Tables for
Volume F
Crystallography of biological macromolecules
Edited by E. Arnold, D. M. Himmel and M. G. Rossmann

International Tables for Crystallography (2012). Vol. F, ch. 11.7, pp. 311-316

Chapter 11.7. Detecting twinning by merohedry

T. O. Yeatesa* and Y. Tsaia

aDepartment of Chemistry and Biochemistry, University of California, Los Angeles, PO Box 951569, Los Angeles, CA 90095–1569, USA
Correspondence e-mail:

The chapter defines various kinds of twinning and focuses on the detection and analysis of twinning by hemihedry. The perfect overlap of lattices causes distinct intensities to be added together, thereby affecting the intensity distribution as well as the apparent symmetry properties of the crystal, to a degree determined by the twin fraction. Equations are presented for identifying the presence of twinning and estimating the twin fraction from an analysis of diffraction intensities. Complicating factors, including anisotropy, pseudo-centring and local noncrystallographic symmetry, are discussed.

11.7.1. Introduction

| top | pdf |

Twinning is a growth disorder in which different regions or domains of a crystal specimen are oriented differently with respect to each other (reviewed in Yeates, 1997[link]; Dauter, 2003[link]; Parsons, 2003[link]; Helliwell, 2008[link]). Numerous types of twinning are possible and a hierarchy among the types has been articulated (Yeates & Fam, 1999[link]) (Fig.[link]). In the most general type, differently oriented twin domains in the crystal have lattices that are distinguishable from each other. As a result, the different domains are associated with distinct reciprocal lattices and therefore with distinguishable interpenetrating diffraction patterns. This type of twinning is therefore generally recognized in X-ray diffraction images. The situation is referred to as twinning by non-merohedry and has also been referred to occasionally in the literature as `epitaxial twinning'. However, the latter term is usually used to describe the growth of a crystal of heterogeneous molecules. Treatment of twinning by non-merohedry usually amounts to a data-collection problem, as only a fraction of the diffraction spots in a given image correspond to a single lattice. A further complication can arise from the presence of a subset of reflections (e.g. those lying on an axis of rotation that relates the twin domains) whose multiple distinct contributions from the different domains overlap in the diffraction pattern. Strategies for unit-cell determination and reflection integration have been developed and implemented in various X-ray data processing programs (Sparks, 1999[link]; Sheldrick, 1997[link], 2002[link], 2004[link], 2008[link]; Duisenberg, 1992[link]; Herbst-Irmer & Sheldrick, 1998[link]; Duisenberg et al., 2003[link]; Dix et al., 2007[link]). Those programs have been developed mainly in the context of small-molecule applications, but are often suitable for problems involving macromolecular crystals as well. We do not consider twinning by non-merohedry further in this chapter, but focus instead on a special category (Fig.[link]) known as twinning by merohedry. The synonymous term `merohedral twinning' is also widely used in the macromolecular literature, although it has been noted that the latter usage introduces potential confusion in that the adjective `merohedral', or `mérohèdre' (Friedel, 1926[link]), is used in mathematical crystallography in a somewhat different context as a specific property of certain symmetries.


Figure | top | pdf |

Hierarchy of multiple types of twinning. Merohedry is elevated over non-merohedry as it results in the exact superposition of diffraction patterns. Tetartohedry is ranked higher than hemihedry because it involves a larger number of domain orientations. Perfect merohedry is given higher rank than partial merohedry as it results in higher apparent symmetry than exists in the crystal. Darker print highlights the scope of the text. (Modified from Yeates & Fam, 1999[link].)

11.7.2. Twinning by merohedry – considerations of lattice symmetry

| top | pdf |

In twinning by merohedry, the lattices of the differently oriented domains are superimposable, although the operations that relate the domains are not operations of the space-group symmetry (Fig.[link]). This is possible only in special symmetries, namely those where the rotational symmetry of the lattice exceeds the underlying rotational symmetry of the crystal space group; we consider only rotational operations here, in order to focus on situations relevant to biological macromolecules. Symmetries where this occurs are listed in Table[link]; they include crystal point groups 4, 3, 312, 321, 6 and 23. Space group P4 provides a simple example. It requires a tetragonal lattice, and that lattice contains twofold axes of rotational symmetry (perpendicular to the fourfold axis) which are not utilized by the P4 symmetry of the crystal. Therefore, two separate crystal domains, each obeying P4 symmetry, can grow together within the same specimen in a way such that they are rotated with respect to each other by one of these twofold symmetry axes of the tetragonal lattice. The operation used to describe the rotation is referred to as the twin operation. Because the lattices of the two domains are indistinguishable, the separate domains give rise to diffraction patterns that exactly superimpose in reciprocal space. Therefore, multiple distinct reflections (e.g. one from each of the two twin domains) fall precisely on top of each other (Fig.[link]). Because warning signs are usually not present during inspection of diffraction patterns, twinning by merohedry must be detected after data collection, and special steps must be taken during structure determination if unaffected crystals cannot be grown.

Table| top | pdf |
Symmetries in which twinning by merohedry can occur in biological macromolecules

Point groupSpace groupsApparent point group when α = 1/2
4 P4 P41 P42 P43 I4 I41 422
3 P3 P31 P32 R3 6 or 321 or 312
312 P312 P3112 P3212 622
321 P321 P3121 P3221 622
6 P6 P61 P62 P63 P64 P65 622
23 P23 P213 I23 I213 F23 432
Excluding cases of reticular twinning, R3 can only be twinned towards R32.
Point group 3 gives apparent symmetry 622 under perfect tetartohedral symmetry (α1–4 = 1/4).

Figure | top | pdf |

A cartoon depicting partial merohedral twinning in space group P4. (a) Two twin domains growing together, related by a twofold twin operation perpendicular to the fourfold symmetry axis. (b) The diffraction patterns of individual domains and their overlapping combination expected in a diffraction experiment. (Adapted from Yeates & Fam, 1999[link].)

The case of space group P4 described above is an example of the simplest type of twinning by merohedry, namely where only two distinct domain orientations are possible. The case of two domain orientations (n = 2) is referred to as twinning by hemihedry. We focus mainly on hemihedry in this chapter but provide brief comments on higher-order cases, such as tetartohedry (n = 4), at the end. Another distinct category of twinning arises in crystals where the lattice very nearly obeys some higher rotational symmetry, not by virtue of a strict requirement, as in P4 and its tetragonal lattice, but because of a fortuitous geometry of the unit cell. These cases are referred to as twinning by pseudo-merohedry. A simple example is provided by a monoclinic crystal with the unique β angle very nearly equal to 90°. The lattice is nearly orthorhombic, in which it is possible to have two domain orientations related by a twofold rotation about one of the principal axes perpendicular to the crystallographic axis of twofold symmetry. Situations of twinning by pseudo-merohedry are relatively common (Rudolph et al., 2004[link]; Hamdane et al., 2009[link]). The method of treatment is similar to those of typical merohedry, but detection requires special attention since the phenomenon depends on specific unit-cell geometries. Modern software programs are able to detect cases where twinning by pseudo-merohedry is possible by identifying potentially higher metric symmetry within an apparently lower-symmetry unit cell (Adams et al., 2002[link], 2004[link]; Zwart et al., 2005[link]b). Yet another special case is reticular twinning, wherein only a subset of reflections are superimposed on others by the twinning operation while the remaining reflections are not overlapped (Herbst-Irmer & Sheldrick, 2002[link]). The simplest case is space group R3, twinned about a twofold axis along c (for example); reflections for which l = 3n overlap with other reflections in the same layer under the twin operation, while reflections in the other layers do not overlap. To date, reticular twinning is not well documented in macromolecular crystals and is not covered further here.

11.7.3. Considerations of length scale and effects in reciprocal space

| top | pdf |

Twinning is distinguished from other types of disorder by the length scale over which the disorder occurs. In cases of twinning, individual twin domains behave as ideal crystals whose sizes are presumed to be large compared with the coherence length of the X-ray beam. Therefore, the scattering from twinned specimens behaves as a sum of intensities from the separate twin domains, rather than as a sum of complex-valued structure factors; the X-ray scattering from one domain is presumed not to interfere with the scattering from the other domain(s). The organization of differently oriented domains in twinned crystal specimens has been studied in select cases (Efremov et al., 2004[link]; Ko et al., 2001[link]), but is generally not well understood for macromolecular crystals. Based on the behaviour of twinned crystals of small molecules, various types of physical arrangements are likely.

11.7.4. Extent of twinning: the twin fraction

| top | pdf |

The twin fraction describes the fractional volumes of the different domain orientations within a crystal specimen. In the case of twinning by hemihedry, just one twin fraction, α, needs to be specified. Its value is taken to be the minor component, meaning a value less than or equal to 1/2. The twin fraction dictates the relative weights with which two twin-related reflections contribute to each observed intensity. The basic system of equations is [\eqalignno{I_{\rm obs}({\bf h}_1) &= (1 - \alpha) I({\bf h}_1) + \alpha I({\bf h}_2), \cr I_{\rm obs} ({\bf h}_2) &= \alpha I({\bf h}_1) + (1 - \alpha) I({\bf h}_2), & (}]where h1 and h2 are twin-related reflections, I represents the true crystallographic intensity of a reflection (i.e. the square of the structure-factor magnitude) and Iobs represents an observed intensity (Fig.[link]).

The value of the twin fraction has a critical effect on how twinning manifests itself in practice. Two essentially distinct scenarios arise. First, if the twin fraction is significantly less than 1/2, a situation typically referred to as `partial twinning' occurs, and partial higher symmetry may be evident during data processing and analysis. For example, if the true space group is P4 and the crystal specimen is partially twinned, one might find that the merging R value is lower than random when attempts are made to reduce the data under the higher tetragonal symmetry P422. Similarly, when the data set reduced in P4 is analysed using a self-rotation function, partial higher symmetry may be evident along what would be symmetry axes in P422. However, such symptoms can also be consistent with an alternative explanation, namely local or noncrystallographic symmetry (NCS) nearly coincident with potentially higher crystallographic symmetry (P422 in the present example). Discriminating partial twinning from pseudo-crystallographic NCS is therefore a key problem in analysing data in space groups that support twinning.

In the second scenario, the twin fraction α is very nearly 1/2, giving rise to a situation typically referred to as `perfect twinning'. In that case, pairs of twin-related observations are necessarily equal. The recorded data set therefore obeys higher symmetry. For example, if a crystal in P4 is perfectly twinned by merohedry, the data set will obey 422 symmetry. The challenge in a case of perfect twinning is to recognize that the correct space-group symmetry is lower than that implied by the observed intensity data.

11.7.5. Indications of twinning

| top | pdf |

Unusual crystal morphology sometimes provides an early suggestion of twinning. In particular, concave crystals or crystals having `re-entrant' faces are suspect. However, most cases of twinning in macromolecules are not identified until structure determination is in progress. It is relatively rare, but in some cases twinning can be identified decisively when an impossibly dense molecular packing would be required to generate a crystal with symmetry as high as that implied by the observed data (Redinbo & Yeates, 1993[link]). Sometimes, the possibility of twinning is raised during atomic refinement when, for unexplained reasons, the refinement R values cannot be brought down to acceptable limits. However, it should be emphasized that problematic refinement is never sufficient evidence by itself to conclude that a crystal is twinned; instead, a careful analysis of the observed data is required. Modern data-reduction programs now incorporate various statistical analyses for this purpose, making it relatively straightforward to identify the presence of twinning before proceeding further with structure determination.

Twinning has two related but distinct effects on intensity data. First, the observed intensities of twin-related reflections tend to be more similar to each other than expected for independent reflections; in the limit of perfect twinning they become equal. Second, the overall statistical distribution of intensities is affected, so that classic Wilson statistics are not obeyed. Therefore, two essentially different kinds of tests exist to analyse intensity statistics for the presence of twinning.

11.7.6. Twinning tests based on overall intensity statistics

| top | pdf |

Twinning causes a shift in the typical intensity distribution from an exponential to a sigmoidal shape (Fig.[link]a). Various measures of the intensity distribution have been used to diagnose twinning. The Wilson ratio, <F>2/<F2> (where the angle brackets denote the average), is 0.785 for ordinary acentric data and 0.885 for perfectly twinned data (Stanley, 1972[link]; Wilson, 1949[link]). Similarly, the value of <|E2 − 1|> (where E2 is defined as F2/<F2>) is 0.736 for ordinary acentric data and 0.541 for perfectly twinned data (Bruker, 2000[link]). The moments of the intensity distribution have simple, rational values. For example, the ratio <I2>/<I>2, sometimes referred to as the Stanley factor, is 2.0 for ordinary acentric data and 1.5 for perfectly twinned data (Stanley, 1972[link]). For partial twinning, intermediate behaviour is observed and complex expressions for the resulting intensity distributions have been worked out (Rees, 1980[link]). A key difficulty with these statistical tests is that they assume that the members of the set of reflections being evaluated all have the same expected value. If the diffraction is nearly isotropic, this condition can be satisfied relatively easily by normalizing the reflection intensities in thin resolution shells. In contrast, anisotropy presents a problem that must be corrected if these traditional tests are to be useful. If the anisotropy has a relatively simple form (e.g. if it follows an elliptical Gaussian function), then correction may be straightforward, but otherwise correction can remain problematic. Crystals in which distinct molecules in the asymmetric unit are related by translational shifts create similar problems: classes of strong and weak reflections are present in a single resolution shell. Cases of anisotropy and translational pseudo-symmetry both give intensity distributions that are shifted counter to the shift caused by twinning, and can therefore mask its presence.


Figure | top | pdf |

Detection of high or perfect twinning by analysing overall intensity statistics. (a) A traditional examination of cumulative intensity statistics. z is the normalized intensity and N(z) is its cumulative distribution. Theoretical curves are shown for perfectly twinned (dashed line) and untwinned acentric data (solid line). The experimental curve (open circles) is for the protein Dicer [Protein Data Bank (PDB; Berman et al., 2000[link]) code 2qvw; MacRae & Doudna, 2007[link]], which is essentially perfectly twinned. (b) A plot of the local intensity difference, L (see text). Theoretical curves for perfectly twinned (dashed line) and untwinned (solid line) data are shown. Data from a perfectly twinned crystal, VP5CT (Padilla & Yeates, 2003[link]), are plotted with open circles. The crystal suffered from pseudo-centring, which did not interfere with the interpretation of twinning using local intensity statistics; reflection pairs were chosen to have indices differing by even numbers.

One method for examining overall intensity statistics in a way that seeks to overcome the challenges described above has been described (Padilla & Yeates, 2003[link]). Reflections near each other in reciprocal space are considered in pairs. The difference between the two intensities divided by their sum is taken as the quantity to be analysed, [L \equiv {{(I_{{\rm obs}, A} - I_{{\rm obs}, B})} \over {(I_{{\rm obs}, A} + I_{{\rm obs}, B})}}, \eqno (]where L is the so-called local intensity difference, and Iobs,A and Iobs,B are the observed intensities of two reflections that are not related by any crystal symmetry or potential twin operation. Choosing the reflection pairs to be near in reciprocal space ensures that anisotropy effects will not affect their expected values differently. Countering translational pseudo-symmetry is a more difficult problem, but common cases of pseudo-centring or unit-cell doubling can be treated effectively by requiring that reflection pairs have the same parity (i.e. differ by an even value for all three indices). In principle, more complex types of pseudo-symmetry can be handled with additional care (Zwart et al., 2005[link]a).

For two ordinary (untwinned) acentric intensities following Wilson statistics, |L| has a uniform probability distribution from 0 to 1. For perfectly twinned intensities, the distribution of L has a quadratic form. An observed data set can therefore be analysed and the cumulative distribution of |L| compared with simple theoretical curves (Fig.[link]b). Likewise, the expected values and moments of L are simple fractions for both untwinned and perfectly twinned cases (Table[link]). Values for <|L|> are 1/2 and 3/8 for untwinned and perfectly twinned data, respectively.

Table| top | pdf |
Twinning statistics for acentric intensity data from a crystal specimen with different numbers of twin domain orientations (n)

Values are given for crystals exhibiting no twinning (n = 1), perfect twinning by hemihedry (n = 2) and perfect twinning by tetartohedry (n = 4).

1 1/2 1/3 2
2 3/8 1/5 3/2
4 35/128 1/9 5/4

When applying tests of overall intensity statistics, situations of partial twinning produce values intermediate between those expected for untwinned and perfectly twinned data. The twinning fraction can therefore be estimated by such tests. However, owing to systematic deviations of the types noted above, this is generally not the best way to estimate α. In particular, if α is small the presence of partial twinning can be missed. The strength of these tests is in identifying cases of high or perfect twinning. Consequently, an examination of overall intensity statistics is sometimes referred to as a test for perfect twinning. Such tests typically provide decisive support for a claim of high or perfect twinning.

11.7.7. Tests for partial twinning based on comparison of twin-related reflections

| top | pdf |

Once model-derived intensities become available during structure determination, the presence of partial twinning can be detected reliably and α can be estimated with high precision (see Chapter 18.12[link] ). However, at the outset of structure determination, identifying partial twinning unambiguously and estimating α are more challenging problems, and numerous methods have been developed. One method, referred to as a Britton plot and popularized by Fisher & Sweet (1980[link]), is based on the behaviour of equation ([link]. Given a pair of twin-related observed intensities and a presumed value for α, the two underlying crystallographic intensities can be calculated by inverting the linear equations. If α is chosen to be higher than its correct value, this process leads to some of the crystallographic intensities having negative calculated values. The correct value of α can therefore be estimated by plotting the number of negative calculated intensities obtained as a function of α. For error-free data, the correct value for α is the highest value that gives no negative calculated intensities.

Other methods of estimating α evaluate the degree of similarity between twin-related intensities. Rees (1982[link]) analysed the behaviour of the R value when calculated as a comparison over potentially twin-related intensities. If twinning is absent, the comparison is effectively between independent intensities, resulting in a random R value, which is 1 for acentric data. In the presence of twinning, the expected R value is 1 − 2α, providing an equation for extracting α from the R value over twin-related pairs. A related method takes twin-related reflections one pair at a time and evaluates the difference divided by their sum, to give a variable designated H (Yeates, 1988[link]),[H \equiv {{\left [ I_{\rm obs} ({\bf h}_1) - I_{\rm obs} ({\bf h}_2) \right ]} \over {\left [ I_{\rm obs} ({\bf h}_1) + I_{\rm obs} ({\bf h}_2) \right ] }}, \eqno (]where Iobs(h1) and Iobs(h2) are the observed intensities of two twin-related reflections, h1 and h2. The value of |H| ranges from 0 to 1 for untwinned (acentric data) and has a uniform probability distribution. A plot of the cumulative distribution of |H| is therefore linear. In the presence of twinning, |H| takes on a restricted range from 0 to 1 − 2α but remains uniformly distri­buted. A plot of |H| therefore provides a relatively simple means of extracting α (Fig.[link]). In addition, the expected value and moments for H relate simply to α. The value of <|H|> is (1 − 2α)/2, while the value of <H2> is (1 − 2α)2/3, providing equations for determining α from measured values of <|H|> or <H2>.


Figure | top | pdf |

Estimation of the twin fraction α based on the variable H (see text). Two cases of twinning by hemihedry are shown, along with theoretical lines for different values of α. Results from diffraction data on dihaem c-type cytochrome DHC2 (PDB code 2czs; Heitmann & Einsle, 2008[link]) are shown as a dashed curve. Results from Escherichia coli elongation factor Tu (PDB code 2fx3; Heffron et al., 2006[link]) are shown as a dotted curve. α is typically estimated from the central linear region of the observed curve.

Measurement errors complicate methods for estimating α. In a Britton plot, the ideally sharp transition that indicates the correct value for α becomes more gradual, leading to uncertainty. In methods that use measures of the average difference between twin-related reflections, ignoring the effects of errors causes α to be underestimated. Equations for incorporating the effects of errors have been discussed by Rees (1982[link]), Fisher & Sweet (1980[link]), Dumas et al. (1999[link]) and Lunin et al. (2007[link]). In general, measurement errors have a relatively small effect on estimates of α when the twin fraction is low, but have a much larger effect when the intensity measurement errors (on a fractional scale) are comparable with the deviation of α from 1/2.

Local or noncrystallographic symmetry (NCS) often presents challenging complications in relation to twinning. If a local symmetry operator nearly coincides with a crystallographic operator of some higher crystal symmetry, this gives rise to higher pseudo-crystallographic symmetry. Especially at lower resolution, reflections related by this operation will have similar intensities, and this can be mistaken for partial twinning. A further complication is that bona fide twinning often occurs in conjunction with NCS. One approach for resolving such complications is to analyse intensity statistics as a function of resolution. In cases where NCS is present but twinning is not, the similarity between potentially twin-related reflections typically decreases at higher resolution. In contrast, similarity between twin-related reflections persists out to the maximum resolution if twinning is present.

NCS can also occur in conjunction with perfect twinning. As before, resolution-dependent behaviour can be informative. Tests of overall intensity statistics (i.e. tests of perfect twinning) described earlier can appear relatively normal at low resolution where, owing to NCS, the true crystallographic intensities of twin-related reflections may be very similar to each other. If the true crystallographic intensities of twin-related reflections are nearly equal, twinning amounts to an averaging of values that are already nearly equal, which has little effect on intensity statistics. The effects of twinning on intensity statistics may be prominent only at higher resolution, where the true crystallographic intensities of twin-related reflections have diverged.

11.7.8. Higher forms of twinning

| top | pdf |

Types of twinning beyond hemihedry are possible. When four distinct twin domain orientations are involved, the situation is referred to as tetartohedry. For macromolecular crystals, twinning by tetartohedry is possible when the true symmetry of the crystal is primitive trigonal, e.g. P3. The crystal lattice has 622 rotational symmetry (order 12), which exceeds the rotational symmetry of P3 by a factor of four. Multiple cases have been reported (Rosendal et al., 2004[link]; Barends et al., 2005[link]; Gayathri et al., 2007[link]; Yu et al., 2009[link]). Twinning by pseudo-tetartohedry is possible in symmetries other than P3 if the unit-cell geometry obeys fortuitously high symmetry (Anand et al., 2007[link]). In twinning by tetartohedry, four crystallographic intensities contribute to each observed intensity. The effects on intensity statistics can therefore be even more pronounced compared with twinning by hemihedry. This is especially true when all four domain orientations are equally represented (i.e. the four underlying twin fractions are all 1/4), a situation referred to as perfect tetartohedry. In that case, the ratio <I2>/<I>2 is expected to be 5/4 (Stanley, 1972[link]) and the value of <|L|> is expected to be 35/128 ≃ 0.273 (Table[link]). Less extreme values are expected for partial tetartohedry.

Twinning by partial tetartohedry can manifest itself as a crystal that appears to be partially twinned according to multiple distinct twin laws. However, cases of tetartohedry cannot be described properly in terms of a set of estimated hemihedral twin fractions (e.g. one for each twin law). Whereas the twin fraction α (and its complement 1 − α) can be estimated in the case of hemihedry by comparing pairs of twin-related reflections, the four twin fractions (which sum to unity) in the case of tetartohedry must be estimated by performing statistical tests on quadruplets of twin-related reflections. Equations and methods for that purpose are provided by Yeates & Yu (2008[link]).

11.7.9. Other kinds of disorder

| top | pdf |

Macromolecular crystals can suffer from disorders other than twinning. These are not covered in depth here, but similarities and differences between twinning and a class of problems known as order–disorder (OD) phenomena are worth noting. In OD phenomena, alternative molecular positions or orientations occur over a short range within a crystal, from one layer to another for instance, in a stochastic fashion (Dornberger-Schiff, 1956[link]; Dornberger-Schiff & Grell-Niemann, 1961[link]). When the stochastic variation between layers or between rows is a translation, this is referred to as lattice-translocation disorder (LTD). Several cases of OD have been described and most have been of the LTD type (Wang, Kamtekar et al., 2005[link]; Wang, Rho et al., 2005[link]; Hwang et al., 2006[link]; Tanaka et al., 2008[link]; Zhu et al., 2008[link]; Hare et al., 2009[link]; Tsai et al., 2009[link]), while cases involving alternately oriented layers are known, but less common (Pletnev et al., 2009[link]). Similar to twinning, OD phenomena can complicate structure determination and prevent proper atomic refinement. However, the symptoms and treatments differ from those for twinning. A distinguishing feature is the length scale of the disorder. In OD cases, differently oriented or translated molecules are nearby in the crystal and therefore scatter coherently. The intensity statistics are therefore not shifted in the same way as for twinning. Furthermore, the interference between alternatively positioned molecules in the case of LTD is typically evident by major non-origin peaks in a native Patterson map, often appearing to be impossibly close to the origin. In addition, LTD diffraction patterns often show prominent streaking in all or a subset of reflections. Another situation that often arises in OD cases is the appearance of sterically impossible (e.g. interpenetrating) electron density during model building. This can arise when the model accounts for some of the molecules in the unit cell, but others (typically in alternative, mutually exclusive, orientations or positions) have not been accounted for. In contrast, unexplained or interpenetrating electron density is generally not observed in twinning, owing to the absence of coherent scattering between differently oriented twin domains.

11.7.10. Summary

| top | pdf |

In situations where symmetry permits it, twinning is fairly common in macromolecular crystals. If twinning is present but goes undetected, problems arise in later stages of structure determination. Partial and perfect twinning lead to distinct scenarios, but statistical tests make it possible to uncover both kinds of problem. Local or noncrystallographic symmetry, which very often occurs together with twinning, can complicate the interpretation. Resolution-dependent analysis can help to clarify these cases. Additional challenges occur with more complex types of twinning, such as tetartohedry, which is being recognized with increasing frequency.


The authors thank Michael Sawaya, Regine Herbst-Irmer and George Sheldrick for helpful comments on the manuscript.


Adams, P. D., Gopal, K., Grosse-Kunstleve, R. W., Hung, L.-W., Ioerger, T. R., McCoy, A. J., Moriarty, N. W., Pai, R. K., Read, R. J., Romo, T. D., Sacchettini, J. C., Sauter, N. K., Storoni, L. C. & Terwilliger, T. C. (2004). Recent developments in the PHENIX software for automated crystallographic structure determination. J. Synchrotron Rad. 11, 53–55.
Adams, P. D., Grosse-Kunstleve, R. W., Hung, L.-W., Ioerger, T. R., McCoy, A. J., Moriarty, N. W., Read, R. J., Sacchettini, J. C., Sauter, N. K. & Terwilliger, T. C. (2002). PHENIX: building new software for automated crystallographic structure determination. Acta Cryst. D58, 1948–1954.
Anand, K., Schulte, A., Fujinaga, K., Scheffzek, K. & Geyer, M. (2007). Cyclin box structure of the P-TEFb subunit cyclin T1 derived from a fusion complex with EIAV tat. J. Mol. Biol. 370, 826–836.
Barends, T. R. M., de Jong, R. M., van Straaten, K. E., Thunnissen, A.-M. W. H. & Dijkstra, B. W. (2005). Escherichia coli MltA: MAD phasing and refinement of a tetartohedrally twinned protein crystal structure. Acta Cryst. D61, 613–621.
Berman, H. M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T. N., Weissig, H., Shindyalov, I. N. & Bourne, P. E. (2000). The Protein Data Bank. Nucleic Acids Res. 28, 235–242.
Bruker (2000). XPREP Software Reference Manual. Bruker AXS Inc., Madison, Wisconsin, USA.
Dauter, Z. (2003). Twinned crystals and anomalous phasing. Acta Cryst. D59, 2004–2016.
Dix, I., Sevvana, M., Herbst-Irmer, R. & Sheldrick, G. M. (2007). Processing and scaling algorithms for small molecules, macromolecules and twins. Acta Cryst. A63, s162.
Dornberger-Schiff, K. (1956). On order–disorder structures (OD structures). Acta Cryst. 9, 593–601.
Dornberger-Schiff, K. & Grell-Niemann, H. (1961). On the theory of order–disorder (OD) structures. Acta Cryst. 14, 167–177.
Duisenberg, A. J. M. (1992). Indexing in single-crystal diffractometry with an obstinate list of reflections. J. Appl. Cryst. 25, 92–96.
Duisenberg, A. J. M., Kroon-Batenburg, L. M. J. & Schreurs, A. M. M. (2003). An intensity evaluation method: EVAL-14. J. Appl. Cryst. 36, 220–229.
Dumas, P., Ennifar, E. & Walter, P. (1999). Detection and treatment of twinning: an improvement and new results. Acta Cryst. D55, 1179–1187.
Efremov, R., Moukhametzianov, R., Büldt, G. & Gordeliy, V. (2004). Physical detwinning of hemihedrally twinned hexagonal crystals of bacteriorhodopsin. Biophys. J. 87, 3608–3613.
Fisher, R. G. & Sweet, R. M. (1980). Treatment of diffraction data from crystals twinned by merohedry. Acta Cryst. A36, 755–760.
Friedel, G. (1926). Leçons de Cristallographie. Paris: Berger-Levrault.
Gayathri, P., Banerjee, M., Vijayalakshmi, A., Azeez, S., Balaram, H., Balaram, P. & Murthy, M. R. N. (2007). Structure of triosephosphate isomerase (TIM) from Methanocaldococcus jannaschii. Acta Cryst. D63, 206–220.
Hamdane, D., Lechauve, C., Marden, M. C. & Golinelli-Pimpaneau, B. (2009). Pseudo-merohedral twinning in monoclinic crystals of wild-type human brain neuroglobin. Acta Cryst. D65, 388–392.
Hare, S., Cherepanov, P. & Wang, J. (2009). Application of general formulas for the correction of a lattice-translocation defect in crystals of a lentiviral integrase in complex with LEDGF. Acta Cryst. D65, 966–973.
Heffron, S. E., Moeller, R. & Jurnak, F. (2006). Solving the structure of Escherichia coli elongation factor Tu using a twinned data set. Acta Cryst. D62, 433–438.
Heitmann, D. & Einsle, O. (2008). Pseudo-merohedral twinning in crystals of the dihaem c-type cytochrome DHC2 from Geobacter sulfurreducens. Acta Cryst. D64, 993–999.
Helliwell, J. R. (2008). Macromolecular crystal twinning, lattice disorders and multiple crystals. Crystallogr. Rev. 14, 189–250.
Herbst-Irmer, R. & Sheldrick, G. M. (1998). Refinement of twinned structures with SHELXL97. Acta Cryst. B54, 443–449.
Herbst-Irmer, R. & Sheldrick, G. M. (2002). Refinement of obverse/reverse twins. Acta Cryst. B58, 477–481.
Hwang, W. C., Lin, Y., Santelli, E., Sui, J., Jaroszewski, L., Stec, B., Farzan, M., Marasco, W. A. & Liddington, R. C. (2006). Structural basis of neutralization by a human anti-severe acute respiratory syndrome spike protein antibody, 80R. J. Biol. Chem. 281, 34610–34616.
Ko, T.-P., Kuznetsov, Y. G., Malkin, A. J., Day, J. & McPherson, A. (2001). X-ray diffraction and atomic force microscopy analysis of twinned crystals: rhombohedral canavalin. Acta Cryst. D57, 829–839.
Lunin, V. Y., Lunina, N. L. & Baumstark, M. W. (2007). Estimates of the twinning fraction for macromolecular crystals using statistical models accounting for experimental errors. Acta Cryst. D63, 1129–1138.
MacRae, I. J. & Doudna, J. A. (2007). An unusual case of pseudo-merohedral twinning in orthorhombic crystals of Dicer. Acta Cryst. D63, 993–999.
Padilla, J. E. & Yeates, T. O. (2003). A statistic for local intensity differences: robustness to anisotropy and pseudo-centering and utility for detecting twinning. Acta Cryst. D59, 1124–1130.
Parsons, S. (2003). Introduction to twinning. Acta Cryst. D59, 1995–2003.
Pletnev, S., Morozova, K. S., Verkhusha, V. V. & Dauter, Z. (2009). Rotational order–disorder structure of fluorescent protein FP480. Acta Cryst. D65, 906–912.
Redinbo, M. R. & Yeates, T. O. (1993). Structure determination of plastocyanin from a specimen with a hemihedral twinning fraction of one-half. Acta Cryst. D49, 375–380.
Rees, D. C. (1980). The influence of twinning by merohedry on intensity statistics. Acta Cryst. A36, 578–581.
Rees, D. C. (1982). A general theory of X-ray intensity statistics for twins by merohedry. Acta Cryst. A38, 201–207.
Rosendal, K. R., Sinning, I. & Wild, K. (2004). Crystallization of the crenarchaeal SRP core. Acta Cryst. D60, 140–143.
Rudolph, M. G., Wingren, C., Crowley, M. P., Chien, Y. & Wilson, I. A. (2004). Combined pseudo-merohedral twinning, non-crystallographic symmetry and pseudo-translation in a monoclinic crystal form of the T-cell ligand T10. Acta Cryst. D60, 656–664.
Sheldrick, G. M. (1997). SHELX. University of Göttingen, Germany.
Sheldrick, G. M. (2002). TWINABS. University of Göttingen, Germany.
Sheldrick, G. M. (2004). CELL_NOW. University of Göttingen, Germany.
Sheldrick, G. M. (2008). A short history of SHELX. Acta Cryst. A64, 112–122.
Sparks, R. A. (1999). GEMINI. Bruker AXS, Madison, Wisconsin, USA.
Stanley, E. (1972). The identification of twins from intensity statistics. J. Appl. Cryst. 5, 191–194.
Tanaka, S., Kerfeld, C. A., Sawaya, M. R., Cai, F., Heinhorst, S., Cannon, G. C. & Yeates, T. O. (2008). Atomic-level models of the bacterial carboxysome shell. Science, 319, 1083–1086.
Tsai, Y., Sawaya, M. R. & Yeates, T. O. (2009). Analysis of lattice-translocation disorder in the layered hexagonal structure of carboxysome shell protein CsoS1C. Acta Cryst. D65, 980–988.
Wang, J., Kamtekar, S., Berman, A. J. & Steitz, T. A. (2005). Correction of X-ray intensities from single crystals containing lattice-translocation defects. Acta Cryst. D61, 67–74.
Wang, J., Rho, S.-H., Park, H. H. & Eom, S. H. (2005). Correction of X-ray intensities from an HslV-HslU co-crystal containing lattice-translocation defects. Acta Cryst. D61, 932–941.
Wilson, A. J. C. (1949). The probability distribution of X-ray intensities. Acta Cryst. 2, 318–321.
Yeates, T. O. (1988). Simple statistics for intensity data from twinned specimens. Acta Cryst. A44, 142–144.
Yeates, T. O. (1997). Detecting and overcoming crystal twinning. Methods Enzymol. 276, 344–358.
Yeates, T. O. & Fam, B. C. (1999). Protein crystals and their evil twins. Structure, 7, R25–R29.
Yeates, T. O. & Yu, F. (2008). Equations for determining tetartohedral twin fractions. Acta Cryst. D64, 1158–1164.
Yu, F., Song, A., Xu, C., Sun, L., Li, J., Tang, L., Yu, M., Yeates, T. O., Hu, H. & He, J. (2009). Determining the DUF55-domain structure of human thymocyte nuclear protein 1 from crystals partially twinned by tetartohedry. Acta Cryst. D65, 212–219.
Zhu, X., Xu, X. & Wilson, I. A. (2008). Structure determination of the 1918 H1N1 neuraminidase from a crystal with lattice-translocation defects. Acta Cryst. D64, 843–850.
Zwart, P. H., Grosse-Kunsteleve, R. W. & Adams, P. D. (2005a). Characterization of X-ray data sets. CCP4 Newslett. 42, contribution 10.
Zwart, P. H., Grosse-Kunsteleve, R. W. & Adams, P. D. (2005b). Xtriage and Fest: automatic assessment of X-ray data and substructure structure factor estimation. CCP4 Newslett. 43, contribution 7.

to end of page
to top of page