universität Freiburg, Freiburg, Germany. 5Department of Medicinal Chemistry and Molecular Pharmacology, Purdue University, West Lafayette, IN, USA.
6
Division of Biochemistry, Department of Neuroscience, Carl-von-Ossietzky-Universität Oldenburg, Oldenburg, Germany. 7Department of Neuroscience, University of Texas Southwestern
Medical Center, Dallas, TX, USA. 8Howard Hughes Medical Institute, University of Texas Southwestern Medical Center, Dallas, TX, USA. 9Research Centre for Neurosensory Science, University of
Oldenburg, Oldenburg, Germany. 10Department of Physics, Carl-von-Ossietzky Universität Oldenburg, Oldenburg, Germany. 11High Magnetic Field Laboratory, Key Laboratory of High Magnetic
Field and Ion Beam Physical Biology, Hefei Institutes of Physical Science, Chinese Academy of Sciences, Hefei, PR China. 12International Magnetobiology Frontier Research Center, Hefei, PR
China. 13Centre for Advanced Electron Spin Resonance (CAESR), Department of Chemistry, University of Oxford, Oxford, UK. ✉e-mail: [email protected]; [email protected];
[email protected]; [email protected]; [email protected]; [email protected]

Nature | Vol 594 | 24 June 2021 | 535

Article
a b ErCRY4 ClCRY4 c
ErCRY4 kDa
WT WAF WBF WC F WDF WT TrpB
2
400 80 180 3

Conductivity (mS cm–1)

130 FAD
Absorbance (mAu)

100 W372
300 60 70 W395 TrpC
55
200 40 40 W318
TrpD
35 TrpA
100 20 25 1
4
0 0 W369
15 Tyr319
0 20 40 60 80 100 120 10
Volume (ml)

d FADox FADH
•
e 1.0 f
0.08
1.0 0.35 WDF 435–455 nm

Absorbance (580–600 nm)

Absorbance (435–455 nm)
ErCRY4 WT 0s 0.8 ErCRY4 ErCRY4 0.06
Relative absorbance

Relative absorbance
0.8 WT 435–455 nm
42 s
102 s
0.6 0.6 WT 450 nm WDF 450 nm 0.30
162 s
0.04
222 s
0.4 282 s 0.4 WT 580–600 nm
WT 600 nm
402 s
0.02
0.2 582 s 0.2 0.25
867 s WDF 600 nm
WDF 580–600 nm
0 0 0
300 400 500 600 700 800 0 200 400 600 800 0 10 20 30 40 50
Wavelength (nm) Time (s) Time (min)

Fig. 1 | Purification, electron transfer pathway and photoreduction of characteristic of correctly bound FADox. e, Time dependence of FADox
European robin CRY4. a, Purification of wild-type ErCRY4 by anion-exchange depletion (450 nm) and FADH• formation (600 nm). The faster photoreduction
chromatography. The protein fractions in the first peak of the gradient elution kinetics seen for the wild-type protein are consistent with a greater
contained yellow ErCRY4 with FAD bound. b, Verification of protein purity by stabilization of photoinduced radicals than in the WDF mutant. The decrease in
SDS–PAGE Coomassie blue staining: wild-type (WT) and TrpX-to-Phe (X = A–D) radical absorbance (600 nm) for the wild-type after 500 s illumination arises
mutants of ErCRY4, and wild-type ClCRY4. c, Structural homology model of from photoreduction of FADH• to FADH−. f, Time-dependence of the oxidation
ErCRY4 showing the flavin group of the FAD chromophore, the Trp-tetrad and of FADH• (580–600 nm) and the recovery of FADox (435–455 nm) under aerobic
Tyr 319. Sequential electron transfers are indicated by arrows. Robin conditions after a 2-min period of illumination at 450 nm. No oxidizing or
photograph by Thomas Eisenhut (© www.xeta.at). d, Photoreduction of reducing agents were added to the samples. Note the different vertical
wild-type ErCRY4 by continuous irradiation at 450 nm. UV–visible absorption scales for the two wavelength bands. See Supplementary Figs. 15–17 and
spectra show depletion of FADox (<500 nm) and accumulation of FADH• Supplementary Table 5 for further details. See Supplementary Methods for
(550–650 nm). The shoulders at around 480 nm on the peak at 450 nm are experimental details.

To characterize its photochemistry and magnetic properties, we 200 ps, around 1 ns, greater than 1 μs and greater than 1 μs, respectively)
recombinantly expressed and purified33 wild-type ErCRY4 as a yellow as the electron transfer chain is extended and as charge recombina-
protein containing the fully oxidized chromophore, FADox, which is tion (that is, the return to the ground state: [FAD•− TrpXH•+] → FADox +
essential for radical pair formation4 (Fig. 1a, b). Optical spectroscopy TrpXH) becomes less efficient (Supplementary Fig. 2). No radicals were
(Fig. 1d) and mass spectrometry (Supplementary Fig. 13) experiments observed for WAF, which lacks the tryptophan nearest the FAD; this
confirmed the incorporation of FADox (greater than 97%). ErCRY4 con- implies the absence of an alternative electron donor that is capable of
tains a chain of four tryptophan residues stretching from the FAD to the reducing the FAD and forming radical pairs with lifetimes longer than
surface of the protein25 (Fig. 1c). To determine the roles of these residues around 100 ps. Unlike WBF and WCF, the radical pairs formed in both
in radical-pair production, we also prepared site-specific mutants of WDF and in wild-type ErCRY4 are sufficiently long-lived that they could
ErCRY4 by replacing each of the tryptophans with a redox-inactive be sensitive to the geomagnetic field.
phenylalanine: W395F, W372F, W318F and W369F (Fig. 1b). We refer Continuous blue-light illumination of both wild-type ErCRY4 and
to these four residues as TrpX (X = A, B, C or D, respectively) and to the ErCRY4 WDF resulted in the reduction of FADox to the neutral semiqui-
corresponding mutants as WXF. RPX denotes the radical pair, [FAD•− none radical, FADH• (Fig. 1d); this confirms the existence of an intra-
TrpXH•+], containing the protonated cation radical form of TrpX. All the molecular electron transfer pathway and the stabilization of FAD•− by
mutated ErCRY4 forms bind FADox (Supplementary Fig. 1). protonation to form FADH• in both proteins on a timescale (greater
We have developed and continuously refined several spectroscopic than 10 μs) that is beyond the range of the transient absorption meas-
techniques specifically to study magnetic field effects on cryptochrome urements. Wild-type ErCRY4 is more easily photoreduced than is WDF
photochemistry. Cavity ring-down spectroscopy (CRDS)34 and broad- (Fig. 1e).
band cavity-enhanced absorption spectroscopy35 both take advan- Figure 2a shows the effect of an applied magnetic field on the absorb-
tage of the huge optical path lengths that are achievable by multiple ance of FAD and Trp radicals in wild-type ErCRY4, measured by CRDS
passes of the probe light through a sample placed between highly 2 μs after photoexcitation (450 nm) at 5 °C. Millitesla fields suppress
reflective mirrors. Higher time-resolution was provided by a conven- radical yields by favouring return to the ground state, as expected for
tional single-pass transient absorption spectrometer equipped with radical pairs that are formed in a spin-correlated singlet state17. Under
magnetic field coils. physiological conditions (40–43 °C), the pH of most bird cells36,37 is 7.3.
Transient absorption measurements were used to investigate the The change in the charge states of proteins caused by a reduction in
formation and stabilization of light-induced radicals. Laser excita- temperature—principally due to a shift in the acid–base equilibrium
tion (at 450 nm) of WBF, WCF, WDF and wild-type ErCRY4 reduced FADox of histidine side chains—can be reversed by an increase in pH (Sup-
ground-state absorption and produced signals that are characteristic plementary Information), which suggests that a pH of approximately
of FAD•− and TrpH•+ radicals with lifetimes that steadily increase (around 8 at 5 °C should be equivalent to physiological conditions (pH 7.3 at

536 | Nature | Vol 594 | 24 June 2021

 a Magnetic field (mT) b Magnetic field (mT) above the noise level for either ClCRY4 or GgCRY4. Taking the results
B1/2
10 20 30 10 20 30 of the two experiments together, the magnetic sensitivity of CRY4
0 0
from the night-migratory robin is substantially larger than that of the
Magnetic field effect (%)

Magnetic field effect (%)

–2
–5
B1/2 CRY4 proteins from the non-migratory—primarily diurnal—pigeon
WT ErCRY4 pH 7
–4 AtCRY1 WT pH 7 and chicken.
–10 B1/2 For a molecule to act as a receptor, it must switch on in response to a
–6
stimulus and either switch off again by returning to its resting state or
–8 –15 ErCRY4 WDF pH 7
be replaced by freshly expressed protein. For ErCRY4, it would there-
–10 WT ErCRY4 pH 8
–20
fore be advantageous if the FADH• and/or FADH− states were to be
oxidized back to the FADox state in minutes (rather than hours) in vivo.
c Wavelength (nm) d Time (μs) Illumination of ErCRY4—in a cuvette in contact with air and without
1
500 550 600 650 700 750 2 4 6 8 10 added oxidants, and at an intensity approximately equivalent to that
0 0
of sunlight—produced a steady state in which the proportions of the
–1
–2 ClCRY4 ErCRY4 oxidation states were approximately 50% FADox and approxi-
105 ΔΔA

ClCRY4 GgCRY4
–2 mately 50% FADH•. When the light was turned off, the recycling times
106 ΔΔA

–4 ErCRY4
GgCRY4
ErCRY4 (10.4 s) –3 from FADH• to FADox and from FADH− to FADH• were around 13 min
–6
ErCRY4 (0.78 s)
–4 and around 1 min, respectively (Fig. 1f, Supplementary Information,
–8
pH 8 pH 8 Supplementary Figs. 15–17, Supplementary Table 5). The re-oxidation
–5
–10 • •
times for ClCRY4 and GgCRY4 have similar values (Supplementary
Trp FADH
Information, Supplementary Table 5). Re-oxidation of ErCRY4 in the
Fig. 2 | Magnetic field effects on the yields of photoinduced radicals in presence of 5 mM dithiothreitol is slower: about 5 h at 25 °C and about 1 h
CRY4 proteins. a, b, Magnetic field effect (percentage change in absorbance at the physiological temperature of 40 °C (Supplementary Information,
induced by the magnetic field) on the optical absorbance of photoinduced Supplementary Fig. 13, Supplementary Table 6). Although re-oxidation
radicals 2 μs after a 450-nm laser pulse in wild-type ErCRY4 at pH 7 and pH 8 (a) could be slower or faster in vivo, these measurements show that there
and in WDF and in wild-type AtCRY1 at pH 7 (b). Data were measured by CRDS at are conditions under which ErCRY4 could cycle fast enough to act as
530 nm. The smaller value of B1/2 (the magnetic field that produces 50% of the
a magnetoreceptor.
limiting change seen at high field) for wild-type ErCRY4 (4.9 ± 1.6 mT) compared
The slower photoreduction of WDF compared to the wild-type protein
to wild-type AtCRY1 (14.3 ± 1.7 mT) suggests that, other factors being equal, the
(Fig. 1e) suggests that TrpD participates in the formation of radical pairs
former would be more sensitive to weak magnetic fields. c, Change in the
optical absorbance of photoinduced radicals in three avian CRY4 samples
in wild-type ErCRY4 in vitro (as it does in DmCry13,14) by prolonging the
induced by a 30-mT magnetic field. All four broadband cavity-enhanced lifetime of photoreduced states of the protein. To determine whether
absorption spectra measured using continuous illumination at 450 nm are the electron-transfer chain in wild-type ErCRY4 stops at TrpC or extends
dominated by the field-induced reduction in the yield of Trp• radicals, which to include TrpD, we performed time-resolved electron paramagnetic
absorb in the 500−600 nm range. ErCRY4 spectra are shown at two different resonance and out-of-phase electron spin echo envelope modula-
times after the start of illumination. The equivalent spectra of ClCRY4 and tion experiments on wild-type ErCRY4 and its WDF mutant. Both the
GgCRY4 showed no time dependence and are averages over the first 9 s. The spin polarization of the time-resolved electron paramagnetic reso-
weak signals at wavelengths greater than 600 nm are thought to arise from nance spectra38 (Fig. 3a) and the existence of an ‘out of phase’ (that is,
magnetic field effects on the formation of FADH•. d, Change in the 90°-phase-shifted) spin echo13 (Fig. 3b) confirm the rapid (within 100 ns)
photoinduced optical absorbance of three avian CRY4 samples upon formation of singlet radical pairs possessing the non-equilibrium
application of a 30 mT magnetic field, measured by CRDS at 530 nm. Within spin-polarization that is a necessary—but not sufficient—condition
error, the magnetic field effects on the CRY4 proteins from the non-migratory
for magnetic sensitivity. In the latter measurement, the modulation
pigeon and chicken could not be distinguished from zero (apart from one very
frequencies arise mainly from the electron dipolar coupling, which
early data point for ClCRY4). See Supplementary Methods for experimental
depends on the centre-to-centre radical separation, r, as r−3. The clearly
details and statistical information. Data in a, b, d are mean ± s.e.m.
different modulation frequencies in Fig. 3b are consistent with r being
approximately 0.3-nm larger for the wild-type protein than for the
mutant (Supplementary Fig. 4). Molecular dynamics simulations pre-
40–43 °C)36,37. It may be relevant, therefore, that the magnetic field dict r = 2.2 nm for RPD and 1.8 nm for RPC, and the X-ray structure of
effect on wild-type ErCRY4 at 5 °C is substantially larger at pH 8 than ClCRY4 reveals FAD–TrpD and FAD–TrpC distances of 2.13 nm and 1.76 nm,
at pH 7 (Fig. 2a). respectively15 (Supplementary Figs. 4, 8, Supplementary Table 2),
If the magnetic field effect on CRY4 forms the basis of the magnetic both of which are consistent with the involvement of TrpD in electron
compass sense in night-migratory robins, it should have been optimized transport in wild-type ErCRY4. Very similar results have recently been
by evolution. We therefore compared the magnetic field effects on reported39 for pigeon CRY4. The electron paramagnetic resonance
ErCRY4 with those on CRY4 proteins from two non-migratory birds, spectra of the two proteins bear a strong resemblance to those of the
pigeon33 (C. livia) and chicken (G. gallus), in which we expect there to FAD–Trp radical pairs in DmCry14, in which the differences in the spec-
have been less evolutionary pressure to optimize any light-dependent tral shapes were related to the larger dipolar coupling in RPC and the
magnetic compass sense. Figure 2c shows the change in the optical different side-chain orientations of TrpC and TrpD radicals. In summary,
absorption spectra of the three proteins caused by a 30-mT magnetic electron paramagnetic resonance spectroscopy provides strong
field under conditions of continuous illumination at 450 nm. The band evidence that photoinduced electron transfer in wild-type ErCRY4
between 500 nm and 550 nm suggests a considerably bigger magnetic results in the formation of [FAD•− TrpDH•+] in a singlet state.
field effect on the yield of neutral Trp• radicals in ErCRY4 than in ClCRY4 Having established that the electron transfer pathway in wild-type
or GgCRY4. This difference between the three proteins was supported ErCRY4 extends as far as TrpD, it is important to know the role of RPD
by CRDS (Fig. 2d) by measuring the effect of a 30-mT magnetic field in generating the magnetic responses of the protein. Figure 2b shows
on the radical absorption signal as a function of time after irradiation the magnetic field effects on radical yields in two proteins, ErCRY4 WDF
with blue light. The field produces a significant reduction in the yield of and AtCRY1, that have a Trp triad and therefore only RPC. Both are more
radicals in ErCRY4 that persists for more than 10 μs. Apart from one of sensitive to magnetic fields than is wild-type ErCRY4 (Fig. 2a) at pH 7.
the earliest data points for ClCRY4, no such change could be observed This would seem to be consistent with the 0.3-nm-larger separation

Nature | Vol 594 | 24 June 2021 | 537

Article
a b c
ErCRY4 WT 600–650 nm

Out-of-phase echo intensity

A 0 2

104 ΔA
0 0.2 0.4 0.6 0.8 1.0
Time (μs)
ErCRY4 WDF –2
E
ErCRY4 WT 0
ErCRY4 WDF –4 ErCRY4 WT
460–490 nm ErCRY4 WDF
344 346 348 350 352 0 0.5 1.0 1.5 2.0 –6
Magnetic field (mT) W (μs)

Fig. 3 | Electron paramagnetic resonance and optical spectroscopy of τ  is the interval between the microwave pulses in the spin echo pulse sequence.
photoinduced FAD–Trp radical pairs in ErCRY4. a, Time-resolved X-band Slight irregularities in the oscillations could arise from nuclear electron spin
(9.75 GHz) electron paramagnetic resonance spectra of wild-type ErCRY4 and echo envelope modulations. c, Time dependence of the decay of the transient
its WDF mutant recorded at 1 °C, 0.5 μs after a 450-nm laser pulse. The spectra radical absorption at 600–650 nm and the recovery of the ground state
have been normalized to the same absolute intensity. A, absorptive absorption at 460–490 nm for the wild-type and WDF forms of ErCRY4.
contribution; E, emissive contribution. b, Out-of-phase electron spin echo The lines are guides for the eye. See also Supplementary Fig. 12. See
envelope modulation signals recorded at Q-band frequencies (34 GHz) at 80 K. Supplementary Methods section for experimental details.

between the radicals in RPD, which should make radical recombination combinations of ⟨kr⟩ and ⟨kf ⟩ for which the radical pair lifetime is
much slower than for RPC12. However, we found that the light-induced approximately 100 ns (white line) and the points on that line (white
radicals formed in the two proteins have very similar decay kinetics spots) that correspond to pure RPC (fC = 1.0), and a composite radical
during the period (up to around 1 μs) in which the magnetic field effects pair with fC = 0.1. The values of the individual rate constants that under-
are generated. As shown in Fig. 3c, in both proteins a substantial frac- lie these average rate constants are given in the Supplementary Infor-
tion of the radicals disappears and the ground state recovers with time mation. The magnetic sensitivity for fC = 1.0 (modelling WDF) is
constants on the order of 100 ns. considerably larger than that for fC = 0.1 (modelling the wild-type
To gain further insight, we performed molecular dynamics simu- protein) because ⟨kr⟩ and ⟨kf ⟩ for the latter are far from the 1:3 ratio that
lations for five redox states of ErCRY4—the diamagnetic ground is required for a strong magnetic response.
state and the four radical-pair states, RPX (X = A, B, C or D)—to obtain In vivo, a cryptochrome magnetoreceptor would need to be capable
order-of-magnitude estimates of the rate constants for electron trans- of sensing and signalling. The sensing state cannot survive for more
fer along the tryptophan chain (Supplementary Table 3, Supplementary than a few microseconds, otherwise any magnetic field effect on the
Figs. 9, 10). The first two steps, RPA → RPB and RPB → RPC, are exergonic, singlet–triplet interconversion would be lost by spin relaxation40. How-
fast (>1010 s−1) and outpace the reverse reactions (RPA ← RPB and RPB ← ever, to trigger a biochemical signalling cascade, a lifetime of millisec-
RPC) by factors of more than 104. However, the third electron transfer onds, seconds or longer is required. No single state of the protein could
(RPC ↔ RPD), is predicted to have similar forward (kCD) and backward fulfil both roles. Evolutionary optimization of these functions might
(kDC) rate constants of around 1010 s−1 (Fig. 4a). The molecular dynamics be favoured by a composite ErCRY4 radical pair formed rapidly in high
simulations also provided estimates of the rate constants for the recom- yield, with inter-radical spin interactions that are not so large that they
bination of RPC and RPD directly to the ground state, confirming that the block singlet–triplet mixing41 and, as shown by the WDF mutant, an RPC
former should be one to two orders of magnitude faster than the latter component that is capable of generating strong magnetic field effects.
(Fig. 4a, Supplementary Table 3). These rate constants, and those of the The RPD component, with its larger radical separation, is better placed
other relevant processes (singlet–triplet interconversion and radical to generate the signalling state. Being closer to the surface of the pro-
(de)protonation), are so much slower than 1010 s−1 that the RPC and RPD tein, TrpDH•+ could deprotonate more efficiently than could TrpCH•+.
states of ErCRY4 could be in dynamic equilibrium. The consequence Alternatively, TrpDH•+ could be stabilized by receipt of an electron from
would be a ‘composite’ radical pair, the spin dynamics and reaction an external reductant or, possibly, from Tyr31911 (which is only 0.37 nm
rates of which are weighted averages of the properties of RPC and RPD, away15,25) (Fig. 1c). On a similar (greater than 1 μs) timescale, FAD•− would
with the weights given by the fractional populations of the two states, become protonated (Figs. 1d, 2c), which would produce the FADH• form
fC and fD, where fC/fD = kDC/kCD. A reaction scheme including this pos- of the protein as a potential signalling state (Fig. 4a).
sibility, and showing various estimates of the relevant rate constants, To be sensitive to the Earth’s magnetic field (around 50 μT) in vivo,
is provided in Fig. 4a. As detailed in the Supplementary Information, a composite radical pair would need a minimum lifetime4 of around
the difference in the dipolar couplings measured for wild-type ErCRY4 1 μs—that is, an order of magnitude longer than the approximately
and for WDF would not be inconsistent with a composite radical pair, 100-ns lifetime that is observed here for purified wild-type ErCRY4
provided that the proportion of RPC at equilibrium is no larger than in vitro. Figure 4c shows a simulation performed under identical con-
about 10% (that is, fC ≤ 0.1). ditions to those in Fig. 4b except that the magnetic field was 50 μT
To explore whether a rapid equilibrium between the RPC and RPD instead of 30 mT. The maximum magnetic field effect (which occurs
states of wild-type ErCRY4 could account for both its smaller magnetic when ⟨kf ⟩ ≈ 8 × 105 s−1, ⟨kr⟩ ≈ 2 × 106 s−1) is smaller than at 30 mT, has
field effect (Fig. 3a) and the kinetic similarity to WDF (Fig. 3c), we used opposite phase (the ‘low field effect’42), and is shifted to lower values
spin dynamics simulations to calculate the magnetic field effect of ⟨kr⟩ and ⟨kf ⟩, reflecting the smaller electron Larmor frequency in the
(Fig. 4b) as a function of the weighted average reaction rate constants, weaker magnetic field. Superimposed on the contour plot is the locus
⟨kr⟩ (return of the singlet pairs to the ground state) and ⟨kf ⟩ (the com- of rate constants for which the radical pair lifetime is around 1 μs. As
peting forward reaction). Strong magnetic sensitivity requires that these expected, it passes close to the maximum magnetic field effect. Figure
reactions proceed at similar rates, are not too slow to compete with 4c also shows the predicted rate constants for RPC and a composite
spin relaxation, and are not so fast that there is little time for singlet– radical pair—as in Fig. 4b, but with the values of ⟨kf ⟩ scaled down by a
triplet interconversion. For ⟨kf ⟩ in the range 105−108 s−1, the maximum factor of 10. Such a change in the average rate constant of the forward
magnetic field effect occurs when ⟨kr⟩ ≈ 3⟨kf ⟩ . Figure 4b shows the reaction would optimize the predicted magnetic field effect for the

538 | Nature | Vol 594 | 24 June 2021

 a b

Change in yield of kf reaction channel

9
–0.01
–0.02
FAD* 10 12 8
Composite –0.03
RPA

log10(〈kr〉 (s–1))
1011 radical pair –0.04
RPB 7 RPC
kCD = 1010 –0.05
107
RPC RPD –0.06
106 kDC = 1010 6 {RPC/RPD} –0.07
S T –0.08
1010 kCr = 107 5 –0.09
〈kr〉 〈kf〉
Blue light 109 〈kf〉 –0.10
kCf = 106 kDf = 106 4
4 5 6 7 8 9
Stabilized log10(〈kf〉 (s–1))
RP2C radical RP2D
c

Change in yield of kf reaction channel

pairs 9
0.0050

kDr = 105 8 0.0040

FADH
• RPC

log10(〈kr〉 (s–1))
Potential
7 0.0030
signalling state

6 {RPC/RPD} 0.0020
FADox
0.0010
5
0
4
4 5 6 7 8 9
log10(〈kf〉 (s–1))

Fig. 4 | Reaction scheme and simulated magnetic field effects for ErCRY4. ⟨kr ⟩ are weighted averages of the corresponding rate constants for RPC and RPD.
a, Proposed reaction scheme for magnetic sensing and signalling by CRY4 in The contour plot shows the change in the yield of the ⟨kf ⟩ pathway caused by a
night-migratory songbirds. The arrows are labelled with approximate rate 30-mT magnetic field as a function of ⟨kf ⟩ and ⟨kr ⟩. On the white line the lifetime
constants (in s−1): red, calculations based on molecular dynamics simulations; of the composite radical pair is approximately 100 ns. The positions of the
purple, transient absorption measurements; green, estimated from magnetic white spots correspond to the rate constants estimated for RPC and the
field effects. b, Spin dynamics simulation for the reaction scheme shown composite pair with fC = 0.1. See Supplementary Information for details.
at the bottom left, in which the curved arrows represent the coherent c, As b, except that the magnetic field is 50 μT and the white line corresponds to
interconversion of the singlet (S) and triplet (T) states of a composite radical a lifetime of the composite radical pair of approximately 1 μs. The white spots
pair and the straight arrows indicate the competing reaction pathways. ⟨kf ⟩ and have been moved −1.0 log units along the ⟨kf ⟩ axis.

composite radical pair and cause it to outperform RPC in a 50-μT field. would be required. We hope that such experiments will become
A reduction in ⟨kf ⟩ of this magnitude could be achieved in vivo by a possible in the future.
combination of smaller changes in the individual forward rate constants
for RPC and RPD (kCf and kDf ) and the equilibrium proportions of the two Online content
radical pairs (fC). Small changes in the two recombination rates (kCr and Any methods, additional references, Nature Research reporting sum-
kDr) might also be advantageous. Such variations could arise from pro- maries, source data, extended data, supplementary information,
tein–protein interactions that reduced the solvent accessibility of the acknowledgements, peer review information; details of author con-
two TrpH•+ radicals (and slowed their deprotonation) and produced tributions and competing interests; and statements of data and code
small shifts in the position of the RPC/RPD equilibrium. It is anticipated availability are available at https://doi.org/10.1038/s41586-021-03618-9.
that a viable cryptochrome magnetoreceptor would require protein
1. Mouritsen, H. Long-distance navigation and magnetoreception in migratory animals.
interaction partners in vivo both for signalling and to restrict its ori- Nature 558, 50–59 (2018).
entational freedom (as some degree of immobilization is required for 2. Wiltschko, W., Munro, U., Ford, H. & Wiltschko, R. Red-light disrupts magnetic orientation
directional sensing43). Protein–protein interactions could also enhance of migratory birds. Nature 364, 525–527 (1993).
3. Zapka, M. et al. Visual but not trigeminal mediation of magnetic compass information in a
the sensitivity to magnetic fields of around 50 μT in vivo by restricting migratory bird. Nature 461, 1274–1277 (2009).
the internal mobility of the radicals and so reducing the effect of spin 4. Hore, P. J. & Mouritsen, H. The radical-pair mechanism of magnetoreception. Annu. Rev.
relaxation40,44. CRY4 molecules are not expected to show substantial Biophys. 45, 299–344 (2016).
5. Ritz, T., Adem, S. & Schulten, K. A model for photoreceptor-based magnetoreception in
responses to Earth-strength magnetic fields in vitro unless they are birds. Biophys. J. 78, 707–718 (2000).
anchored, aligned and associated with the appropriate signalling part- 6. Ritz, T., Thalau, P., Phillips, J. B., Wiltschko, R. & Wiltschko, W. Resonance effects
ners. None of the proteins required for these interactions is currently indicate a radical-pair mechanism for avian magnetic compass. Nature 429, 177–180
(2004).
known (but see ref. 45). Furthermore, magnetic sensitivity could be 7. Engels, S. et al. Anthropogenic electromagnetic noise disrupts magnetic compass
enhanced by receptor alignment41, biochemical amplification45 and orientation in a migratory bird. Nature 509, 353–356 (2014).
neural processing3. These aspects can also be optimized by evolution 8. Steiner, U. E. & Ulrich, T. Magnetic field effects in chemical kinetics and related
phenomena. Chem. Rev. 89, 51–147 (1989).
and could substantially improve the magnetic sensitivity of the robin’s 9. Maeda, K. et al. Chemical compass model of avian magnetoreception. Nature 453,
magnetic sense. 387–390 (2008).
In conclusion, we have demonstrated that CRY4 from the night- 10. Kerpal, C. et al. Chemical compass behaviour at microtesla magnetic fields
strengthens the radical pair hypothesis of avian magnetoreception. Nat. Commun. 10,
migratory European robin seems to be fit for purpose as a magnetic 3707 (2019).
sensor, and that it is more magnetically sensitive than CRY4 from the 11. Giovani, B., Byrdin, M., Ahmad, M. & Brettel, K. Light-induced electron transfer in a
non-migratory pigeon and chicken under identical measurement con- cryptochrome blue-light photoreceptor. Nat. Struct. Biol. 10, 489–490 (2003).
12. Müller, P., Yamamoto, J., Martin, R., Iwai, S. & Brettel, K. Discovery and functional
ditions in vitro. Our data also suggest that nature might have found analysis of a 4th electron-transferring tryptophan conserved exclusively in
ways to independently optimize the magnetic sensing and signalling animal cryptochromes and (6-4) photolyases. Chem. Commun. 51, 15502–15505
properties of CRY4 through control of its photochemistry. To deter- (2015).
13. Nohr, D. et al. Determination of radical–radical distances in light-active proteins and their
mine whether CRY4 acts as a magnetoreceptor molecule in vivo, direct implication for biological magnetoreception. Angew. Chem. Int. Ed. 56, 8550–8554
manipulations of this protein in the eyes of night-migratory songbirds (2017).

Nature | Vol 594 | 24 June 2021 | 539

Article
14. Nohr, D. et al. Extended electron-transfer pathways in animal cryptochromes mediated by 30. Müller, P. & Ahmad, M. Light-activated cryptochrome reacts with molecular oxygen to
a tetrad of aromatic amino acids. Biophys. J. 111, 301–311 (2016). form a flavin–superoxide radical pair consistent with magnetoreception. J. Biol. Chem.
15. Zoltowski, B. D. et al. Chemical and structural analysis of a photoactive vertebrate 286, 21033–21040 (2011).
cryptochrome from pigeon. Proc. Natl Acad. Sci. USA 116, 19449–19457 (2019). 31. Bolte, P. et al. Cryptochrome 1a localisation in light- and dark-adapted retinae of several
16. Sheppard, D. M. W. et al. Millitesla magnetic field effects on the photocycle of an animal migratory and non-migratory bird species: no signs of light-dependent activation. Ethol.
cryptochrome. Sci. Rep. 7, 42228 (2017). Ecol. Evol. https://doi.org/10.1080/03949370.2020.1870571 (2021).
17. Maeda, K. et al. Magnetically sensitive light-induced reactions in cryptochrome are 32. Kutta, R. J., Archipowa, N., Johannissen, L. O., Jones, A. R. & Scrutton, N. S. Vertebrate
consistent with its proposed role as a magnetoreceptor. Proc. Natl Acad. Sci. USA 109, cryptochromes are vestigial flavoproteins. Sci. Rep. 7, 44906 (2017).
4774–4779 (2012). 33. Qin, S. et al. A magnetic protein biocompass. Nat. Mater. 15, 217–226 (2016).
18. Fedele, G., Green, E. W., Rosato, E. & Kyriacou, C. P. An electromagnetic field disrupts 34. Maeda, K. et al. Following radical pair reactions in solution: a step change in sensitivity
negative geotaxis in Drosophila via a CRY-dependent pathway. Nat. Commun. 5, 4391 using cavity ring-down detection. J. Am. Chem. Soc. 133, 17807–17815 (2011).
(2014). 35. Neil, S. R. T. et al. Broadband cavity-enhanced detection of magnetic field effects in chemical
19. Gegear, R. J., Casselman, A., Waddell, S. & Reppert, S. M. Cryptochrome mediates models of a cryptochrome magnetoreceptor. J. Phys. Chem. B 118, 4177–4184 (2014).
light-dependent magnetosensitivity in Drosophila. Nature 454, 1014–1018 (2008). 36. Roos, A. & Boron, W. F. Intracellular pH. Physiol. Rev. 61, 296–434 (1981).
20. Hoang, N. et al. Human and Drosophila cryptochromes are light activated by flavin 37. Reeves, R. B. The interaction of body temperature and acid–base balance in ectothermic
photoreduction in living cells. PLoS Biol. 6, e160 (2008). vertebrates. Annu. Rev. Physiol. 39, 559–586 (1977).
21. Pooam, M. et al. Magnetic sensitivity mediated by the Arabidopsis blue-light receptor 38. Weber, S. et al. Origin of light-induced spin-correlated radical pairs in cryptochrome.
cryptochrome occurs during flavin reoxidation in the dark. Planta 249, 319–332 J. Phys. Chem. B 114, 14745–14754 (2010).
(2019). 39. Hochstoeger, T. et al. The biophysical, molecular, and anatomical landscape of
22. Xu, C. X., Lv, Y., Chen, C. F., Zhang, Y. X. & Wei, S. F. Blue light-dependent pigeon CRY4: a candidate light-based quantal magnetosensor. Sci. Adv. 6, eabb9110
phosphorylations of cryptochromes are affected by magnetic fields in Arabidopsis. (2020).
Adv. Space Res. 53, 1118–1124 (2014). 40. Kattnig, D. R., Solov’yov, I. A. & Hore, P. J. Electron spin relaxation in cryptochrome-based
23. Giachello, C. N. G., Scrutton, N. S., Jones, A. R. & Baines, R. A. Magnetic fields modulate magnetoreception. Phys. Chem. Chem. Phys. 18, 12443–12456 (2016).
blue-light-dependent regulation of neuronal firing by cryptochrome. J. Neurosci. 36, 41. Efimova, O. & Hore, P. J. Role of exchange and dipolar interactions in the radical pair
10742–10749 (2016). model of the avian magnetic compass. Biophys. J. 94, 1565–1574 (2008).
24. Watari, R. et al. Light-dependent structural change of chicken retinal cryptochrome4. 42. Timmel, C. R., Till, U., Brocklehurst, B., McLauchlan, K. A. & Hore, P. J. Effects of weak
J. Biol. Chem. 287, 42634–42641 (2012). magnetic fields on free radical recombination reactions. Mol. Phys. 95, 71–89 (1998).
25. Günther, A. et al. Double-cone localization and seasonal expression pattern suggest a 43. Worster, S., Mouritsen, H. & Hore, P. J. A light-dependent magnetoreception
role in magnetoreception for European robin cryptochrome 4. Curr. Biol. 28, 211–223.e4 mechanism insensitive to light intensity and polarization. J. R. Soc. Interface 14,
(2018). 20170405 (2017).
26. Nießner, C., Denzau, S., Peichl, L., Wiltschko, W. & Wiltschko, R. Magnetoreception in 44. Kattnig, D. R., Sowa, J. K., Solov’yov, I. A. & Hore, P. J. Electron spin relaxation can enhance
birds: I. Immunohistochemical studies concerning the cryptochrome cycle. J. Exp. Biol. the performance of a cryptochrome-based magnetic compass sensor. New J. Phys. 18,
217, 4221–4224 (2014). 063007 (2016).
27. Nießner, C., Denzau, S., Peichl, L., Wiltschko, W. & Wiltschko, R. Magnetoreception: 45. Wu, H., Scholten, A., Einwich, A., Mouritsen, H. & Koch, K. W. Protein–protein interaction
activation of avian cryptochrome 1a in various light conditions. J. Comp. Physiol. A 204, of the putative magnetoreceptor cryptochrome 4 expressed in the avian retina. Sci. Rep.
977–984 (2018). 10, 7364 (2020).
28. Wiltschko, R., Ahmad, M., Nießner, C., Gehring, D. & Wiltschko, W. Light-dependent
magnetoreception in birds: the crucial step occurs in the dark. J. R. Soc. Interface 13, Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in
20151010 (2016). published maps and institutional affiliations.
29. Player, T. C. & Hore, P. J. Viability of superoxide-containing radical pairs as
magnetoreceptors. J. Chem. Phys. 151, 225101 (2019). © The Author(s), under exclusive licence to Springer Nature Limited 2021

540 | Nature | Vol 594 | 24 June 2021

 S. Y. Wong for assistance with spin dynamics calculations, and W. Myers (CAESR, Engineering
and Physical Sciences Research Council, grant no. EP/L011972/1) for obtaining a threefold
Reporting summary
improvement in the time-resolved electron paramagnetic resonance signal. We are grateful to
Further information on research design is available in the Nature the staff of the mechanical and electronic workshops in the Oxford Chemistry Department and
Research Reporting Summary linked to this paper. at the University of Oldenburg.

Author contributions J.X., L.E.J., T.Z., M.K., K.B.H., S.R. and M.J.G. made particularly important
experimental contributions. J.X. cloned wild-type ErCRY4 and all the mutants and developed
Data availability the protocols for expression and purification of the proteins with FAD bound. J.X. and J.S.
produced the protein samples. L.E.J. developed the continuous illumination experiment for
The complete set of molecular dynamics simulation and quantum studying photoreduction and the picosecond transient absorption experiment for measuring
chemistry data (300 GB) can be downloaded from the University of magnetic field effects, and recorded and interpreted data. T.Z. and M.J.G. developed the CRDS
Oldenburg repository: https://cloud.uol.de/s/NrTYpoEzL6RbPq7. experiment for measuring magnetic field effects and recorded and interpreted data. M.K.
developed the broadband cavity-enhanced absorption spectroscopy experiment for
Specific molecular dynamics data can also be obtained directly from measuring magnetic field effects and recorded and interpreted data. S.R. and S.W. recorded
I.A.S. on request. Source data are provided with this paper. and interpreted the EPR data. K.B.H. participated in all five of the above experiments and
recorded and interpreted spectroscopic data. J.F., with K.B.H., recorded and interpreted some
of the transient absorption data and all of the re-oxidation data. M.K. helped with the global
Acknowledgements This work was supported by the Air Force Office of Scientific Research analysis of the re-oxidation data. M.J.G., V.D., J.R.W. and P.D.F.M. made spectroscopic
(Air Force Materiel Command, USAF award no. FA9550-14-1-0095, to P.J.H., H.M., C.R.T., S.R.M. measurements of magnetic field effects. D.J.C.S. helped to develop the picosecond TA
and K.-W.K.); by the European Research Council (under the European Union’s Horizon 2020 apparatus. J.L. and Y.W. performed spin dynamics calculations. T.L.P. and G.M. reproduced and
research and innovation programme, grant agreement no. 810002, Synergy Grant: helped to interpret the EPR data. A.S.G. recorded and interpreted mass spectra. M.B.
‘QuantumBirds’, awarded to P.J.H. and H.M.); by the Office of Naval Research Global, award no. expressed and purified chicken CRY4. M.H., S.H., G.D. and S.J.K. expressed and purified some
N62909-19-1-2045, to P.J.H., C.R.T. and S.R.M.; by the Deutsche Forschungsgemeinschaft of the ErCRY4 protein samples. Y.C., J.S.T. and J.X. expressed and purified pigeon CRY4. H.Y.,
(SFB 1372, ‘Magnetoreception and navigation in vertebrates’, project number: 395940726 to H.W., K.-W.K., R.B. and C.X. provided advice on protein expression. I.A.S. performed molecular
H.M., K.-W.K., I.A.S. and P.J.H., and GRK 1885, ‘Molecular basis of sensory biology’ to K.-W.K., dynamics simulations and provided advice on cryptochrome structure and dynamics. L.E.J.
I.A.S. and H.M.); by a DAAD (German Academic Exchange Service, Graduate School had oversight of the organization and administration of the optical spectroscopy
Scholarship Programme, ID 57395813) stipend to J.X.; by funding for G.M. from the SCG measurements. P.J.H., H.M., C.R.T. and S.R.M. conceived the study. P.J.H., H.M., C.R.T., S.R.M.
Innovation Fund; by the Electromagnetic Fields Biological Research Trust (to P.J.H., C.R.T. and and C.X. supervised the work. P.J.H. and H.M. wrote the manuscript, and all authors
S.R.M.); by the National Natural Science Foundation of China, grant no. 31640001, and the commented on the manuscript.
Presidential Foundation of Hefei Institutes of Physical Science, Chinese Academy of Sciences,
grant no. BJZX201901 (to C.X.); and by the Lundbeck Foundation, the Danish Councils for Competing interests The authors declare no competing interests.
Independent Research, and the Volkswagen Foundation (to I.A.S.). V.D. is grateful to the
Clarendon Fund, University of Oxford. M.J.G. thanks the Biotechnology and Biological Additional information
Sciences Research Council, grant number BB/M011224/1 and the Clarendon Fund. We Supplementary information The online version contains supplementary material available at
acknowledge use of the Advanced Research Computing (ARC) facility of the University of https://doi.org/10.1038/s41586-021-03618-9.
Oxford. J.S.T. is an Investigator and Y.C. is a Research Specialist in the Howard Hughes Medical Correspondence and requests for materials should be addressed to I.A.S., C.X., S.R.M., C.R.T.,
Institute. I.A.S. is grateful to the DeiC National HPC Center, University of Southern Denmark for H.M. or P.J.H.
computational resources. We thank B. Grünberg, I. Fomins, A. Günther and A. Einwich for Peer review information Nature thanks Aurelien de la Lande, Thorsten Ritz, Eric Warrant and
laboratory assistance and for providing protein sequence information. J.X. thanks Y. Tan for the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
training in protein expression and purification. We thank S. Chandler for mass spectrometry, Reprints and permissions information is available at http://www.nature.com/reprints.