Autoantibodies to alpha-synuclein in inherited Parkinson’s disease

Abstract

Neurodegeneration in Parkinson’s disease (PD) is accompanied by a local immune reaction in the affected brain regions. It is well established that α-synuclein is directly implicated in the pathogenesis of PD. Development of the disease is often associated with changes of expression and cellular compartmentalisation of this protein; moreover, its oligomers or protofibrils are often released to the CSF and plasma of patients. Aggregated α-synuclein can trigger the activation of microglia; however, its capacity to induce production of specific autoantibodies (AAb) has not been assessed. In this study, we examined the presence of AAb against synuclein family members in the peripheral blood serum of PD patients and control individuals. Presence of AAb against β-synuclein or γ-synuclein showed no association with PD. Multi-epitopic AAb against α-synuclein were detected in 65% of all patients tested and their presence strongly correlated with an inherited mode of the disease but not with other disease-related factors. The frequency of the presence of AAb in the studied group of patients with sporadic form of PD was not significantly different from the frequency in the control group but very high proportion (90%) of patients with familial form of the disease were positive for AAb against α-synuclein. We hypothesise that these AAb could be involved in pathogenesis of the inherited form of PD.

The principal pathological feature of Parkinson’s disease (PD) is a massive death of dopaminergic neurons. With the exception of a few hereditary and toxin-induced forms, the aetiology and mechanism of the development of this pathology are not known. Recently, immunological components of the disease came to the limelight, because local inflammatory and immune reactions in PD patients were identified as factors that exacerbate the neurodegenerative processes in the affected regions of the brain. Pro-inflammatory events that might trigger these reactions, like head trauma and intrauterine foetal exposure to certain viruses, are considered to be pre-disposing factors to the disease (Ringheim and Conant 2004), and epidemiological studies revealed that chronic use of non-steroidal anti-inflammatory drugs may reduce the risk of PD by about 45% (Chen et al. 2003).

Local inflammation and immune response in PD involve both CNS resident cells (mainly activated microglia) and infiltrating peripheral monocytes and leukocytes (T-cells and B-cells). A profound microglia proliferation is found in the substantia nigra (SN) and striatum of PD patients as well as in the basal ganglia of the experimental animal models of PD (Hunot and Hirsch 2003). Activated microglia could develop into functional antigen presenting cells, activate the classical complement pathway, induce opsonisation of damaged neurons and promote chemotaxis (Wucherpfennig 1994). Possible consequences of this initial immune activation in the affected region of PD brains are the local permeabilisation of blood–brain barrier. These could lead to infiltration of the affected regions by B- and/or T-lymphocytes, which is believed to be a critical step for the development of autoimmune reactions (Racke et al. 2000).

Recent studies linked the PD-associated immune reaction with the dysfunction of α-synuclein, an abundant neuronal protein with developmentally regulated pattern of expression in many neuronal populations, including the dopaminergic neurons of the SN. The normal function(s) of α-synuclein is not known but the ability of over-expressed α-synuclein to functionally compensate for the loss of certain chaperone complexes involved in the regulation of neurotransmitter release in neuronal synapses (Chandra et al. 2005) is consistent with the hypothesis that synucleins are scaffold/adaptor proteins that enhance efficiency of various presynaptic processes (Papachroni et al. 2005). The implication of α-synuclein in PD is well documented and supported by observations that (i) aggregates of fibrillated α-synuclein are the main constituents of Lewy bodies, (ii) certain mis-sense mutations, as well as duplication or triplication of the α-synuclein gene, cause autosomal dominant PD and (iii) the principal molecular, cellular and pathophysiological aspects of PD can be modelled by expression of α-synuclein in neuronal cell lines or transgenic organisms (reviewed in Maries et al. 2003; Forman et al. 2005). Certain intermediates of α-synuclein aggregation are believed to be toxic to neurons, particularly to dopaminergic neurons (Lundvig et al. 2005). Death of neurons or damage of axons and synaptic terminals could result in the release of various soluble or aggregated forms of α-synuclein into the extracellular space and, in accord with this scenario, monomeric and oligomeric forms of α-synuclein have been found in the CSF and plasma of PD patients (El-Agnaf et al. 2006). Experiments in primary mesencephalic neuron–glia cultures have demonstrated that the extracellular aggregated form of human α-synuclein induces microglia activation, which aggravates the degeneration of neurons (Zhang et al. 2005). One of the effects of activated microglia on co-cultured dopaminergic neurons is nitration of α-synuclein, which is believed to be a crucial step that triggers death of these neurons (Shavali et al. 2006). These data suggest that α-synuclein is involved in the initial steps of the local immune response associated with PD. However, its involvement in further steps, particularly in the induction of autoimmune reactions, has never been studied. In this study, we assessed the presence of autoantibodies (AAb) against all three synucleins in the peripheral blood serum of PD patients and control healthy individuals.

Materials and methods

Patients and healthy controls

The study sample consisted of three groups of subjects – patients with sporadic (n = 31) or familial (n = 20) PD (fPD) (total 32 men, 19 women; mean age 65.22 ± 12.08 years) and a control group included 26 healthy individuals (16 men, 10 women; mean age 64.9 ± 10.9 years) with matching gender, age and ethnic characteristics and with no history of neurological illness. Demographics data of patients and control individuals are shown in Table 1. All patients were residents of Thessaly (Central Greece) and were recruited from the outpatient clinic for movement disorders in the University Hospital of Larissa and followed up for at least 1 year. Mutations in autosomal dominant PD genes (α-synuclein, ubiquitin carboxylterminal hydrolase L1 (UCHL1) and leucine rich repeat kinase 2 (LRRK2)) were excluded by direct sequencing in all fPD patients. Dosage alterations in α-synuclein gene were excluded by quantitative duplex PCR. The LRRK2 G2019S mutation that may cause sporadic PD was excluded in all sporadic PD patients. Skilled neurologists (G.M.H and A.P.) performed all clinical assessments including PD diagnosis, staging according to Hoehn and Yahr scale, age-at-onset, etc. Fifteen probands with fPD had five affected family members and five probands with fPD had six affected family members, apparently on autosomal dominant mode of inheritance based on family history and pedigree analysis (see Fig. 1 and Xiromerisiou et al. 2006).

Table 1.

Demographic data of PD patients

	Sporadic (n = 31)	Familial (n = 20)	Controls (n = 26)
Age at time of study, average, years (SD)	65.1 (11.6)	66.1 (12.7)	64.9 (10.9)
Age at onset, average, years (SD)	57.9 (13.7)	56.7 (12.5)	NA
Female, (%)	12 (38.7)	7 (35.0)	10 (38.5)
History of smoking (%)	12 (38.7)	3 (15.0)	7 (26.9)
History of traumatic brain injury, TBI (%)	5 (16.1)	3 (15.0)	5 (19.2)
History of exposure in pesticides (%)	14 (45.2)	7 (35.0)	8 (30.8)
Clinical characteristics and findings, number
Onset with tremor (%)	12 (38.7)	11 (55.0)	NA
Bradykinesia (%)	16 (51.6)	14 (70.0)	NA
Muscular rigidity (%)	15 (48.4)	9 (45.0)	NA
Postural instability (%)	17 (54.8)	10 (50.0)	NA
Persistent asymmetry of signs (%)	2 (6.5)	1 (0.05)	NA
Rest tremor (%)	23 (74.2)	17 (85.0)	NA
Hoehn and Yahn score (staging 0–5) mean (SD)	2.5 (1.3)	2.4 (1.7)	NA

Fig. 1.

Pedigrees for the patients with familiar form of PD participated in the study. Affected members of the family are shown as filled circles or squares. Arrows show studied patients. Of 20 patients with familiar form of the disease, only patients GRFPD14 and GRFPD17 did not have detectable levels of AAb against α-synuclein.

In the absence of knowing the causative gene in these families, or haplotype data around a known locus, it is premature to conclude whether or not these families are unrelated. In two families, two affected members were examined in each family, while diagnosis of other members was based on medical records or self-report of the proband. In the remaining families (n = 18), at least three affected members were examined by movement disorders’ specialists (G.M.H or A.P). Moreover, PD was excluded in most unaffected family members after clinical examination.

All PD patients were under pharmaceutical treatment. Controls were subjects living in the same geographical area, who visited other, non-neurological outpatient clinics and were free of disease (PD included). The samples of peripheral blood serum of all subjects were aliquoted and stored at −80°C. This study was approved by the institutional ethical review committees. All subjects or their families were informed of the investigational nature of the study and informed consent was obtained for their participation.

Preparation of bacterially expressed recombinant proteins

The cDNA fragments encoding full-length α-synuclein, β-synuclein and γ-synuclein, and overlapping peptides of α-synuclein were PCR amplified from corresponding plasmid templates (α-syn/pRK172 and β-syn/pRK172 and γ-syn/H1) using Pfu polymerase (Stratagene, La Jolla, CA, USA) and cloned into pGEX-4T-1 (Amersham Pharmacia Biotech, St. Albans, UK) vector in frame with glutathione-S-transferase (GST). Nucleotide sequences of all plasmid constructs were verified by DNA sequencing. Expression of recombinant proteins was induced with 1 mmol/L isopropyl-b-D-thiogalactosidase in exponentially growing Escherichia coli BL21 cells transformed with corresponding recombinant plasmids. GST-fusion proteins were purified using standard affinity purification protocol described by the manufacturer of Glutathione Sepharose 4B (Amersham Pharmacia). When required, eluted GST fusion proteins were treated with 10 units of human thrombin (Sigma, St. Louis, MO, USA) per 0.5 mg of protein at 18–21°C for 1.5 h and the GST fragment was removed by re-absorption on the fresh Glutathione Sepharose 4B beads.

Immunoblot analysis

Sodium dodecyl sulfate – polyacrylamide gel electrophoresis (SDS–PAGE) of recombinant synucleins was performed as described (Buchman et al. 1998), using 4–12% polyacrylamide gels. Separated recombinant proteins (1–2 μg per lane) were transferred to nitrocellulose membrane (ECL Hybond, Amersham Pharmacia). Following blocking with 5% non-fat dry milk in Tris-buffered saline–0.1% Tween 20 (TBST), the membranes were incubated with human serum (1 : 100 dilution in 5% milk/TBST) for 14–16 h at 4°C. The presence of AAb was revealed using secondary horseradish peroxidase-conjugated anti-human antibody (Sigma) and ECL detection system (Amersham Pharmacia). In some experiments, filters were re-probed with specific anti-synuclein antibodies. Monoclonal anti-α-synuclein (sc-32280, Santa Cruz Biotechnology), polyclonal anti-β-synuclein (sc-9565, Santa Cruz Biotechnology) and anti-γ-synuclein SK109 (Ninkina et al. 1998) antibodies were used in 1 : 1000 dilution.

Statistical analysis

Data were analysed and power analysis was performed with the statistical software SPSS v12 (SPSS Inc., Chicago, IL, USA). Two by two associations were tested using the Fisher’s exact test and the odds ratio with the corresponding 95% confidence interval. The difference in continuous variables was tested using the Mann–Whitney U-test. The concordance in positive immunosignal between any two constructs was tested using the McNemar’s test and the concordance for positive immunosignal for more than two constructs was tested using the Cochran’s Q-test. p-values >0.05 for Conhran’s Q-test and McNemar’s test are indicative of concordance. In all cases, p < 0.05 was considered statistically significant.

Results

To assess the presence of AAb against synucleins in the peripheral blood serum of PD patients and healthy control individuals, we used each diluted sera as a source of primary antibody for immunoblot detection of denatured recombinant human α-, β- and γ-synucleins. A positive signal for at least one synuclein was detected with 69% of all serum samples when they were used in a 1 : 100 dilution. The use of less diluted sera produced the same results but on significantly higher background over the whole surface of the filter. Typical examples of immunoblots probed with human peripheral blood sera are shown in Figs 2 and 3. The AAb against α-synuclein was present with significantly (p = 0.007) higher frequency in the tested PD patient samples (65%) than in the control group samples (31%). This association was due to high frequency of AAb in fPD patients (90%; p < 0.001), while there was no statistically different distribution of AAb between sporadic PD patients (48%) and controls (p = 0.17) (Table 2). The observed frequency of α-synuclein AAb presence in the control population is in accordance with other reports referring to the frequency of natural antibodies against various autoantigens in healthy subjects (Terryberry et al. 1998). Control population showed higher frequency of a positive immune reaction for β-synuclein; however, the difference from the PD group did not bear statistical significance (Table 2). Only six serum samples (two from the PD and four from the control group) reacted with γ-synuclein. The analysis of correlations between the presence of AAb against either α- or β-synuclein in the PD subjects and the age or sex of the patients, the stage of the disease, the inheritance mode and other pre-disposing factors (pesticide exposure, traumatic brain injury (TBI), smoking) demonstrated that only inheritance mode of the disease (which is considered the primary investigated parameter) significantly correlates with the presence of the autoantibody – the unadjusted odds ratio was 9.60 with 95% confidence interval (1.90–48.60) and the power to detect the observed association at 5% significance level was 91%. No association was established with the other factors (adjusted ORs are shown in Table 3). The age of onset or the disease duration also did not show correlation (two-tailed Mann–Whitney U-test) with the presence of α-synuclein AAb (p = 0.89 for the age of onset, p = 0.98 for the duration) or β-synuclein AAb (p = 0.33 for age-at-onset, p = 0.50 for duration).

Fig. 2.

Examples of immunoreactivity of sera obtained from PD patients with recombinant full-length synucleins. To confirm that protein bands detected by the PD serum GRFPD16 (a, top panel) represent α-synuclein, the membrane was re-probed with an anti-α-synuclein antibody (a, bottom panel, see also Materials and methods section). The right panels show the same membrane stained with Ponceau-S to visualize full-length recombinant synucleins (b, top panel) and probed with another PD serum GRFPD6, which has AAb against α- and β- but not γ-synuclein (b, bottom panel). Panel c shows examples of immunoblots probed with a control serum, sera from patients with sporadic form of PD (GRSPD) or sera from patients from PD families (GRFDP, designated by arrows in Fig. 1).

Fig. 3.

Example of immunoreactivity of sera obtained from a PD patient and a control healthy individual with recombinant synuclein proteins. A membrane with full-length recombinant synucleins (Ponceau-S staining is shown in the top panel) was subsequently probed with PD serum GRSPD4 (middle panel) and control serum [c_3880] (bottom panel). α-S: α-synuclein, α-S^A53T: α-synuclein with A53T mutation, β-S: β-synuclein, γ-S: γ-synuclein.

Table 2.

Incidence of α- and β-synuclein AAb in the peripheral blood serum of PD and control group. The presence of AAb was revealed by immunoblot assay with full-length recombinant proteins as antigens. The number of serum samples in each group positive for α-synuclein autoantibody (α-syn⁺), negative for α-synuclein autoantibody (α-syn⁻), positive for β-synuclein autoantibody (β-syn⁺) and negative for β-synuclein autoantibody (β-syn⁻) is shown

	α-Syn−	α-Syn+	Total	OR (95% CI)	p	β-Syn−	β-Syn+	Total	OR (95% CI)	p
Control	18	8	26	Reference	NA	13	13	26	Reference	NA
Familial PD	2	18	20	20.25 (3.77–108.83)	<0.001	15	5	20	0.33 (0.09–1.19)	0.129
Sporadic PD	16	15	31	2.11 (0.71–6.28)	0.278	22	9	31	0.41 (0.14–1.22)	0.172
Total PD	18	33	51	4.13 (1.50–11.34)	0.007	37	14	51	0.38 (0.14–1.01)	0.076

Table 3.

Correlation for the presence of α- and β-synuclein AAb in the peripheral blood serum of PD patients with predisposing (traumatic brain injury, pesticide exposure) or other factors (inheritance, disease stage, sex, age and smoking) related to the disease

	α-Synuclein		β-Synuclein
Factor	p (Fisher’s exact test)	OR (95% CI)	p (Fisher’s exact test)	OR (95%CI)
Sex	0.07	3.33 (0.99–11.12)	0.99	1.10 (0.31–3.94)
Age	0.99	0.85 (0.27–2.69)	0.99	0.85 (0.25–2.91)
Smoke	0.20	2.86 (0.68–11.92)	0.73	1.50 (0.40–5.57)
TBI	0.20	0.48 (0.11–2.22)	0.19	3.3 (0.70–15.64)
Pesticides	0.20	2.45 (0.71–8.43)	0.53	1.64 (0.48–5.68)
PD stage	0.20	2.13 (0.64-7.02)	0.99	0.88 0.26–3.10)
Inheritance	0.02	5.67 (1.37-23.46)	0.52	0.58 (0.15–2.24)

To reveal whether α-synuclein AAb in PD patients display increased specificity to particular epitopes of the protein, overlapping α-synuclein fragments fused to a GST-tag, were designed (Fig. 4a), expressed in E. coli and purified as described in Materials and methods section. For larger fragments, fusion proteins were digested with thrombin to remove the GST-tag, while fragments 3, 317 and 99 were used in GST-fusion form, because their sizes (<5 kDa) were too small for efficient SDS–PAGE analysis. The fragments were separated by SDS–PAGE, transferred to nitrocellulose membrane and incubated with diluted serum of control or PD subjects. Serum samples that were previously found positive for the presence of AAb against the full-length α-synuclein reacted with one or several α-synuclein peptides. A sample immunoblot is shown in Fig. 4b.

Fig. 4.

(a). Schematic representation of the GST-fused peptides of α-synuclein used in this study. The domain organisation of the α-synuclein molecule with the positions of the three known point mutations associated with hereditary forms of PD is shown above the scheme of the overlapping GST-fusion peptides 3, 6, 99, 305 and 317. (b). Immunoreactivity of PD serum GRSPD5 against the full-length synuclein proteins and the GST-peptides of α-synuclein. The numbers on top of the lanes correspond to the peptides shown on panel a. This serum shows immune reaction with full-length α-synuclein and α-synuclein peptide 99 but not with other α-synuclein peptides.

The association between antigenicity of each of the fragments and disease-related factors (sex, mode of inheritance, age, disease stage, smoking, TBI, pesticide exposure) was assessed and no statistically significant correlations were found (Table 4). However, combined immunoreactivity to all three peptides (N-terminal peptide 3, C-terminal peptide 99 and middle protein region peptide 317) or combined immunoreactivity for any two peptides of those three is concordant, as was shown by statistical analysis (Cochran’s Q-test analysis of antigenicity of the three peptides, p = 0.78; McNemar’s two-sided test for immunopositivity between any two peptides of the three: 3 versus 317, p = 0.75; 99 versus 3, p = 1.0; 317 versus 99, p = 0.73).

Table 4.

Associative analysis of the antigenicity of α-synuclein fragments against disease-related factors

Construct		Sex	Inheritance	Age	Smoke	TBI	Pesticides	PD stage
3	p	1.00	0.29	0.30	1.00	0.61	0.73	1.00
	OR	0.95 (0.20–4.41)	2.75 (0.66–11.54)	2.36 (0.58–9.58)	1.27 (0.30–5.33)	2.80 (0.26–30.18)	0.70 (0.18–2.77)	0.88 (0.22–3.45)
99	p	1.00	1.00	0.73	0.47	1.00	0.49	0.49
	OR	0.95 (0.20–4.41)	0.98 (0.25–3.96)	1.43 (0.36–5.66)	2.2 (0.50–9.61)	0.81 (0.10–6.58)	1.88 (0.47–7.53)	0.53 (0.13–2.14)
317	p	1.00	0.28	0.08	0.47	1.00	0.73	0.49
	OR	0.70 (0.14–3.49)	2.74 (0.64–11.75)	4.18 (0.94–18.61)	0.50 (0.12–2.13)	2.11 (0.19–22.90)	0.70 (0.17–2.85)	0.51 (0.12–2.12)
6	p	0.46	0.50	0.49	0.30	0.11	0.30	0.49
	OR	2.00 (0.40–9.91)	1.69 (0.41–6.88)	1.88 (0.47–7.53)	2.28 (0.54–9.67)	0.64 (0.13–3.17)	2.36 (0.58–9.58)	1.88 (0.47–7.53)
305	p	0.71	0.07	0.73	0.07	0.34	0.08	0.17
	OR	1.48 (0.32–6.90)	4.80 (1.07–21.45)	0.69 (0.18–2.73)	4.88 (1.01–23.57)	0.27 (0.03–2.92)	4.03 (0.95–17.22)	3.06 (0.74–12.63)

Discussion

A growing body of evidence from clinical studies and experiments with animal models emphasise a role for an immune component in the development of PD. Neurodegeneration in the brain of PD patients is accompanied by inflammatory-like response, primarily characterised by activation of microglia. Direct inhibition of microglial activation (Delgado and Ganea 2003) or therapeutic immunisation using adoptive transfer of immune cells (Benner et al. 2004) is neuroprotective in model systems. Further development of the local immune response is characterised by infiltration of the damaged regions with various types of immune cells, including B- and T-lymphocytes. Particularly interesting is the high representation in this lymphocyte population of activated (CD25+) γδ-T cells that play a key role in autoimmunity. The increased number of γδ-T cells is often seen in peripheral blood of PD patients but the elevated levels of activated (CD25+) γδ-T cells are found solely in the CSF and not peripherally, suggesting that activation of these cells is associated with degenerative processes in the CNS (Fiszer et al. 1994). Therefore, degeneration of dopaminergic neurons and associated local inflammation in the SN of PD patients sets up conditions for the development of an autoimmune reaction. Indeed, AAb against various antigens that recognise dopaminergic neurons have been found in the CSF and peripheral blood serum of patients with PD (Defazio et al. 1994; Chen et al. 1998; Rowe et al. 1998; Bas et al. 2001; Orr et al. 2005).

Here, we have demonstrated for the first time the presence of AAb against synucleins in the human peripheral blood serum. A positive correlation with PD has been found only for AAb against α-synuclein, which is consistent with the role of this protein but not two other synuclein family members in the aetiology and pathogenesis of PD. This correlation was evident only for PD patients with familial but not sporadic form of the disease, suggesting that production of AAb against α-synuclein is associated with this commonly more aggressive form of PD. We have shown that the majority of studied antisera were multi-epitopic and the epitopes were spread across the length of α-synuclein molecule. This observation confirms that not a single peptide but the whole α-synuclein molecule is autoimmunogenic and dismisses a possibility that the observed immune reaction was the result of an accidental cross-reactivity with a α-synuclein epitope of AAb produced against an antigenically similar epitope of another protein. Although the protocol used in this study (immunoblot analysis of denatured recombinant synucleins) was designed to detect only AAb against linear antigens, it is feasible that using different detection methods AAb specific to covalently modified α-synuclein species or aggregation-dependent epitopes could also be found in PD patients.

The question about functional importance of antibodies against disease-associated neuronal proteins remains wide open. It has been demonstrated that an immunoglobulin G fraction purified from serum of PD patients causes death of dopaminergic neurons in vivo following stereotaxic injection in the SN of experimental animals (Chen et al. 1998), and the presence of immunoglobulins in PD brain tissue could lead to the targeting of dopaminergic nigral neurons for destruction (Orr et al. 2005). Thus, it is feasible to suggest that AAb against α-synuclein could also be neurotoxic. However, results of other experiments suggest that antibodies against certain proteins implicated in neurodegenerative disorders could be beneficial for slowing down diseases progression. Immunisation-mediated neuroprotection following central or peripheral nervous trauma has been demonstrated (Moalem et al. 1999) and vaccination approaches have been effective in the removal of huntingtin (Luthi-Carter 2003) and β-amyloid_1–42 peptide (Janus et al. 2000) deposits in animal models. Recently, a neuroprotective role of anti-α-synuclein antibodies has been shown in a human α-synuclein transgenic mouse model of PD. Mice that produced high affinity antibodies following vaccination with human α-synuclein displayed reduced accumulation of the aggregated human α-synuclein in the neuronal cell bodies and the synapses, and decreased degeneration of neurons, probably because the antibodies recognised the aggregated human α-synuclein and promoted its degradation via lysosomal pathways (Masliah et al. 2005). It is possible that α-synuclein AAb in the serum of fPD patients are also part of a protective reaction aimed at elimination of extracellular toxic α-synuclein species or even cells bearing these species. Such cleansing could not eradicate the cause of the disease, which in these cases is obviously genetic, but could at least break the vicious cycle of events and slow the disease progression. To clarify a role for α-synuclein AAb, further longitudinal studies of larger cohorts of PD patients and healthy individuals, including healthy members of the affected families, as well as animal models of the disease are necessary. These studies should reveal whether the monitoring of the AAb titres could have a diagnostic value and whether immunisation with α-synuclein could be used as a novel approach in PD therapy.

Abbreviations used

AAb

autoantibodies

fPD

familial PD

GST

glutathione-S-transferase

PD

Parkinson’s disease

SDS–PAGE

sodium dodecyl sulfate – polyacrylamide gel electrophoresis

SN

substantia nigra

TBI

traumatic brain injury

TBST

Tris-buffered saline with 0.1% Tween 20